OPINION
RENDELL, Circuit Judge.James Dennis has spent almost twenty-four years unsuccessfully challenging his conviction for the murder of Chedell Williams. The Pennsylvania Supreme Court repeatedly affirmed Dennis’s first-degree murder conviction and sentence and denied his applications for post-conviction relief. Thereafter, Dennis filed an application under 28 U.S.C. § 2254, and the United States District Court for the Eastern District of Pennsylvania granted Dennis habeas corpus relief, concluding that the Pennsylvania Supreme Court had unreasonably applied Brady v. Maryland, 373 U.S. 83, 83 S.Ct. 1194, 10 L.Ed.2d 215 (1963), with respect to three pieces of evidence suppressed by the Commonwealth. The suppressed Brady material — a receipt corroborating Dennis’s alibi, an inconsistent statement by the Commonwealth’s key eyewitness, and documents indicating that another individual committed the murder — effectively gutted the Commonwealth’s case against Dennis. The withholding of these pieces of evidence denied Dennis a fair trial in state court. We will therefore affirm the District Court’s grant of habeas relief based on his Brady claims.
I. Background
A. Factual Background
On October 22, 1991, Chedell Williams and Zahra Howard, students at Olney High School, climbed the steps of the Fern Rock SEPTA station, located in North Philadelphia. Two men approached the girls and demanded “give me your fucking earrings.” App. 465. The girls fled down the steps; Howard ran to a nearby fruit vendor’s stand and Williams ran into the intersection at Tenth and Nedro Streets. The men followed Williams. The perpetrators tore Williams’s gold earrings from her earlobes. One of the men grabbed her, held a silver handgun to her neck, and shot her. The men then ran up the street to a waiting getaway car and fled the scene. The precise time of injury was 1:54 p.m. Emergency personnel responded within minutes, but Williams was pronounced dead at the hospital less than an hour later.
B. Police Investigation and the Trial
The police undertook an investigation into the Williams murder, primarily aimed at determining the identity of the shooter. Frank Jastrzembski led a team of detectives who pursued the investigation based on rumors that “Jimmy” Dennis from the Abbottsford Homes projects in East Falls1 committed the crime, despite being unable to identify the source of the rumors. Resting on tips by neighbors from *270the projects, police proceeded with Dennis as the primary, if not the sole, suspect.2
Detectives obtained eyewitness reports and identifications, very few of which aligned with Dennis’s appearance. Nearly all of the eyewitnesses who gave height estimates of the shooter described him as between 5'9" and 5'10". He was described as having a dark complexion and weighing about 170 to 180 pounds. The victim, Williams, had a similar build as the shooter; she was 5'10" and weighed 150 pounds. Dennis, on the other hand, is 5'5" tall and weighed between 125 and 132 pounds at the time of trial.
Prior to trial, three eyewitnesses identified Dennis in a photo array, at an in-person lineup, and at a preliminary hearing: Williams’s friend, Zahra Howard; a man working on a garage near the intersection, Thomas Bertha; and a SEPTA employee who was standing in front of the station at the time of the murder, James Cameron.3
Zahra Howard
• Photo Array: Howard identified Dennis, saying “this one looks like the guy, but I can’t be sure ... He looks a little like the guy that shot Chedell.” App. 1537. When asked if she could be sure, she replied “No.” Id.
• Lineup: Howard indicated that she “thought” Dennis was the shooter. App. 586.4
*271• Preliminary Hearing and Trial: Howard testified at trial that she had identified Dennis as the shooter at a preliminary hearing. App. 474-75. She also made an in-court identification during trial. Id.
Thomas Bertha
• Photo Array: Bertha initially said that the first photo, which was a photo of Dennis, looked like the man running with the gun and later confirmed his identification.
• Lineup: When asked to identify the shooter, Bertha simply stated “three,” which was Dennis. App. 586.
• Trial: Bertha identified Dennis as the shooter at trial.
James Cameron
• Photo Array: Cameron said that Dennis looked like the shooter, but wavered “I can’t be sure.” App. 1548.
• Lineup: Cameron identified Dennis, who was in the third position in the lineup, by simply stating “number three” without reservation. App. 689.
• Preliminary Hearing and Trial: At trial, Cameron identified Dennis as the shooter and confirmed that he had identified Dennis at the preliminary hearing.
At trial, the prosecutor introduced testimony from detectives who verified that Howard, Bertha, and Cameron each identified Dennis in the photo array and lineup. No other eyewitness identifications were referenced.
Dennis was arrested on November 22, 1991. His signed statement indicated that he stayed at his father’s house until about 1:30 p.m. on the day in question, when his father drove him to the bus stop and watched him get on the “K” bus toward Abbottsford Homes to attend singing practice that evening. Dennis rode the K bus for approximately thirty minutes to the intersection of Henry and Midvale Avenues. During the trip, Dennis saw Latanya Cason, a woman he knew from Abbotts-ford Homes. In his statement to police, which was read into the record at trial, Dennis asserted that when he and Cason disembarked the bus “[he] waved to her.” App. 710. After getting off the bus, Dennis walked to Abbottsford Homes, where he spent the rest of the day with his friends. Dennis’s father, James Murray, corroborated Dennis’s story. He stated that they spent the morning together, and that he drove Dennis to the bus stop shortly before 2:00 p.m. to catch the K bus to Ab-bottsford Homes:
The Commonwealth’s case rested primarily on eyewitness testimony, which Assistant District Attorney Roger King emphasized in his opening statement to the jury. Though ADA King acknowledged that the Commonwealth had no physical evidence — the silver handgun and the earrings were never recovered — he contended that the eyewitness identifications were sufficient for a conviction. Three eyewitnesses were called to testify at trial: Zahra Howard, Thomas Bertha, and James Cameron.
Zahra Howard, who was present with the victim at the time of the murder, led the Commonwealth’s case. She recounted what had occurred, noting that the shooter was “right in front of’ her and Williams, about one or two feet away, and that she looked the shooter in the face. App. 467-68. About ten seconds passed between the first time she saw the men until she turned around and ran away from the scene; she also saw the shooter for about five to ten seconds while he was grabbing Williams in the street. Howard identified Dennis in a photo array, at an in-person lineup, and at a preliminary hearing. Defense counsel focused his cross-examination on her hesitation in prior identifications. Howard described the shooter as wearing a black hooded sweatshirt and a red sweat suit. In *272her statement, Howard said that the shooter was about same height as Detective Danks, who was 55'9" or 5'10", or taller. Howard testified at trial that she had never seen the shooter or his accomplice before in her life.
Thomas Bertha and his partner, Anthony Overstreet, were installing stones on a garage near Tenth and Nedro Streets on the day in question. After hearing the gunshot, they came down from their ladders and looked down the street from the sidewalk. The two perpetrators ran past them. The shooter passed between three to eight feet in front of Bertha, and Bertha ran after him. Bertha made visual contact with the shooter, who was running toward him, for about three to four seconds. Defense counsel impeached Bertha by recalling that, at the preliminary hearing, Bertha testified that he could not have seen the shooter for longer than about a second. Bertha viewed the photo array and attended the lineup, identifying Dennis at both. He described the shooter as wearing red sweat pants, a red hooded sweatshirt, a black cap, and a leather jacket. Bertha testified at trial that he remembered telling the police that the shooter was 5'9" and 180 pounds.
James Cameron was working as a SEPTA cashier on the day of the murder. He was about eight to ten feet from Williams when she was shot and saw the shooter for a few seconds. Cameron saw the shooter’s face several times but acknowledged that he “didn’t really pay attention.” App. 664. He testified at trial that he saw the shooter for about thirty to forty seconds collectively. This .estimate contradicted Cameron’s prior testimony at the preliminary hearing where he claimed that about twenty seconds passed between when he first saw the shooter and when the shooter ran away. Cameron viewed the array, attended the lineup, and testified at the preliminary hearing, identifying Dennis at each instance, as well as at trial. Cameron stated that Dennis looked like the shooter, “especially from the side.” App. 676. He described the shooter as wearing a red sweat suit and a dark jacket, carrying a small silver revolver. He did not remember giving detectives a specific height and weight description, but remembered telling them that the shooter was “stocky.”5 App. 664.
Aside from eyewitness testimony, the Commonwealth presented testimony from Charles “Pop” Thompson and Latanya Ca-son, who spoke about their interactions with Dennis on October 22, 1991, the day of the murder. Thompson was in Dennis’s singing group, which held rehearsal at Ab-bottsford Homes that day. Thompson did not remember what Dennis was wearing, but told detectives that he saw Dennis with a gun that night. He also identified an illustrative .32 chrome revolver, which had been admitted as a Commonwealth exhibit, as being similar to the one he saw in Dennis’s possession. Thompson had an open drug possession charge at the time of trial, but testified that he was not expecting any help from the Commonwealth with the drug charge in exchange for his testimony. Three years after trial, Thompson attested in a statement that he had never *273seen Dennis with a gun and that his testimony at trial was false.
Latanya Cason, who knew Dennis “by living up [her] way” at Abbottsford Homes, testified that she saw him between 4:00 and.4:30 p.m. at Henry and Midvale Avenues on October 22, 1991. App. 731. Cason’s estimate that she saw Dennis between 4:00 and 4:30 p.m. was “strictly a guess” on her part — she did not know exactly what time she saw Dennis — but there was no question she saw him that day. App. 745. Prior to seeing Dennis, Cason took public transportation to the 3-2 center where she picked up her public assistance check, signing a document to confirm pick up. She then filled her daughter’s prescription, got some fish, ran a few additional errands, and went home via the K bus. Cason testified at trial that she did not see Dennis at 2:00 p.m. that day because she was just leaving work at 2:00 p.m. Although the Commonwealth introduced a schedule of payment and food stamps at trial, which stated that Cason was slated to pick up her public assistance check and food stamps on October 22, 1991, nothing was introduced at trial indi-eating the precise time of day she retrieved her benefits.
Detective Jastrzembski executed a search warrant of Dennis’s father’s home and seized two black jackets, a pair of red pants, and a pair of white sneakers. The police lost the items prior to trial. Detectives and two experts testified at trial about physical aspects of the crime, but the Commonwealth did not introduce any physical evidence at trial.6
Dennis’s defense strategy centered on his alibi, good character, and mistaken identity.7 His defense comprised of testimony by his father, James Murray, Dennis himself, a few members of his singing group, and character witnesses. Dennis did not have evidence to support an “other suspect” defense.
Dennis’s father testified that the two of them were together from the evening of October 21, 1991, until about 1:50 p.m. on October 22, 1991. Murray lives about fifteen to eighteen blocks from the Fern Rock Station, roughly a five-minute drive with traffic. Murray testified that “[he] kn[ew] for a fact that [Dennis] was on [the K bus]” at the time of Williams’s murder *274because he drove Dennis to the stop and watched from his car as Dennis boarded the bus. App. 804. The Commonwealth pointed out that Murray had visited Dennis forty times since his arrest.
Willis Meredith, James Smith, and Marc Nelson, members of Dennis’s singing group who had known Dennis for ten years or more, testified on Dennis’s behalf about rehearsal on the day of the murder. Meredith saw Dennis for about twenty minutes around 2:15 or 2:30 p.m., which aligned with Dennis’s account. Smith testified that Dennis was dressed in dark sweats and a dark hooded shirt at rehearsal that night — he was not wearing any red. Meredith, Smith, and Nelson each testified that Thompson and Dennis frequently got into arguments. Each testified that they had not seen a handgun in Dennis’s possession.8 Other defense witnesses, including Dennis’s brothers, friends, and church leaders, testified to Dennis’s reputation for being honest, truthful, peaceful, and law-abiding.9
Finally, Dennis took the stand. He testified that he had nothing to do. with Williams’s shooting and was not in the area at the time of her murder.10 In line with his father’s testimony, Dennis said he spent the previous night at his father’s house and left at 1:30 or 1:45 p.m. to take the bus to Abbottsford Homes for singing practice. When Dennis left his father’s house, he was wearing a dark blue jeans set; he changed into black sweats at Merri-weather’s house before rehearsal. Dennis testified that he took the K bus, where he “thought” he saw Tammy Cason, to Henry and Midvale Avenues in East Falls, arriving around 2:30 p.m.11 App. 1028. Dennis then went to Willis Meredith’s house for twenty to thirty minutes. Dennis acknowledged getting into frequent arguments with Thompson about Thompson’s desire to be the leader of the singing group.
Counsels’ closings reiterated the trial’s themes — eyewitness identifications and Dennis’s alibi. Defense counsel pointed to eyewitness identifications as the key question in the Commonwealth’s case, but he had no means, of impeaching Howard, the eyewitness with the closest view of the shooter. Defense counsel highlighted Thompson’s motive to lie, but Thompson’s testimony did not directly link Dennis to Williams’s murder. Finally, defense counsel had to backtrack from using Cason to bolster Dennis’s timeline due to the timing discrepancy between her version — that they saw one another between 4:00 and 4:30 — and Dennis’s account that he saw Cason at 2:30. In his closing statement to the jury, counsel urged that Dennis had not, in fact, seen Cason on the bus to detract from the inconsistency.
ADA King similarly saw Howard as the key witness at trial and instructed the jury that “if you believe Zahra Howard, that’s enough to convict James Dennis.” App. 1207. King attacked Dennis’s testimony that he saw Tammy Cason on the K bus as incredible, and undercut Dennis’s father’s testimony by urging that “blood is thicker *275than water,” leaving no disinterested witnesses to support Dennis’s account. App. 1208-09.
The jury found Dennis guilty of first-degree murder, robbery, carrying a firearm without a license, criminal conspiracy, and possession of an instrument of a crime. It found Dennis’s lack of significant criminal history a mitigating factor during the penalty phase, but it also found that the killing was committed in the course of a felony, amounting to an aggravating circumstance. The jury sentenced Dennis to death.
C. Undisclosed Evidence
The prosecution failed to disclose to Dennis’s counsel three pieces of exculpatory and impeachment evidence: (1) a receipt revealing the time that Cason had picked up her welfare benefits, several hours before the time she had testified to at trial, thus corroborating Dennis’s alibi (the “Ca-son receipt”); (2) a police activity sheet memorializing that Howard had given a previous statement inconsistent with her testimony at trial, which provided both invaluable material to discredit the Commonwealth’s key eyewitness and evidence that someone else committed the murder (the “Howard police activity sheet”); and (3) documents regarding a tip from an inmate detailing his conversation with a man other than Dennis who identified himself as the victim’s killer (the “Frazier documents”). .
1. Cason receipt
Detectives interviewed ■ Latanya Cason, the woman identified in Dennis’s initial statement, at Abbottsford Homes a few months after Dennis’s arrest. Cason told detectives that she. thought she remembered seeing Dennis the day of the murder, but her timeline contradicted the one Dennis outlined. She said that she worked until 2:00 p.m., went to the 3-2 center to pick up her public assistance check, picked up a prescription and some fish, boarded the K bus, and got off near Abbottsford Homes. According to Cason, she saw Dennis when she got off the K bus between 4:00 and 4:30. p.m., not between 2:00 and 2:30 p.m. as Dennis indicated. The only discrepancy between Dennis’s testimony and Cason’s was the time of their interaction. Police records indicate that Cason gave detectives a Department of Public Welfare (“DPW”) card marked “Schedule of check payment” at the time of her interview, which was introduced at trial. However, the Commonwealth' possessed another DPW document not disclosed at trial — a receipt bearing the time Cason picked up her check. Cason testified at trial as a witness for the prosecution and her testimony aligned with her initial statement to detectives.
On appeal, Dennis’s new appellate counsel obtained Cason’s time-stamped receipt from the DPW.12 Cason stated in an affidavit that police had a copy of the timestamped receipt when they interviewed her and that she gave police her only copy of the receipt. The receipt indicated that Cason picked up her welfare check at 13:03, or 1:03 p.m. In complete contradiction to her trial testimony, then, Cason could not have been working until 2:00 p.m. that day. Cason attributed her prior incorrect testimony to misunderstanding military time, so that she “may have thought that the 13:03, which is on the receipt, was 3:03 p.m.” App. 1736. Based *276on the discrete time indicated on the receipt, Cason’s affidavit stated she would have seen Dennis “between 2:00 and 2:30 p.m. at the Abbottsford Homes, and hot 4:00 to 4:30 that is in my statement.” Id.
2. Howard police activity sheet
Two days after the murder, detectives interviewed Diane and Mannasett Pugh, Williams’s aunt and uncle. Diane Pugh told detectives that, the night after the murder, Zahra Howard told them that she recognized the assailants from Olney High School, where she and Williams were students. Dennis did not attend Olney High School. Howard’s assertion that she recognized the assailants from school contradicted her prior statements to police that she had never seen the men before and did not recognize them from school. Police recorded in their “THINGS TO DO” list that they planned to interview Howard about her inconsistent statements.
Howard further told the Pughs that two people named “Kim” and “Quinton” had also been present at the murder. The following day, another of Williams’s aunts, Elaine Parker, told police that Howard mentioned Kim and Quinton were at the scene. The Commonwealth disclosed Parker’s statement prior to trial. However, the prosecution did not disclose information about Howard’s inconsistent statement to the Pughs. Mere hours after meeting with Parker and receiving additional information that Howard had omitted or misstated facts in her initial statement to police, two detectives met with Howard, ostensibly to follow up on their “things to do.” Ignoring their recorded intentions, however, the detectives only questioned Howard about a photo array and did not inquire about the inconsistent statements.
S. Frazier documents
Prior to Dennis’s arrest, Philadelphia detectives received a call from Montgomery County police relaying a tip from an inmate at the Montgomery County Correctional Facility, William Frazier. Frazier told Montgomery County detectives that he spoke with the man who may have murdered Williams during a three-way call with Frazier’s friend, Tony Brown, facilitated by Frazier’s aunt. During the call, Brown told Frazier and Frazier’s aunt that he “fucked up” and murdered Williams when the gun went off accidentally during a botched robbery of her earrings. App. 1692. He also said that two other men, Ricky Walker and “Skeet,” aided in committing the crime. Frazier told detectives that Brown had a brown car, that he “like[d] to wear sweat suits,” and that the men knew the victim as “Kev[’s] ... girl.”13 App. 1694-95.
Frazier told police that Brown and the others had hid in Frazier’s empty apartment for two days following the murder. Frazier provided addresses for the men, including their parents’ and girlfriends’ addresses, an address and phone number for his aunt, and an address for the pawn shop Brown frequented. Frazier volunteered to take detectives on a “ride along” to point out the houses and pawn shop.
Following the tip, Detectives Santiago and Jastrzembski interviewed Walker, who admitted to knowing Williams from Olney High School, but denied knowing Brown or Skeet. Walker denied any involvement in the murder, and claimed that his mother could verify that he was sleeping when Williams was murdered. Walker admitted to hanging out around Broad and Olney, the exact area where Overstreet said he had seen the perpetrator before. Detectives never verified Walker’s alibi nor *277showed his photo to any of the eyewitnesses. Detectives never located Brown or Skeet.
Detectives, including Jastrzembski, spoke with Frazier’s landlord, who had no knowledge of anyone entering Frazier’s apartment. Detectives did not interview Frazier’s aunt to obtain her account of the call with Brown.
The Commonwealth suppressed at least six documents relating to the Frazier tip from Dennis’s trial counsel: (1) Frazier’s initial statement to the Montgomery County police (Oct 31, 1991); (2) Frazier’s statement to the Philadelphia police (Nov. 1, 1991); (3) Police Activity Sheet regarding Frazier’s landlord (Nov. 1, 1991); (4) Police Activity Sheet regarding Ricky Walker (Nov. 2, 1991); (5) Frazier’s signed search consent; and (6) Ricky Walker’s statement (Nov. 2, 1991). The Commonwealth concedes that these documents were not disclosed to Dennis until a decade after trial.
D. Review of State Court Conviction
Like many habeas cases, this case has a lengthy procedural history. Only those decisions and arguments relevant to the instant appeal aré summarized below.
On July 22, 1998, the Pennsylvania Supreme Court affirmed Dennis’s conviction and death sentence on direct appeal by a vote of four to three. Commonwealth v. Dennis, 552 Pa. 331, 715 A.2d 404 (1998) (“Dennis /”). Dennis argued on direct appeal that the Commonwealth violated his due process rights by failing to disclose Cason’s time-stamped receipt prior to trial, in opposition to Brady v. Maryland, 373 U.S. 83, 83 S.Ct. 1194, 10 L.Ed.2d 215 (1963).14
On September 15, 1998, Dennis filed a timely pro se petition pursuant to Pennsylvania’s • Post Conviction Relief Act (“PCRA”), received new counsel, and also received discovery. In December 1999, PCRA counsel was appointed and filed an amended petition, and, subsequently,' a supplemental amended petition and a second supplemental petition on December 1, 2000, and July 10, 2002, respectively. Two pieces of evidence at issue in this appeal were disclosed during PCRA discovery. First, Dennis received the police activity sheet memorializing Howard’s statements to Diane Pugh the night after the murder, which indicated that she recognized the shooter from Olney High School. Second, Dennis received the six documents relating to the Frazier lead that police had abandoned. The PCRA court denied Dennis’s claims that the prosecution violated Brady by failing to disclose the Howard statement and the Frazier documents. Dennis again appealed to the Pennsylvania Supreme Court.
The Pennsylvania Supreme Court affirmed the PCRA court in part and vacated in part, and remanded two claims. Commonwealth v. Dennis, 597 Pa. 159, 950 A.2d 945 (2008) (‘Dennis III”). The court found that the Commonwealth’s failure to disclose the Frazier documents did not violate Brady because the prosecution was not required to disclosé “every fruitless lead” and that “inadmissible evidence cannot be the basis for a Brady violation.” Id. at 968 (internal quotation marks omitted) *278(quoting Commonwealth v. Lambert, 584 Pa. 461, 884 A.2d 848, 857 (2005)).
The Pennsylvania Supreme Court remanded to the PCRA court Dennis’s claim that the Commonwealth violated Brady by suppressing the contents of the police activity sheet memorializing Zahra Howard’s inconsistent statement. After evidentiary hearings on remand, the PCRA court again dismissed Dennis’s petition. Commonwealth v. Dennis, Case No. 92-01-0484, slip op. (Pa. Ct. Com. PL Mar. 17, 2010). The Pennsylvania Supreme Court concluded that it was not relevant that Howard denied her prior inconsistent statement at the evidentiary hearing before the PCRA court. See, e.g., Commonwealth v. Dennis, 609 Pa. 442, 17 A.3d 297, 309 (2011) (“Dennis /V”).
The Pennsylvania Supreme Court affirmed the PCRA denial on appeal. Id. It concluded that the police activity sheet was not material under Brady because “Howard was extensively cross-examined” and because “there were two eyewitnesses other than Howard who observed the shooting ■ at close range ... [and] positively identified [Dennis] as the shooter in a photo array, in a line up, and at trial.” Id.
Following the Pennsylvania Supreme Court ruling, Dennis filed a habeas corpus petition under 28 U.S.C. § 2254 in the United States District Court for the Eastern District of Pennsylvania for review of his conviction and death sentence. The District Court granted Dennis habeas relief based on Dennis’s Brady claims as to the Commonwealth’s failure to disclose the Cason receipt, the Frazier documents, and the police activity sheet containing Howard’s inconsistent statement. Dennis V, 966 F.Supp.2d at 518.
The District Court concluded that the state court’s ruling regarding the Cason receipt involved an unreasonable determination of the facts. The Pennsylvania Supreme Court had concluded that the receipt was not exculpatory because (1) “[Cason’s] testimony would not support Appellant’s alibi”; (2) it would have been cumulative of testimony by another witness; and (3) there was no evidence that the Commonwealth withheld the receipt from the defense. Dennis I, 715 A.2d at 408. The District Court determined that the receipt corroborated Dennis’s alibi, provided direct evidence that Cason’s testimony was false, and would have been strong impeachment evidence. Therefore, the state court’s determination that the receipt was not “exculpatory” was an unreasonable determination of the facts. Dennis V, 966 F.Supp.2d at 508.
The District Court also concluded the Pennsylvania Supreme Court had engaged in a similarly unreasonable determination of facts regarding whether the receipt was actually suppressed by the police. In its opinion, the Pennsylvania Supreme Court stated that the police came into possession of the receipt when interviewing Cason, and that the Commonwealth never claimed to have disclosed the receipt to defense counsel. The District Court relied on Kyles v. Whitley, 514 U.S. 419, 437, 115 S.Ct. 1555, 131 L.Ed.2d 490 (1995), for the proposition that favorable evidence in the police’s possession is imputed to the prosecution. Dennis V, 966 F.Supp.2d at 509-10. It also interpreted the three-factor balancing test in United States v. Pelullo, 399 F.3d 197 (3d Cir. 2005), to come out in favor of required disclosure by the Commonwealth. Further, the state court’s conclusion that the receipt was not material was an unreasonable application of clearly established federal law because the “receipt and Ca-son’s accompanying corrected testimony would have provided independent, disinterested corroboration of Dennis’[s] explanation for where he was at the time of Williams’[s] murder,” would have transformed Cason from a government witness into a defense witness who supported Den*279nis’s alibi, and would have provided impeachment evidence to challenge Cason’s testimony that she had worked until 2:00 p.m. that day, which otherwise could not have been challenged. Dennis V, 966 F.Supp.2d at 511.
The District Court also granted habeas relief on the basis of Dennis’s Brady claim regarding the Frazier documents, concluding that the state court had adopted an unreasonably narrow reading of Brady. The Pennsylvania Supreme Court had held that the prosecution did not violate Brady by failing to disclose the Frazier documents because Dennis did not show that the documents were admissible and material. The District Court rejected the assertion that inadmissible evidence cannot be the basis of a Brady claim, reasoning that the United States Supreme Court has never stated such a rule and that most circuit courts, including the Third Circuit, have held to the contrary. Id. at 503. Additionally, that the United States Supreme Court proceeded with the Brady analysis after acknowledging that the polygraph results at issue in Wood v. Bartholomew, 516 U.S. 1, 116 S.Ct. 7, 133 L.Ed.2d 1 (1995), were not admissible indicated to the District Court that there is no admissibility requirement for Brady evidence. Dennis V, 966 F.Supp.2d at 503.
The Pennsylvania Supreme Court had also held that the prosecution need not disclose every “fruitless lead” in order to comply with Brady. The District Court determined that this conclusion was unreasonable under Kyles. The Frazier documents contained “internal markers of credibility,” such as a description of the victim as “Kev[’s] ... girl,” which was accurate, an admission to shooting the victim in the correct location on her body, and a description of the alleged perpetrators that matched other descriptions of the shooter more closely than Dennis did. Id. at 504. The District Court reasoned that the Frazier documents would have led to further investigation that could have proved vital to the defense and could have been used to impeach the police investigation or provide a defense that another person committed the murder. Id. at 505.
Lastly, the District Court granted habe-as relief on the basis of Dennis’s claim that the Commonwealth violated Brady when it withheld the police activity sheet containing Howard’s inconsistent statements. The District Court concluded that the Pennsylvania Supreme Court had unreasonably applied Brady and its progeny in rejecting the Howard Brady claim. First, the Pennsylvania Supreme Court had unreasonably dismissed the impeachment value of the evidence and incorrectly concluded that cross-examination of Howard rendered new impeachment evidence immaterial. The District Court noted that the United States Supreme Court has directly rejected the notion that there can be no Brady claim relating to impeachment evidence where a witness was already impeached with other information. See Banks v. Dretke, 540 U.S. 668, 702, 124 S.Ct. 1256, 157 L.Ed.2d 1166 (2004) (rejecting the state’s argument that no Brady violation occurred because the witness was “heavily impeached at trial,” where the withheld evidence was the only impeachment evidence that the witness was a paid informant).15 The District Court emphasized that, although Howard was cross-examined at trial, she was not impeached. Dennis V, 966 F.Supp.2d at 514-15. Second, the District Court concluded that the Pennsylvania Supreme Court had incorrectly applied a sufficiency of the evidence test in direct contravention of Kyles’ s directive that Brady material be viewed in light of all of the *280evidence. Rather, the state court should have focused on whether the defendant received a fair trial in the absence of the disclosed evidence. Id. at 516. Finally, the District Court found it unreasonable that the state court had failed to consider the effect of the evidence on trial counsel’s investigation, pretrial preparation, decision to interview or call certain witnesses, or the effect of cross-examining detectives on their investigation into Howard. Given that the police themselves thought it was important to follow up with Howard about her possible statements to Pugh, the District Court concluded it was clear that the lead was material from an investigatory point of view. Id.
The District Court also concluded that the Pennsylvania Supreme Court had failed to undertake a cumulative materiality analysis as required by Kyles. Id. at 517-18. It did not rule on Dennis’s remaining claims. Id. at 491, 501 n.19 & 510 n.27. The Commonwealth filed a timely notice of appeal.
A panel of this Court issued an opinion on February 9, 2015. Dennis v. Sec’y, Pa. Dep’t of Corr., 777 F.3d 642 (3d Cir. 2015). This opinion was vacated and rehearing en bane was granted on May 6, 2015.
II. Jurisdiction and Standard of Review
The District Court had jurisdiction under 28 U.S.C. §§ 2241, and 2254 over Dennis’s habeas corpus petition. This Court has appellate jurisdiction under 28 U.S.C §§ 1291 and 2253. The District Court based its decision on a review of the state court record and did not conduct an evidentiary hearing, so our review of its order is plenary and we apply the same standard the District Court applied. Branch v. Sweeney, 758 F.3d 226, 232 (3d Cir. 2014); Brown v. Wenerowicz, 663 F.3d 619, 627 (3d Cir. 2011).
The Antiterrorism and Effective Death Penalty Act (AEDPA) dictates the manner in which we conduct our review. Federal habeas courts cannot grant relief “with respect to any claim that was adjudicated on the merits in State court” unless the adjudication (1) “resulted in a decision that was contrary to, or involved an unreasonable application of, clearly established Federal law, as determined by the Supreme Court of the United States”; or (2) “resulted in a decision that was based on an unreasonable determination of the facts in light of the evidence presented in the State court proceeding.” 28 U.S.C. § 2254(d).
Under § 2254(d)(1), “clearly established federal law” means “the governing legal principle or principles set forth by the Supreme Court at the time the state court renders its decision.” Lockyer v. Andrade, 538 U.S. 63, 71-72, 123 S.Ct. 1166, 155 L.Ed.2d 144 (2003). It “refers to the holdings, as opposed to the dicta, of [the Supreme Court’s] decisions as of the time of the relevant state-court decision.” Williams v. Taylor, 529 U.S. 362, 412, 120 S.Ct. 1495, 146 L.Ed.2d 389 (2000). AED-PA allows federal courts to grant habeas relief only if the state court decision is contrary to, or an unreasonable application of, clearly established federal law. 28 U.S.C. § 2254(d)(1). .
A state court decision is “contrary to” clearly established federal law if the state court (1) “applies a rule that contradicts the governing law” set forth in Supreme Court precedent or (2) “confronts a set of facts that are materially indistinguishable from a decision of [the Supreme] Court and nevertheless arrives at a result different” from that reached by the Supreme Court. Williams, 529 U.S. at 405-06, 120 S.Ct. 1495. Interpreting Supreme Court precedent in a manner that adds an additional element to the legal standard for proving a constitutional violation is “contrary to” clearly established federal *281law. Id. at 393-94, 397, 120 S.Ct. 1495 (reasoning that the Virginia Supreme Court’s interpretation of Strickland v. Washington, 466 U.S. 668, 104 S.Ct. 2052, 80 L.Ed.2d 674 (1984), which increased the burden on petitioners, was “contrary to” Supreme Court precedent).
A state court decision is an “unreasonable application of federal law” if the state court “identifies the correct governing legal principle,” but “unreasonably applies that principle to the facts of the prisoner’s case.” Id. at 413, 120 S.Ct. 1495. A strong case for habeas relief “does not mean the state court’s contrary conclusion was unreasonable.” Harrington v. Richter, 562 U.S. 86, 102, 131 S.Ct. 770, 178 L.Ed.2d 624 (2011). Habeas relief may not be granted on the basis that the state court applied clearly established law incorrectly; rather, the inquiry is “whether the state court’s application of clearly established federal law was objectively unreasonable.” Williams, 529 U.S. at 409, 120 S.Ct. 1495 (emphasis added). A rule’s unreasonable application corresponds to the specificity of the rule itself: “[t]he more general the rule, the more leeway courts have in reaching outcomes in case-by-case determinations.” Richter, 562 U.S. at 101, 131 S.Ct. 770 (internal quotation marks and citation omitted). “A state court’s determination that a claim lacks merit precludes federal habeas relief so long as fairminded jurists could disagree on the correctness of the state court’s decision.” Id. (internal quotation marks omitted).
Finally, under 28 U.S.C. § 2254(d)(2), a state court decision is based on an “unreasonable determination of the facts” if the state court’s factual findings are “objectively unreasonable in light of the evidence presented in the state-court proceeding,” which requires review of whether there was sufficient evidence to support the state court’s factual findings. See Miller-El v. Cockrell, 537 U.S. 322, 340, 123 S.Ct. 1029, 154 L.Ed.2d 931 (2003). Determinations of factual issues made by state courts are presumed to be correct. 28 U.S.C. § 2254(e)(1); Miller-El, 537 U.S. at 340, 123 S.Ct. 1029. However, “[d]eference does not by definition preclude relief. A federal court can disagree with a state court’s credibility determination and, when guided by AEDPA, conclude the decision was unreasonable or that the factual premise was incorrect by clear and convincing evidence.” Miller-El, 537 U.S. at 340, 123 S.Ct. 1029.
Judges Fisher and Hardiman advance an interpretation of Richter that far exceeds its reach. Further, their approach would have the federal habeas courts “rewrite” state court opinions, as Judge Jordan’s thorough concurrence observes. We recognize that the AEDPA standard is “difficult to meet ... because it was meant to be. As amended by AEDPA, § 2254(d) stops short of imposing a complete bar on federal-court relitigation of claims already rejected in state proceedings.” Richter, 562 U.S. at 102, 131 S.Ct. 770. The highly deferential standard “reflects the view that habeas corpus is a guard against extreme malfunctions in the state criminal justice systems, not a substitute for ordinary error correction through appeal.” Id. at 102-OS, 131 S.Ct. 770 (internal quotation marks omitted). This level of deference stems from deep-rooted concerns about federalism. Williams, 529 U.S. at 406, 120 S.Ct. 1495 (noting that Congress intended to “further the principles of comity, finality, and federalism” in passing AEDPA). That said, Richter and its progeny do not support unchecked speculation by federal ha-beas courts in furtherance of AEDPA’s goals. While we must give state court decisions “the benefit of the doubt,” as Judge Fisher recognizes, federal habeas review does not entail speculating as to what other theories could have supported the state court ruling when reasoning has been pro*282vided, or buttressing a state court’s scant analysis with arguments not fairly presented to it. Make no mistake about it, the Dissents justify the state court ruling based on an argument never presented to it. No case decided by our court or the United States Supreme Court permits this approach. We now write to clarify how we interpret the Supreme Court’s jurisprudence as to when and how federal courts ought to “fill the gaps” in state court opinions on federal habeas review subject to AEDPA.
The United States Supreme Court has clearly laid out the analytical path for federal habeas courts confronted with a state court opinion devoid of reasoning — i.e., a bare ruling. When a state court decision lacks reasoning, the Supreme Court instructed habeas courts to “determine what arguments or theories supported or, as here, could have supported, the state court’s decision; and then it must ask whether it is possible fairminded jurists could disagree that those arguments or theories are inconsistent with the holding in a prior decision of this Court.” Richter, 562 U.S. at 102, 131 S.Ct. 770 (emphasis added). Richter is that case. This is not.
In Richter, the Court faced the question of whether AEDPA deference “applies when a state court’s order is unaccompanied by an opinion explaining the reasons relief has been denied.” Id. at 98; 131 S.Ct. 770. The United States Supreme Court admonished the Ninth Circuit’s de novo review of the California Supreme Court’s one-sentence summary denial of petitioner’s claim under Strickland, and held that state court decisions that are devoid of reasoning, i.e., a bare ruling, constitute adjudications on the merits that trigger AEDPA deference. Richter, 562 U.S. at 98, 131 S.Ct. 770 (“[T]he habeas petitioner’s burden still must be met by showing there was no reasonable basis for the state court to deny relief. This is so whether or not the state court reveals which of the elements in a multipart claim it found insufficient....”). In pther words, state courts need not articulate a statement of reasons to invoke AEDPA deference by federal habeas courts. Id. (“[Djetermining whether a state court’s decision resulted from an unreasonable legal or factual conclusion does not require that there be an opinion from the state court explaining the state court’s reasoning.”). The California Supreme Court had provided no reasoning; accordingly, in order to determine whether the state court had made a decision that was contrary to, or involved an unreasonable application of, clearly established'federal law, or an unreasonable determination of fact, the federal habeas court was required to theorize based on what was presented to the state court.
We suggest that the concept of “gap filling” is fairly limited. It should be reserved for those cases in which the federal court cannot be sure of the precise basis for the state court’s ruling. It permits a federal court to defer while still exploring the possible reasons. It does not permit a federal habeas court, when faced with a reasoned determination of the state court, to fill a non-existent'“gap” by coming up with its own theory or argument, let alone one, as here, never raised to the state court. In Premo v. Moore, 562 U.S. 115, 131 S.Ct. 733, 178 L.Ed.2d 649 (2011), decided on the same day as Richter, the state court had concluded that the petitioner had not received ineffective assistance of counsel under Strickland, but did not specify on which Strickland prong — performance or prejudice — petitioner failed to meet his burden. As in Richter, the Supreme Court instructed the Ninth Circuit to assume “that both findings would have involved an unreasonable application of clearly established federal law.” 562 U.S. at 123, 131 S.Ct. 733. Unsure as to which *283prong formed the basis for the state court’s ruling, the federal court could fill the gap by exploring the two prongs of Strickland.
In contrast, when the state court pens a clear, reasoned opinion, federal habeas courts may not speculate as to theories that “could have supported” the state court’s decision. The Supreme Court established this limitation on Richter “gap filling” in Wetzel v. Lambert, — U.S. -, 132 S.Ct. 1195, 182 L.Ed.2d 35 (2012), where it described the proper analytical path for state court decisions accompanied by reasoning:
Under § 2254(d), a habeas court must determine what arguments or theories supported ... the state court’s decision; and then it must ask whether it is possible fairminded jurists could disagree that those arguments or theories are inconsistent with the holding in a prior decision of this Court.
Id. at 1198 (quoting Richter, 562 U.S. at 102, 131 S.Ct. 770; alterations in original; emphasis added). This is fairly straightforward. As explained above, the Court in Richter included the language “or, as here, could have supported” when it initially instructed courts on gap filling. Courts were tasked with considering what theories “could have supported” the state court decision in cases akin to those “as here,” or, summary denials. Removing the clause “or, as here, could have supported” from the instruction when the state court provides a fully-reasoned decision removed the task of speculative gap-filling from the habeas court’s analysis. Instead, federal habeas courts reviewing reasoned state court opinions are limited to “those arguments or theories” that actually supported, as opposed to “could have supported,” the state court’s decision. The Supreme Court’s intent to limit deference to the state court to those reasons that it articulated in its opinion is further supported by the Supreme Court’s instruction that the court on remand consider whether “each ground supporting the state court decision is examined and found to be unreasonable under AEDPA.” Id. at 1199.
When a state court ruling is based on a reasoned, but erroneous, analysis, federal habeas courts are empowered to engage in an alternate ground analysis— relying on any ground properly presented — but, in such a case, the federal court owes no deference to the state court. In Lafler v. Cooper, — U.S.-, 132 S.Ct. 1376, 182 L.Ed.2d 398 (2012), the state court had “simply found that respondent’s rejection of the plea was knowing and voluntary” in rejecting defendant’s ineffective counsel claim and “failed to apply Strickland, ” despite referencing the performance and prejudice prongs of Strickland in its opinion. Id. at 1390. “By failing to apply Strickland to assess the ineffective-assistance-of-counsel claim respondent raised, the state court’s adjudication was contrary to clearly established federal law” and the Supreme Court analyzed the Strickland claim de novo. Id. at 1390. The Court was not filling a gap in Lafler. Instead, it was employing different analysis that was very much a part of the case, and supplied an alternate ground for concluding, on de novo review, that there was no ineffectiveness of counsel.
Justices of the Supreme Court have indicated in a concurrence from the denial of a petition for certiorari that federal courts are bound to the text of state court opinions. Justice Ginsburg, joined by Justice Kagan, observed
Richter1 s hypothetical inquiry was necessary, however, because no state court opinion explained the reasons relief had been denied. In that circumstance, a federal habeas court can assess whether the state court’s decision involved an unreasonable application of clearly established *284Federal law only by hypothesizing reasons that might have supported it. But Richter makes clear that where the state court’s real reasons can be ascertained, the § 2254(d) analysis can and should be based on the actual arguments or theories that supported the state court’s decision.
Hittson v. Chatman, — U.S. -, 135 S.Ct. 2126, 2127-28, 192 L.Ed.2d 887, reh’g denied, — U.S. -, 136 S.Ct. 15, 192 L.Ed.2d 984 (2015) (mem.) (internal quotation marks, alterations, and citations omitted). Other courts of appeals have similarly limited Richter’s gap-filling instruction to the bare ruling situation. See Johnson v. Sec’y, DOC, 643 F.3d 907, 910 (11th Cir. 2011) (“When faced with an ineffective assistance of counsel claim that was denied on the merits by the state courts, a federal habeas court ‘must determine what arguments or theories supported or, [if none were stated], could have supported, the state court’s decision[.]” (alterations in. original) (quoting Richter, 562 U.S. at 102, 131 S.Ct. 770)); see also Grueninger v. Dir., Va. Dep’t of Corr., 813 F.3d 517, 525-26 (4th Cir. 2016) (“looking through” a state court summary refusal to hear an appeal to the prior reasoned decision and observing that “where there is no indication of the state court’s reasoning, a federal habeas petitioner must show that there was ‘no reasonable basis for the state court to deny relief,’ and a federal habeas court must defer under AEDPA to any reasonable ‘arguments or theories ... [that] could have supported!] the state court’s decision’ ” (quoting Richter, 562 U.S. at 98, 102, 131 S.Ct. 770) (internal citations omitted; alterations in original)); Montgomery v. Bobby, 654 F.3d 668, 700 (6th Cir. 2011) (Clay, J., dissenting) (“If the- state court articulated its reasons, the habeas court must identify and evaluate those reasons under § 2254(d); only if the state court did not articulate its reasons must the habeas court hypothesize as to the state court’s reasoning, and evaluate those hypothetical reasons.”). Federal courts should only gap-fill when presented with a bare ruling or when it is unsure as to the basis of the state court ruling on the issue presented. See Premo, 562 U.S. at 123, 131 S.Ct. 733 (concluding that when the state court neglected to articulate which prong of Strickland was deficient, the federal habeas court ought to evaluate both prongs of Strickland). We will not gap-fill when the state court has articulated its own clear reasoning. Instead, we will evaluate the state court’s analysis and review de novo any properly presented alternative ground(s) supporting its judgment.
Dennis’s claims at issue on appeal stem from the Commonwealth’s violations of Brady v. Maryland, 373 U.S. 83, 83 S.Ct. 1194, 10 L.Ed.2d 215 (1963). Prosecutors have an affirmative duty “to disclose [Brady] evidence ... even though there has been no request [for the evidence] by the accused,” which may include evidence known only to police. Strickler v. Greene, 527 U.S. 263, 280, 119 S.Ct. 1936, 144 L.Ed.2d 286 (1999); Kyles, 514 U.S. at 438, 115 S.Ct. 1555. To comply with Brady, prosecutors must “learn of any favorable evidence known to the others acting on the government’s behalf ..., including the police.” Strickler, 527 U.S. at 281, 119 S.Ct. 1936 (internal quotation marks omitted) (quoting Kyles, 514 U.S. at 437, 115 S.Ct. 1555).
To prove a Brady violation, a defendant must show the evidence at issue meets three critical elements. First, the evidence “must be favorable to the accused, either because it is exculpatory, or because it is impeaching.” Id. at 281-82, 119 S.Ct. 1936; see also United States v. Bagley, 473 U.S. 667, 676, 105 S.Ct. 3375, 87 L.Ed.2d 481 (1985) (“Impeachment evidence ..., as well as exculpatory evidence, falls within the Brady rule.”). Second, it *285“must have been suppressed by the State, either willfully or inadvertently.” Strickler, 527 U.S. at 282, 119 S.Ct. 1936. Third, the evidence must have been material such that prejudice resulted from its suppression. Id.; see also Banks, 540 U.S. at 691, 124 S.Ct. 1256. The “touchstone of materiality is a ‘reasonable probability’ of a different result.” Kyles, 514 U.S. at 434, 115 S.Ct. 1555. Materiality “does not require demonstration by a preponderance that disclosure of the suppressed evidence would have resulted ultimately in the defendant’s acquittal ... [Rather], [a] ■ ‘reasonable probability’ of a different result is ... shown when the government’s eviden-tiary suppression undermines confidence in the outcome of the trial.” Id. (internal quotation marks omitted).
III. Discussion
The District Court held that the Pennsylvania Supreme Court had unreasonably applied Brady and its progeny in rejecting Dennis’s claims that the prosecution was required under Brady to disclose the Ca-son receipt, the Frazier documents, and the police activity sheet containing Howard’s inconsistent statements. The Pennsylvania Supreme Court issued a thorough decision on each claim. We conclude, like the District Court, that the Pennsylvania Supreme Court’s decisions regarding Dennis’s Brady claims rested on unreasonable conclusions of fact and unreasonable applications of clearly established law, or were contrary to United States Supreme Court precedent. We will affirm the District Court and grant habeas relief on Dennis’s Brady claims based on the Cason receipt, the Howard police activity sheet, the Frazier documents, and their cumulative prejudice.
A. Cason Receipt
1. Facts
The Commonwealth did not disclose the DPW receipt that was in the police’s possession, provided objective impeachment evidence of a key Commonwealth witness, and bolstered Dennis’s alibi. Cason signed the DPW receipt when she picked up her check on October 22, 1991, the day of Williams’s murder. The receipt’s time stamp shows Cason picked up a $94.00 payment for “public assistance” at “13:03,” or 1:03 p.m. During Dennis’s direct appeal, Cason signed an affidavit detailing her recollection of the interview she had with police prior to Dennis’s trial. According to Cason, detectives brought a copy of the time-stamped receipt to the interview, and she “located and gave the detective [her] pink copy of the same receipt. The detective kept [her] copy of the receipt.” App. 1735.
The Commonwealth called Cason to testify at Dennis’s trial. She testified that she left work around 2:00 p.m., picked up her welfare check, ran errands, and saw Dennis when she got off the K bus “between 4:00 and 4:30.” App. 733. The receipt serves two functions: (1) it negates her testimony that she worked until 2:00 p.m. on October 22; and (2) it demonstrates that, contrary to Cason’s testimony at trial that she retrieved her receipt after 3:00 p.m., Cason actually picked up her check at 1:03 p.m. Cason admits in her affidavit that she “may have thought that the 13:03, which was on the receipt, was 3:03 p.m.” App. 1736. In light of the time-stamped receipt, Cason explained in her affidavit, she “would have seen [James] Dennis between 2:00 and 2:30 p.m. at the Abbotts-ford Homes, and not 4:00 to 4:30 that is in my statement.” Id.
2. State Court Decision
The Pennsylvania Supreme Court rejected Dennis’s Brady claim stemming from the Cason receipt. The Court found, consistent with Cason’s affidavit, that the “police came into possession of a Depart-*286merit of Public Welfare (DPW) receipt showing that Cason cashed her check at 1:03 p.m.” Dennis I, 715 A.2d at 408. In denying Dennis’s ineffective assistance of counsel claim, the Court held that Cason’s new version of events “would not support [Dennis’s] alibi [ ] because the murder occurred at 1:50 p.m., forty minutes earlier than Cason’s earliest estimate” of when she saw Dennis. Id. The Court further held that the corrected testimony “would have been cumulative of testimony of witness Willis Meredith, who testified that he saw [Dennis] at the Abbottsford Homes at approximately 2:15 to 2:30 p.m.” Id. The Court dismissed'the Brady claim because the receipt was “not exculpatory, because it had no bearing on [Dennis’s] alibi, and there [was] no evidence that the Commonwealth withheld the receipt from the defense.” Id.
S. AEDPA Review
The state court ruling was a reasoned ruling that the District Court could understand; no gaps needed to be filled. Dennis was entitled to habeas relief based on the Cason Brady claim only if he could demonstrate that the decision was an unreasonable application of, or contrary to, clearly established law, or an unreasonable determination of the facts. 28 U.S.C. § 2254(d). Addressing the reasoned view of the Pennsylvania Supreme Court, we conclude that it unreasonably applied Brady and its progeny in evaluating the Cason receipt and made unreasonable determinations of fact. The receipt would have served as independent documentary corroboration of a key witness for Dennis’s alibi defense, and suppression by the Commonwealth violated Brady.
a) Favorability
The Cason receipt provided exculpatory and impeachment evidence that would have bolstered Dennis’s alibi defense at trial, so it easily meets Brady’s first prong. Banks, 540 U.S. at 691, 124 S.Ct. 1256 (stating that both impeachment and exculpatory evidence satisfy the first Bradiy prong).
The Pennsylvania Supreme Court erred by failing to recognize the impeachment value of the Cason receipt, which would have provided documentary evidence that Cason testified falsely at trial. The United States Supreme Court has made plain that impeachment evidence may be considered favorable under Brady even if the jury might not afford it significant weight. See Kyles, 514 U.S. at 450-51, 115 S.Ct. 1555 (rejecting the state’s argument that the evidence was “neither impeachment nor exculpatory evidence” because the jury might not have substantially credited it; according to the Court, “[s]uch [an] argument ... confuses the weight of the evidence with its favorable tendency”).16
Dennis’s defense strategy pitted his credibility, and that of his witnesses, against eyewitness credibility, Cason’s testimony, and the testimony of the other prosecution witnesses. No physical evidence was admitted at trial. Evidence that challenged Dennis’s credibility, or that of other defense witnesses dike his father, was therefore particularly crucial to the outcome of the trial. As the District Court aptly noted:.
Armed with the receipt, Dennis’s counsel — at the very least — would have been able to show that Cason was mistaken about the timing of the afternoon, by pointing out that she could not possibly *287have worked until 2 p.m. since she was at the DPW center at 1:03 p.m. ... The time stamped receipt would have directly contradicted [Cason’s testimony that she didn’t get off work until 2:00 p.m.].
Dennis V, 966 F.Supp.2d at 508. Without evidence to challenge the veracity of Ca-son’s testimony, Dennis’s assertion that he saw Cason as he got off the K bus lost significant credibility, as did his father’s corroboration of Dennis’s version of his timeline.
Further, the Pennsylvania Supreme Court erroneously concluded that the receipt was not exculpatory because it did not affect Dennis’s alibi. Dennis I, 715 A.2d at 408. It held that Cason’s revised recollection of the day “would not support [Dennis’s] alibi [ ] because the murder occurred at 1:50 p.m., forty minutes earlier than Cason’s earliest estimate.” Id. This conclusion fails to recognize how Cason’s corrected testimony corroborates testimony provided by Dennis and other witnesses, namely, his father.
The Commonwealth argues that the Pennsylvania Supreme Court reasonably concluded that the receipt did not require disclosure pursuant to Brady because Ca-son’s corrected testimony would not have made it impossible for Dennis to have been at Fern Rock station when Williams was murdered. Cason’s affidavit stated that she saw Dennis at 2:30 p.m. at Abbottsford Homes. The Commonwealth contends that Dennis could have committed the murder at Fern Rock at 1:50 p.m. and returned to Abbottsford Homes by 2:30 p.m. because the shooter entered a waiting getaway car after the murder and it was a thirteen minute drive between the two. This view unreasonably discounts the buttressing effect Cason’s corrected testimony would have on Dennis’s alibi theory. Although Cason’s corrected testimony, assuming it would mirror precisely what she said in her affidavit, would not definitively place Dennis in a location where it was impossible for him to commit the murder, Cason’s testimony would have strengthened Dennis’s and his father’s testimony that Dennis had been with his father that afternoon and was on the bus at the time of the murder.
Validating Dennis’s and his father’s testimony about his alibi on the day in question is sufficient to demonstrate favorability under Brady. Exculpatory evidence need not show defendant’s innocence conclusively. Under Brady, “[e]xculpatory evidence includes material that goes to the heart of the defendant’s guilt or innocence as well as that which may well alter the jury’s judgment of the credibility of a crucial prosecution witness.” United States v. Starusko, 729 F.2d 256, 260 (3d Cir. 1984) (citing Giglio v. United States, 405 U.S. 150, 154, 92 S.Ct. 763, 31 L.Ed.2d 104 (1972)). That Cason’s corrected testimony does not wholly undermine the prosecution’s theory of guilt does not sap its exculpatory value. The Commonwealth had an obligation to disclose the receipt under Brady because it would have altered the jury’s judgment about Cason’s credibility. Cason’s evidence is not favorable simply because of where Cason said she saw Dennis as corrected in her affidavit — at Abbottsford Homes. Rather, as Dennis argues, the exculpatory value lies in corroborating testimony of witnesses at trial who otherwise received little objective reinforcement, and whose credibility, as a result of Cason’s mistaken testimony in the absence of the receipt, was seriously undermined.
The only discrepancy between Cason’s testimony and the alibi established by Dennis and his father was the precise time Cason and Dennis saw one another — Ca-son claimed to have seen Dennis around 4:00 or 4:30 p.m., while Dennis said it was around 2:30 p.m. As both parties note, the *288other witnesses that testified on behalf of Dennis were friends and family, who were vulnerable to arguments of bias. To the contrary, Cason offered disinterested testimony that corroborated the government’s theory. Although the Commonwealth indicates that Cason could have been discredited in a similar manner as Dennis’s other witnesses, nothing in the record indicates that Cason shared the type of close relationship with Dennis as other witnesses who testified on his behalf.
The receipt contradicted Cason’s testimony at trial. Her corrected recollection, coupled with a specific documentary basis, would have provided disinterested corroboration of Dennis’s and his father’s testimony. The Pennsylvania Supreme Court made an unreasonable determination of the facts and an unreasonable application of federal law in refusing to acknowledge the receipt’s exculpatory and impeachment value.
b) Suppression of the receipt
The Pennsylvania Supreme Court stated that “the police came into possession of [the] receipt” when interviewing Cason. Dennis I, 715 A.2d at 408. Later, in a section analyzing materiality, it concluded there was “no evidence that the Commonwealth withheld the receipt from the defense.” Id. The Pennsylvania Supreme Court provided no explanation for its latter statement, and we cannot be sure whether the court was assessing the facts or interpreting the law. If it was construing fact, it was clearly unreasonable because the police had the receipt and therefore so did the prosecution.17 See Kyles, 514 U.S. at 437-38, 115 S.Ct. 1555. If it was making a conclusion of law as to the duty to disclose, the conclusion is similarly problematic because the court ignored Kyles. As Judge Jordan observes in his concurrence, “[i]f one follows the instruction of Kyles, those two statements are impossible to harmonize.” J. Jordan Concurring Op. at 352.
Once the Pennsylvania Supreme Court determined that the police detectives had obtained the receipt from Cason, the Commonwealth had constructive possession and was required to disclose the receipt to Dennis prior to trial. In 1995, three years prior to the Pennsylvania Supreme Court’s decision, the United States Supreme Court explained this duty:
[T]he prosecution, which alone can know what is undisclosed, must be assigned the consequent responsibility to gauge the likely net effect of all [favorable] evidence and make disclosure whén the point of “reasonable probability” is reached. This in turn means that the individual prosecutor has a duty to learn of any favorable evidence known to the others acting on the government’s behalf in the case, including the police. But whether the prosecutor succeeds or fails in meeting this obligation (whether, that is, a failure to disclose is in good faith or bad faith), the prosecution’s responsibility for failing to disclose known, favorable evidence rising to a material level of importance is inescapable.
Kyles, 514 U.S. at 437-38, 115 S.Ct. 1555 (internal quotation marks and citation *289omitted). In ignoring Kyles’s instruction that prosecutors must disclose evidence obtained by the police, the Pennsylvania Supreme Court unreasonably applied clearly established federal law. The Commonwealth’s argument that the receipt did not appear in the prosecution file does nothing to undercut its duty to disclose under Kyles and, as the District Court correctly notes, borders on bad faith. It explained:
The Commonwealth admits that the entire homicide file — where one may expect a document recovered by the police to exist — went missing in March 1997, before the Commonwealth had submitted its direct appeal briefing. The Commonwealth may not point to a missing file and declare it the petitioner’s burden to prove that the receipt was, at one point, contained inside.
Dennis V, 966 F.Supp.2d at 509 (citation omitted). The Commonwealth has never asserted that it disclosed the receipt to Dennis. We refuse to allow it to evade its duty under Brady based on failure to adequately search or maintain its own files.
The Commonwealth argues that because Dennis’s appellate counsel was able to obtain the receipt from the DPW nearly five years post-trial, the prosecution had no responsibility under Brady to turn it over to defense counsel when the receipt came into its possession. Judge Fisher adopts this approach and excuses the Commonwealth from its Brady responsibility by injecting an argument that was not even mentioned by the Pennsylvania Supreme Court, much less fairly presented before it.
The Commonwealth did not raise a “due diligence” argument, as such, before the state court. Rather, in its Response to Defendant’s Reply Brief, the Commonwealth argued for the first time that there was no Brady violation because the receipt was publicly available. The entirety of the alleged due diligence argument is below.
[A]lthough defendant does not explain how he obtained a copy of [the Cason receipt], he presumably did so from the Department of Public Welfare, thus establishing its public availability. Brady does not require the Commonwealth to produce evidence that was not in its sole possession, but was available, as this document apparently was.
App. 2026. As Judge Jordan observes, Pennsylvania law generally regards arguments raised for the first time in reply briefs as waived. J. Jordan Concurring Op. at 353 n.9.
Further, our review on habeas is limited to the record as presented to the state court. See Cullen v. Pinholster, 563 U.S. 170, 181-82, 131 S.Ct. 1388, 179 L.Ed.2d 557 (2011). There was no evidence regarding the availability of the receipt. In fact, the Commonwealth’s assertion that the receipt was publicly available was incorrect, as it runs counter to specific Pennsylvania regulations in effect at the time. As they existed during Dennis’s appeal, the DPW’s privacy regulations protected the vast majority of private information; the only exception was that the Commonwealth may disclose “the address and amount of assistance a person is currently receiving” following a direct request about a specific person. 55 Pa. Code § 105.4(a)(1). Even if the DPW receives a subpoena requesting information about a recipient, it must challenge that demand and “plead, in support of its request to withhold information, that under the Public Welfare Code (62 P.S. §§ 101-1503), the rules of the Department prohibit the disclosure of information in records and files, including the names of clients, except as provided in subsection (a).” Id. § 105.4(b)(3). To the extent that information was publicly available regarding Ca-son’s public assistance payments, it was limited to Cason’s address and her amount *290of assistance, which is irrelevant to her interaction with Dennis on the day of Williams’s murder. Only the Commonwealth held information that would support Dennis’s alibi — the time-stamped receipt Cason provided to the police.
Even if we were to imagine that a diligence argument was presented and considered by the state court, the United States Supreme Court has never recognized an affirmative due diligence duty of defense counsel as part of Brady, let alone an exception to the mandate of Brady as this would t clearly be. The Supreme Court has noted that its precedent “lend[s] no support to the notion that defendants must scavenge for hints of undisclosed Brady material when the prosecution represents that all such material has been disclosed.” Banks, 540 U.S. at 695, 124 S.Ct. 1256. To the contrary, defense counsel is entitled to presume that prosecutors have “discharged their official duties.” Id. at 696, 124 S.Ct. 1256 (quoting Bracy v. Gramley, 520 U.S. 899, 909, 117 S.Ct. 1793, 138 L.Ed.2d 97 (1997)). Further, the duty to disclose under Brady is absolute — it does not depend on defense counsel’s actions. United States v. Agurs, 427 U.S. 97, 107, 96 S.Ct. 2392, 49 L.Ed.2d 342 (1976) (“[I]f the evidence is so clearly supportive of a claim of innocence that it gives the prosecution notice of a duty to produce, that duty should equally arise even if no request is made.”). Brady’s mandate and its progeny are entirely focused on prosecutorial disclosure, not de: fense counsel’s diligence.
The emphasis in the United States Supreme Court’s Brady jurisprudence on fairness in criminal trials reflects Brady’s concern with the government’s unquestionable advantage in criminal proceedings, which the Court has explicitly recognized. See, e.g., Strickler, 527 U.S. at 281, 119 S.Ct. 1936 (reasoning that the “special status” of the prosecutor in the American legal system, whose interest “in a criminal prosecution is not that [he] shall win a case, but that justice shall be done ... explains ... the basis fpr the prosecution’s broad duty of disclosure” (quoting Berger v. United States, 295 U.S. 78, 88, 55 S.Ct. 629, 79 L.Ed. 1314 (1935))). Construing Brady in a manner that encourages disclosure reflects the Court’s concern with prosecutorial advantage and prevents shifting the burden onto defense counsel to defend his actions. Requiring an undefined quantum of diligence on the part of defense counsel, however, would enable precisely that result — it would dilute Brady’s equalizing impact on prosecutorial advantage by shifting the burden to satisfy the claim onto defense counsel.
The focus on disclosure by the prosecutor, not diligence by defense, is reiterated in the Supreme Court’s approval of the shift in the traditional adversarial system Brady imposes. In United States v. Bagley, the Court explained that “[b]y requiring the prosecutor to assist the defense in making its case, the Brady rule represents a limited departure from a pure adversary model” because the prosecutor is not tasked simply with winning a case, but ensuring justice. 473 U.S. 667, 675 n.6, 105 S.Ct. 3375, 87 L.Ed.2d 481 (1985). Further, the Court placed the burden of obtaining favorable evidence squarely on the prosecutor’s shoulders. See Kyles, 514 U.S. at 437, 115 S.Ct. 1555 (“[T]he individual prosecutor has a duty to learn of any favorable evidence known to the others acting on the government’s behalf in the ease.”). That the government may be burdened by the Brady rule does not undercut its need to comply with it. The imposition of an affirmative due diligence requirement on defense counsel would erode the prosecutor’s obligation under, and the basis for, Brady itself.
Indeed, the United States Supreme Court has cautioned against such a rule. It *291has rejected the notion that defense counsel’s diligence is relevant in assessing “cause” for the failure to raise a Brady suppression issue in state court proceedings. In Strickler, it reasoned that because counsel was entitled to rely on the prosecutor fulfilling its Brady obligation, and' had no reason for believing it had failed to comply, the failure to raise the issue earlier in habeas proceedings was justified. See Strickler, 527 U.S. at 286-89, 119 S.Ct. 1936. Similarly here, the prosecutor’s duty is clear. Dennis’s counsel was entitled to rely on the prosecutor’s duty to turn over exculpatory evidence.18 Assessing whether he could or should have discovered the receipt is beside the point.19
In Banks, the Supreme Court explicitly rejected the notion that “the prosecution can lie and conceal and the prisoner still has the burden to ... discover the evidence, so long as the potential existence of a prosecutorial misconduct claim might have been detected.” 540 U.S. at 696, 124 S.Ct. 1256 (internal quotation marks and citations omitted). Banks concluded that “[a] rule ... declaring ‘prosecutor may hide, defendant must seek,’ is not tenable in a system constitutionally bound to accord defendants due process.” Id.; see also United States v. Tavera, 719 F.3d 705, 712 (6th Cir. 2013) (recognizing that “the clear holding in Banks” does away with any belief that Brady imposes a due diligence requirement on defense counsel); Bell v. Bell, 512 F.3d 223, 242 (6th Cir. 2008) (Clay, J., dissenting) (“The rule emerging from Strickler and Banks is clear: Where the prosecution makes an affirmative representation that no Brady material exists, but in fact has Brady material in its possession, the petitioner will not be penalized for failing to discover that material.”).
While we think that the United States Supreme Court has made it clear that Brady requires the prosecution to turn over all material favorable evidence in its possession, we acknowledge that it is not totally frivolous under our Third Circuit jurisprudence for the Commonwealth to have argued, as it did here, that because defense counsel could or should have discovered the Cason receipt with due diligence, the prosecution was not required to disclose it.20 That is because our case law, as we discuss below, is inconsistent and could easily confuse. Thus, we need to clarify our position: the concept of “due diligence” plays no role in the Brady analysis.21 To the contrary, the focus of the *292Supreme Court has been, and it must always be, on whether the government has unfairly “suppressed” the evidence in question in derogation of its duty of disclosure. See Gov’t of the V.I. v. Mills, 821 F.3d 448, 460 n.10 (3d Cir. 2016) (“The critical question in assessing constitutional error is to what extent a defendant’s rights were violated, not the culpability of the prosecutor.” (quoting Marshall v. Hendricks, 307 F.3d 36, 68 (3d Cir. 2002))).
In Brady, the United States Supreme Court held “that the suppression by the prosecution of evidence favorable to an accused upon request violates due process where the evidence is material either to guilt or to punishment, irrespective of the good faith or bad faith of the prosecution.” Brady, 373 U.S. at 87, 83 S.Ct. 1194 (emphasis added). Suppression is “[t]he prosecution’s withholding from the defense of evidence that is favorable to the defendant.” Suppression of Evidence, Black’s Law Dictionary (10th ed. 2014). Inquiries into prosecutorial suppression are, by nature, retrospective as to the actions of the prosecutor — they do not place affirmative duties on defense counsel pre-trial. Agurs, 427 U.S. at 108, 96 S.Ct. 2392 (“[T]he prosecutor will not have violated his constitutional duty of disclosure unless his omission is of sufficient significance to result in the denial of the defendant’s right to a fair trial”).
The government must disclose all favorable evidence. Only when the government is aware that the defense counsel already has the material in its possession should it be held to not have “suppressed” it in not turning it over to the defense. Any other rule presents too slippery a slope. In United States v. Perdomo, 929 F.2d 967, 973 (3d Cir. 1991), and United States v. Starusko, 729 F.2d 256, 262 (3d Cir. 1984), we opened the door to a due diligence exception to Brady. Starusko, 729 F.2d at 262 (“ ‘[T]he government is not obliged under Brady to furnish a defendant with information which he already has or, with any reasonable diligence, he can obtain himself.’ ” (quoting United States v. Campagnuolo, 592 F.2d 852, 861 (5th Cir. 1979))). In Grant v. Lockett, 709 F.3d 224, 230-31 (3d Cir. 2013), we may have widened that opening when we combined our conclusion that defense counsel was constitutionally ineffective in violation of the defendant’s rights with a finding that there was no Brady violation because counsel clearly should have discovered the prosecutor’s key witness’s criminal record and been aware that he was on parole when the shooting occurred and when he testified at trial. We did note in Grant that Grant himself had obtained the witness’s criminal records while in custody, but we did not rest our ruling on that fact.
In Wilson v. Beard, 589 F.3d 651, 663-64 (3d Cir. 2009), we got it right. There we concluded that “[i]f the prosecution has the obligation, pursuant to Perdomo, to notify defense counsel that a government witness has a criminal record even when that witness was represented by someone in defense counsel’s office, the fact that a criminal record is a public document cannot absolve the prosecutor of her responsibility to provide that record to defense counsel.” Id. (emphasis added) (internal quotation marks and citations omitted). Thus, we held that a criminal record, which arguably could have been discovered by defense counsel, is suppressed if not disclosed. Defense counsel in Wilson certainly had the ability to obtain the alleged Brady material — a criminal record — by virtue of his legal training. Yet we required disclosure pursuant to Brady. We also got it right in Pelullo when we rejected defendant’s ar*293gument that certain documents were Brady material and somehow “suppressed” when the government had made the materials available for inspection and they were defendant’s own documents. Pelullo, 399 F.3d at 212 (“[T]he government repeatedly made the warehouse documents available to [the defendant] and his attorneys for inspection and copying.”).
To the extent that we have considered defense counsel’s purported obligation to exercise due diligence to excuse the government’s non-disclosure of material exculpatory evidence, we reject that concept as an unwarranted dilution of Brady’s clear mandate. Subjective speculation as to defense counsel’s knowledge or access may be inaccurate, and it breathes uncertainty into an area that should be certain and sure. See Weisburd, swpra, at 164 (“[P]ros-ecutors ... cannot accurately speculate about what a defendant or defense lawyer could discover through due diligence. Prosecutors are not privy to the investigation plan or the investigative resources of any given defendant or defense lawyer.”). The United States Supreme Court agrees. It has recognized that ample disclosure is “as it should be” because it “tend[s] to preserve the criminal trial, as distinct from the prosecutor’s private deliberations, as the chosen forum for ascertaining the truth about criminal accusations.... The prudence of the careful prosecutor should not therefore be discouraged.” Kyles, 514 U.S. at 439-40, 115 S.Ct. 1555 (internal citations omitted).
All favorable material ought to be disclosed by the prosecution. To hold otherwise would, in essence, add a fourth prong to the inquiry, contrary to the Supreme Court’s directive that we are not to do so. In Williams v. Taylor, the Virginia Supreme Court had interpreted the Supreme Court’s decision in Lockhart v. Fretwell, 506 U.S. 364, 369, 113 S.Ct. 838, 122 L.Ed.2d 180 (1993) “to require a separate inquiry into fundamental fairness even when [petitioner] [was] able to show that his lawyer was ineffective and that his ineffectiveness probably affected the out7 come of the proceeding.” 529 U.S. 362, 393, 120 S.Ct. 1495, 146 L.Ed.2d 389 (2000). The Court held that the Virginia Supreme Court’s imposition of this additional test was an unreasonable application of, and contrary to, Strickland v. Washington, 466 U.S. 668, 104 S.Ct. 2052, 80 L.Ed.2d 674 (1984). Williams, 529 U.S. at 393-94, 120 S.Ct. 1495. Adding due diligence, whether framed as an affirmative requirement of defense counsel or as an exception from the prosecutor’s duty, to the well-established three-pronged Brady inquiry would similarly be an unreasonable application of, and contrary to, Brady and its progeny.
The Pennsylvania Supreme Court’s conclusion that the prosecution did not withhold the Cason receipt was an unreasonable application of law and fact. The receipt was in its possession pursuant to Kyles and, under United States Supreme Court precedent, it is clear that there is no additional prong to Brady and no “hide and seek” exception depending on defense counsel’s knowledge or diligence. See Banks, 540 U.S. at 696, 124 S.Ct. 1256.
c) Materiality
Without a doubt, Dennis suffered prejudice due to the Commonwealth’s failure to disclose the receipt. The defense strategy was rooted in Dennis’s alibi that he was getting on the K bus at the time of the murder. The Commonwealth’s withholding of the receipt transformed a witness who would otherwise have been an alibi witness for Dennis into a witness for the prosecution or, at least, left Dennis powerless to impeach Cason’s false testimony if offered by the prosecution. The state court’s conclusion that Dennis suf*294fered no prejudice is an unreasonable determination of fact and law.
Failure to disclose the Cason receipt made testimony by a key government witness, who provided the sole testimony contradicting Dennis’s alibi, unassailable. The Commonwealth highlighted how weighty Cason’s testimony was at trial. In his opening, referring to Cason as simply a “lady from the neighborhood,” ADA King emphasized the discrepancy between Ca-son’s and Dennis’s testimony: “[Cason] had something very interesting to say. Yeah, I saw him when I was on the bus, but it wasn’t 2:00, it was 4:00.” App. 404. At closing, King reiterated the inconsistencies between Cason’s and Dennis’s testimony, and added that “[the Commonwealth] called her, not the defense. She came in and said, I was at work at 2:00. I saw him somewhere between 4:00 and. 4:30. Try again, Jimmy. That one didn’t work.” App. 1209. Disclosure of the receipt would have given defense counsel evidence to demonstrate that Cason falsely testified when she asserted that she worked until 2:00 p.m. on October 22. Disclosure would have allowed defense counsel to undermine Cason’s credibility or would have caused her to correct her testimony — as she did later in an affidavit — so as to support Dennis’s version of events. Impeachment using the receipt may have caused Cason to explain to the jury that her prior testimony rested on a misunderstanding of military time and allowed Cason to correct her timeline during trial. More likely, the prosecution would not have called Cason at all, and Dennis would have called Cason to corroborate his testimony.22 Finally, ADA King would not have, at closing, been able to point out the inconsistencies between Dennis’s and Cason’s testimonies.
Cason’s uncorreeted testimony left the jury with conflicting stories as to Dennis and Cason’s interactions on the day of the murder. Following Cason’s testimony that she could not have seen Dennis between 2:00 and 2:30 p.m., Dennis qualified his trial testimony and said that he only “thought” he saw Cason. App. 1030. During closing, Dennis’s counsel told the jury, “Remember what [Dennis] told you when he got up there? It’s wrong. He didn’t see [Cason] on the bus. He thought he saw her on the bus, but he didn’t.” App. 1179-80. The District Court thoughtfully explained how Dennis’s uncorrected testimony damaged defense counsel’s strategy:
This scrambled explanation left the jury with two options, equally unhelpful to Dennis: believe that Cason and Dennis had seen each other on the bus, as both testified, but that it happened later than Dennis said — and therefore find no alibi for the time of the crime; or believe counsel’s new story that Dennis was on the earlier bus, and thus could not have committed 'the crime, but never saw Ca-son at all. Cason’s corrected testimony would have transformed Cason from a damaging Commonwealth witness to a uniquely powerful, disinterested defense witness who would provide document-supported corroboration for Dennis’fs] alibi....
Dennis V, 966 F.Supp.2d at 512. The impeachment value the receipt provided would have eliminated the conflicting stories for the jury and, given the weight of Cason’s testimony alleged by the prosecution at trial, could have raised significant doubt about Dennis’s guilt. The state court’s determination that Dennis did not suffer prejudice as a result of Cason’s unchallenged testimony was unreasonable. *295In concluding that the Commonwealth had evidence that its witness’s testimony was false, we need not reach whether the prosecutors here intentionally presented false evidence because the inquiry is solely the impact that the absence of evidence had on the trial. See Brady, 373 U.S. at 87, 83 S.Ct. 1194; Mills, 821 F.3d at 460 n.10.
In Banks, the United States Supreme Court admonished prosecutors for letting statements by an informant, which they believed to be false, stand uncorrected throughout the proceedings. The Court concluded that “prosecutors represented at trial and in. state postconviction proceedings that the State had held nothing back ... It was not incumbent on Banks to prove these representations false; rather, Banks was entitled to treat the prosecutor’s submissions as truthful.” 540 U.S. at 698, 124 S.Ct. 1256. Earlier Brady cases indicate similar concern for allowing false testimony. See, e.g., Agurs, 427 U.S. at 103, 96 S.Ct. 2392 (“[Cjonviction obtained by the knowing use of perjured testimony is fundamentally unfair, and must be set aside if there is any reasonable likelihood that the false testimony could have affected the judgment of the jury.”) (footnotes omitted). Letting Cason’s testimony stand when the Commonwealth had evidence it was false unquestionably violated Brady and entitles Dennis to a new trial.
The state court took an unreasonably narrow view of Brady materiality by focusing on the fact that5 Cason would only have been able to say that she saw him around 2:30 p.m. Cason’s testimony need not fully corroborate Dennis’s alibi in order to show materiality under Brady. Kyles explained that Brady materiality does not turn on a determination of the sufficiency of the evidence, but instead requires the court to consider the constitutional error in light of all the evidence to determine whether it “put[s] the whole case in such a different light as to undermine, confidence in the verdict.” 514 U.S. at 435, 115 S.Ct. 1555. Transforming Cason, a disinterested individual with documentary support, into a defense witness meets the requirements of Brady materiality because it would have necessarily bolstered Dennis’s alibi defense narrative and “put the whole case in ... a different light.” Id.
Dennis testified that his father drove him to the bus stop around 1:50 p.m., where he boarded the K bus. Dennis asserted in his statement to police, which was read into the record at trial, that he waved at Cason when “we got off’ the K bus at Abbottsford Homes, a trip that generally took about thirty minutes. App. 710 (emphasis added). Dennis’s statement implies that they rode the K bus together and, setting aside the difference in time-lines, Cason’s testimony aligns with his account since Cason also took the K bus to Abbottsford Homes and saw Dennis there after she disembarked. Regardless of whether the receipt would have refreshed Cason’s memory enough to cause her to testify that she and Dennis were on the 1:56 p.m. K bus together, it certainly would have empowered defense counsel to elicit testimony from Cason that the loca-' tion in which she saw Dennis was consistent with her exiting the bus at the same time he did and to acknowledge that even if she did not notice him on the bus, she had no reason to disbelieve that he was there.
Cason, unlike the other witnesses Dennis called, did not know him well. Cason testified that she knew Dennis, but when ADA King asked her how long she had known him, Cason replied, “I don’t really, you know, know him, I know him by living up my way” at Abbottsford Homes. App. 731. Because Cason simply knew Dennis from the neighborhood, she served as a significantly less interested witness compared to Dennis’s other testifying witnesses, who were all close friends, family, *296and church leaders. As a result, she was less vulnerable to accusations of bias, and her testimony in support would have carried more weight with the jury. This is particularly important given the nature of her testimony compared to Dennis’s other witnesses. Unlike Dennis’s other witnesses, Cason’s testimony would have been supported by documentary proof of her timeline, the time-stamped receipt, to provide independent credibility to her testimony. In light of the receipt, Cason’s testimony on Dennis’s behalf would have been doubly strong — she was disinterested, and the receipt provided documentary corroboration for her version of the events.23
The Commonwealth criticizes the District Court’s analysis of the Cason receipt Brady claim as a misinterpretation of the record. Primarily, this critique rests on the District Court’s conclusion that the Pennsylvania Supreme Court “overlookfed] the fact that both Cason and Dennis testified that they saw each other on the bus.” Dennis V, 966 F.Supp.2d at 511. While it is true that Cason did not testify at trial that she saw Dennis on the K bus, nor did she deny it, and the Commonwealth’s failure to turn over the receipt deprived defense counsel of the opportunity to refresh Ca-son’s memory with the receipt or at least elicit that she saw Dennis immediately upon exiting the bus, thereby corroborating that they exited at the same location. Given that her unrefreshed testimony put the encounter after 4:00 p.m., defense counsel had no reason to elicit such testimony. But whether Cason testified that she saw Dennis on the bus or disembarking the bus, such testimony would have reinforced Dennis’s own testimony that he was on the bus and placed him in a location that would have made it practically impossible for him to murder Williams. Brady, therefore, required that the Commonwealth disclose the receipt.
At minimum, Cason’s time-stamped receipt would have empowered defense counsel to effectively impeach one of the Commonwealth’s strongest witnesses and mitigated the devastating effect of her testimony on Dennis’s credibility and his father’s. At most, the Commonwealth’s ease would have been short one witness, and Dennis’s alibi defense strategy would have been doubly strong due to (1) Cason’s status as a disinterested defense witness with the documentary corroboration and (2) the resulting increase in Dennis’s and his father’s credibility. The Pennsylvania Supreme Court was therefore unreasonable in concluding that the receipt was not favorable to Dennis when it would have bolstered his alibi. It was unreasonable in concluding that there was “no evidence” that the Commonwealth had suppressed the receipt when the state court found that detectives had the receipt in their possession. And finally, it was unreasonable in concluding that the receipt was not material. Had the Commonwealth disclosed the receipt, the jury may well have credited Dennis’s alibi defense.
B. Howard Police Activity Sheet
1. Facts
A suppressed police activity sheet reveals that two days after Williams’s murder, Zahra Howard, an eyewitness and key witness for the Commonwealth at trial, made a statement to Williams’s aunt and *297uncle, Dianne and Mannasett Pugh, that was inconsistent with an earlier statement she had made to police. Shortly after the murder, Howard told police that she did not recognize the shooter from school. The Pughs told police, however, that Howard told them the day after the murder that she knew the perpetrators from Olney High School, and that “Kim” and “Quinton” were at the scene when the shooting occurred. App. 1506. Quinton was Dianne Pugh’s nephew. The police indicated in their “THINGS TO DO” list that they intended to speak with the Pughs again and “[i]nterview Zahra Howard again” in light of her inconsistent statement to the Pughs. App. 1507. When police met with Howard the following day, however, they did not ask Howard about her conversation with the Pughs.
2. State court decision
The Pennsylvania Supreme Court initially characterized Dennis’s Brady claim regarding Howard’s inconsistent statement as one “with at least arguable merit.” Dennis III, 950 A.2d at 969. But the court was not prepared to rule on the record before it, and it remanded the Howard Brady claim to the PCRA court to address that claim in the first instance. Id.
The PCRA court rejected the Brady claim following an evidentiary hearing. The District Court aptly summarized the PCRA hearing and decision by the Pennsylvania Supreme Court:
Dennis sought to argue the merits of the Brady claim on the papers; he objected to the introduction of evidence from Howard and Diane Pugh because, he argued, their recollections now, a decade after the trial, about who the shooter was or what they told the police had no relevance on the question of whether the Commonwealth had violated Brady by failing to disclose the activity sheet. As Dennis’s PCRA counsel told the court:
The testimony has to be evaluated in its trial context. And all we can do at this point is put on paper for the court what we expect the impeachment to have been, assuming, for example, Zahra Howard denies having made the statement. We have to demonstrate on paper how she could have been impeached, and how that evidence relates to other evidence in the case.... Her testimony today about what she remembers from 16 years ago we can cross-examine, but it doesn’t illuminate the question of materiality in the context of the trial.
NT 12/22/08 at 15. The court allowed the testimony over Dennis’s objections. As expected, both Howard and Pugh denied that Howard had ever suggested that she recognized the assailants. Pugh’s testimony should not carry much weight, however, given that she declared before she was even sworn in, “I don’t remember nothing, nothing at all. It’s been 15, 16 years so I don’t remember. They just subpoenaed me and I’m here.” Id. at 56.
The PCRA court ultimately rejected the Brady claim. It noted that, during the hearing, Howard “testified credibly that she did not know the appellant from Olney High School, nor had she seen him prior to the murder.” Commonwealth v. Dennis, Case No. 92-01-0484, slip op. (Pa.Ct.Com.Pl. Mar. 17, 2010), at 18. Although the question whether Howard recognized James Dennis (“the appellant”) or had seen him before the murder is entirely irrelevant to whether she told Diane Pugh that she had seen the shooter before the murder, this is, in fact, the entirety of the. testimony that the Commonwealth elicited from Howard at the PCRA hearing:
Q: And in that conversation [with Diane Pugh] did you ever say anything about recognizing the defendant before?
*298A: No.
Q: Did you ever see the defendant at Olney High School?
A: No.
Q: Did you ever see him around Olney
High School?
A: No.
NT 12/22/08 at 18. On cross, when Dennis’s lawyer asked her about whether she said she had ever seen the shooter before, or whether she had ever told anyone she recognized the shooter from Olney High School, Howard denied recognizing the shooter or having ever said she did. Id. at 25-27.
Given both trial and PCRA counsel’s thorough cross-examination of Howard, the PCRA court determined that it was “unlikely that any additional impeachment evidence contained in the police activity sheet ... would have created a reasonable probability that the result of the proceeding would have been different had it been disclosed.” Dennis, slip op. at 14. The court further noted that the government’s case at trial “did not rest solely on” Howard’s testimony. Id. Finally, the contents of the activity sheet amounted to inadmissible hearsay, which “cannot be the basis for a Brady violation.” Id. at 15.
The Pennsylvania Supreme Court largely accepted the PCRA court’s determinations, despite its seeming, recognition, in Dennis III, of the investigatory value the activity sheet would have had and its earlier dismissal of the admissibility issue. It agreed that Dennis had failed to prove a reasonable probability of a different result had the activity sheet been disclosed. Commonwealth v. Dennis, 609 Pa. 442, 17 A.3d 297, 309 (2011) (“Dennis TP”). It echoed the PCRA court in noting that “Howard was extensively cross-examined by defense counsel in an attempt to impeach her testimony during trial,” and that “there were two eyewitnesses other than Howard” who identified Dennis; “[t]he disclosure of the activity sheet would have had no impact upon these eyewitnesses’ testimony.” Id. It did not specifically address the question of admissibility.
Dennis V, 966 F.Supp.2d at 513-14.
S. AEDPA Review
There is no question that Howard’s inconsistent statement would have been helpful to the defense but was not revealed to defense counsel until PCRA discovery, ten years after trial. The Pennsylvania Supreme Court denied Dennis’s Brady claim regarding the Howard statement on materiality grounds. Although the court articulated the proper standard for materiality, whether a “reasonable probability” of a different outcome has been established, it applied Kyles in a manner inconsistent with Supreme Court precedent.
First and foremost, defense counsel could have used Howard’s inconsistent statement as an effective means of impeachment during trial. As noted above, impeachment evidence unquestionably falls under Brady’s purview and cannot be suppressed by the prosecution. The Commonwealth notes that evidence is not necessarily material under Brady simply because it may open up avenues for impeachment— the focus of the inquiry is on the “reasonable probability of a different result” under Kyles. Such a probability exists here. The type of impeachment evidence provided by the activity sheet would have undercut the credibility of a key prosecution witness in a manner not duplicated by other challenges the defense was able to level at trial. Consequently, the impeachment material provided by the suppressed activity sheet is material under Brady, and it was unreasonable for the Pennsylvania Supreme Court to hold otherwise.
*299Howard was the Commonwealth’s key eyewitness against Dennis and the Commonwealth accordingly highlighted her testimony. ADA King emphasized the importance of Howard’s testimony in his closing argument: “[I]f you believe Zahra Howard, that’s enough to convict James Dennis.” App. 1207. As Williams’s Mend and the person with the closest view of the shooter, Howard’s testimony carried significant emotional and practical weight with the jury.24
Unlike other testifying eyewitnesses, Howard had views of the perpetrator at numerous stages during the incident. At trial, Howard testified that she saw the shooter for approximately twenty seconds total. This comported with her testimony' at the preliminary hearing. The two other testifying eyewitnesses’ views were much briefer. Bertha testified at the preliminary hearing that he saw the assailant for about a second. At trial, he expanded the amount of time he said he saw the shooter to three or four seconds. Cameron initially testified at the preliminary hearing that he saw the assailant for twenty seconds but upped the amount of time to thirty to forty seconds at trial. Notably, Cameron qualified his testimony by admitting that he “didn’t really pay attention.”25 App. 664. In contrast to Bertha and Cameron’s, Howard’s testimony was consistent, lengthy, and involved numerous views of the assailant— on the subway stairs, during the face-to-face encounter and finally, when Williams was shot. Because of the consistency and emotional weight of Howard’s testimony, defense counsel’s strategy was heavily reliant on impeaching Howard by any means — counsel attempted to “discredit her any ... way [he] could.” App. 1326.
Counsel’s ability to discredit Howard was limited, however. Without evidence that would directly contradict Howard’s testimony at trial, defense counsel sought to discredit Howard by pointing out her initial hesitation in identifying Dennis as the perpetrator during the photo array. Counsel could not challenge Howard’s trial testimony on other grounds. But prosecutors held contradictory statements by Howard about whether she recognized the perpetrators. Howard had initially told police, and later testified at trial, that she had never seen the perpetrators before and had not recognized them from school. According to the Pughs, however, Howard had said she recognized the shooter from Olney High School. The Pughs (along with Parker) also stated that Howard had also identified two other individuals, Kim and Quinton, as being present at the scene.
As noted by the District Court, cross-examination does not equate to actu*300al impeachment. Defense counsel cross-examined Howard, but he could only engage in limited questioning focused on challenging her hesitation identifications of Dennis as the shooter. This is decidedly different from the actual impeachment enabled by the activity sheet. In Banks, a witness was heavily impeached at trial, but the prosecution suppressed evidence that the witness served as a paid informant. 540' U.S. at 702, 124 S.Ct. 1256. Accordingly, none of the impeachment conducted at trial covered his status as an informant; the jury weighed his credibility without knowing this. Id. at 702-03, 124 S.Ct. 1256. The Supreme Court rejected the state’s argument that because the witness was heavily impeached, further impeachment evidence was immaterial. Id. at 702, 124 S.Ct. 1256. We have similarly indicated that additional impeachment evidence helps to substantiate Brady claims in a way that might make them material. In Lambert v. Beard, we stated that “it is patently unreasonable to presume — without explanation — that whenever a witness is impeached in one manner, any other impeachment becomes immaterial.” 633 F.3d 126, 134 (3d Cir. 2011), judgment vacated on other grounds sub nom. Wetzel v. Lambert, — U.S. -, 132 S.Ct. 1195, 182 L.Ed.2d 35 (2012). The mere fact that a witness has been heavily cross-examined or impeached at trial does not preclude a determination that additional impeachment evidence is material under Brady.
Indeed, we have granted habeas relief on the basis of a “significant difference” between the suppressed impeachment and other types of impeachment evidence used at trial. Slutzker v. Johnson, 393 F.3d 373, 387 (3d Cir. 2004). In Slutzker, we held that a police report memorializing a witness’s inconsistent statement was significantly different from the reports used to impeach the witness at trial. In the reports used at trial, the witness failed to identify the defendant, but in the suppressed report, she definitively stated that the man she saw was not the defendant. We concluded that “[t]he latter is much more convincing impeachment evidence, and the failure to disclose it leaves us in doubt that the trial verdict was worthy of confidence.” Id. The police activity sheet memorializing Howard’s statement similarly provides distinct and persuasive impeachment material that discredits Howard’s testimony more thoroughly than the identification challenges defense counsel levelled at trial.
The Commonwealth relies on United States v. Walker, 657 F.3d 160 (3d Cir. 2011), and United States v. Perez, 280 F.3d 318 (3d Cir. 2002), in arguing that the activity sheet does not add anything significant to the record and is consequently immaterial, even if the evidence is unique. However, the activity sheet adds to the record in a distinct and significant way, so Walker and Perez do not compel us to find it immaterial. In Walker, defendants sought a new trial based on the state’s suppression of information, unrelated to the trial itself, about an informant witness. The informant, who testified at defendant’s trial, was found with cocaine and marijuana in his pocket on the day of a controlled buy operation in an unrelated case. We held that suppression of that information did not rise tó the level of a Brady violation. 657 F.3d at 188 (noting that another witness for the prosecution provided direct support). Unlike our case, where Howard’s statement to the Pughs directly undercut the credibility of her eyewitness testimony in Dennis’s case, the alleged Brady evidence in Walker was wholly unrelated to defendant’s case. Further, we reiterated the principle in Walker that “there are some instances where specific impeachment evidence is so important (for issues such as the identity of the culprit) that it is material for Brady purposes even when a witness has already been effectively impeached on other issues.” Id. (emphasis *301added). Thus, Walker supports the view that withholding impeachment material that is germane to a critical aspect of the case — as here, the identity of the perpetrator — violates Brady.
Similarly, Perez does not support the Commonwealth’s contention. The alleged Brady material in Perez was a witness’s later statement inculpating another defendant and exculpating Perez. The initial statement, unlike Howard’s initial statement in this case, was corroborated by documentary evidence and co-defendant testimony at trial. Here, Howard’s eyewitness testimony played a pivotal emotional and practical role that could not be replaced by other evidence. There are material differences in impeachment value as well. In Perez, we concluded that cross-examination on the basis of the later statement would not have induced the co-defendant to admit to committing the crime. Perez, 280 F.3d at 350-51. Here, the type of statement at issue is different — Howard would have been confronted with an inconsistent statement, but not one that would have implicated her in the crime.
Armed with the activity sheet, defense counsel could have impeached Howard in a manner that very well may have led her to admit she recognized the perpetrators from her high school. Regardless of whether she actually recognized the shooter, Howard’s credibility would have been placed counter to that of the victim’s aunt and uncle, the Pughs, who would have undoubtedly been called at trial. Consequently, Howard’s impeachment could have changed the jury’s perception of her credibility.
There are significant, material differences between the type of cross-examination defense counsel engaged in and what he could have done had he known of the police activity sheet. As the District Court noted, “the activity sheet would have shown that [Howard] either lied to Williams’[s] close relatives — only days after the murder and in a manner that implicated Diane Pugh’s own nephew — or she was lying at trial.” Dennis V, 966 F.Supp.2d at 515. Thus, the government’s suppression necessarily undermines confidence in the outcome of Dennis’s trial. Discrediting the prosecution’s central witness, and the eyewitness with the most significant exposure to the shooter, would have had devastating effects on the prosecution’s case at trial. The remaining two eyewitnesses were located farther away from the incident, had only brief glimpses of the perpetrators, or were admittedly paying little attention. Challenging Howard’s identification of the shooter did little to undermine her credibility as a witness; but armed with the inconsistent statement, defense counsel could have undercut Howard’s testimony sufficiently that a jury may not have convicted Dennis. There is a reasonable probability that had the activity sheet been disclosed, the result of the proceeding would have been different.
The Commonwealth argues that Howard did not make the statements attributed to, her in the activity sheet. In support of this assertion, the Commonwealth looks to Howard’s and the Pugh’s testimony during PCRA review — over sixteen years after Dennis’s trial. Her statements during PCRA review carry little weight in how we consider a jury’s credibility determination at trial. In Kyles, the Supreme Court explicitly rejected the contention that post-conviction credibility determinations could replicate the jury’s credibility determinations at trial. Kyles, 514 U.S. at 449 n.19, 115 S.Ct. 1555 (“[N]either observation [during post-conviction proceedings] could possibly have affected the jury’s appraisal of Burns’s credibility at the time of Kyles’s trials.”). The court oriented its analysis around how the jury would have weighed the information, not the credibility of the post-conviction testimony itself. Thus, the *302proper inquiry remains whether use of the activity sheet by defense counsel at trial would have resulted in a different outcome at trial. The jury makes the credibility determination, not the Court sixteen years post-trial.
Although the Supreme Court instructed habeas courts in Wood not to ignore testimony at evidentiary hearings that would undermine the potential usefulness of alleged Brady material, the admissions during a post-conviction hearing in Wood differed significantly from those provided by Howard during PCRA review. In Wood, counsel specifically admitted that “disclosure [of the polygraph results] would not have affected the scope of his cross-examination,” and consequently, he did not bother to obtain admissions during post-conviction review. Wood, 516 U.S. at 7-8, 116 S.Ct. 7. The post-conviction testimony at issue here is markedly different. Dennis’s trial counsel testified that discrediting Howard through inconsistent statements was an integral part of the trial strategy. Interpreting Howard’s statements during PCRA hearings as indicating that she did not, in fact, make the statements to the Pughs contained in the activity sheet would allow the Commonwealth to cure its suppression of material evidence through delay. This we will not do.
The Commonwealth’s argument that the information contained in the activity sheet was double hearsay, so not admissible for impeachment purposes, fairs no better. The Pennsylvania Supreme Court did not rest its decision on an admissibility determination. Rather, it rooted its analysis in a misapplication of the Kyles materiality standard: that “any additional impeachment based on the activity sheet would have created a reasonable probability that the result of the proceeding would have been different.” Dennis IV, 17 A.3d at 309.
Counsel could also have used the information to challenge the adequacy of the police investigation. Defense counsel could have questioned Detectives Jastrzembski and Santiago as to why they did not ask Howard questions about her inconsistent statement when they saw her again only a few hours after indicating that confronting her was part of their “things to do.” Their subsequent meeting with Howard centered on reviewing a photo array. The detectives never asked Howard about admitting to the Pughs that she recognized the assailants from Olney High School. They never asked Howard about Kim and Quinton, despite having recently left a discussion with Parker, who stated that Howard mentioned Kim and Quinton to her as well. There is also no indication that they conducted any further investigation into the Pughs and whether they misheard all of these details or had reason to fabricate Howard’s inconsistent statement. Armed with the statement, defense counsel could have highlighted the investigatory failures for the jury, which could have supported Dennis’s acquittal.
Further, defense counsel could have used the Howard inconsistent statement to mount an “other suspect” defense at trial. According to the Pughs, Howard stated that she recognized the shooter from Ol-ney High School where she and Williams were enrolled. Dennis attended Roxbor-ough High School for his entire high school career. The simple conflict between where Dennis attended school and where Howard stated the assailants went to school would have removed Dennis as a suspect and empowered defense counsel to put forth an “other suspect” defense at trial, which he was otherwise unable to do. Together with the failure to follow up on the statements to the Pughs, defense counsel could have urged that Dennis’s was a case where police arbitrarily put blinders on as to the possibility that someone else committed the crime and pursued the easy lead.
*303Although the Pennsylvania Supreme Court acknowledged that “the omission is to be evaluated in the context of the entire record,” Dennis IV, 17 A.3d at 309, it ultimately applied the Brady materiality standard unreasonably by using sufficiency of the evidence as a touchstone. As pointed out by the District Court, the Supreme Court instructed in Kyles that “[a] defendant need not demonstrate that after discounting the inculpatory evidence in light of the undisclosed evidence, there would not have been enough left to convict.” 514 U.S. at 434-35, 115 S.Ct. 1555. Rather, “the Kyles Court rebuked the dissent for assuming that Kyles must lose on his Brady claim because there would still -have been enough to convict, even if the favorable evidence had been disclosed. ‘The rule is clear, and none of the Brady cases has ever suggested that sufficiency of the evidence (or insufficiency) is the touchstone.’ ” Dennis V, 966 F.Supp.2d at 516 (quoting Kyles, 514 U.S. at 435 n.8, 115 S.Ct. 1555). State courts may not “emphasize[ ] reasons a juror might disregard new evidence while ignoring reasons she might not.” Wearry v. Cain, — U.S.-, 136 S.Ct. 1002, 1007, 194 L.Ed.2d 78 (2016).
The Pennsylvania Supreme Court concluded that “[t]he disclosure of the activity sheet would have had no impact upon [two additional] eyewitnesses’ testimony” and consequently, the activity sheet was not material under Brady. Dennis IV, 17 A.3d at 309. In making its conclusion as to the materiality of the activity sheet, the Pennsylvania Supreme Court tied the materiality of the activity sheet to a requirement that Dennis show that Cameron’s and Bertha’s eyewitness testimony would not be sufficient to support the jury’s finding. This analysis is entirely inconsistent with the Court’s instructions on materiality. The Commonwealth argues, and the Dissent appears to accept, that by citing Commonwealth v. Weiss, 604 Pa. 573, 986 A.2d 808 (2009) — which reiterated the Supreme Court’s admonition of the sufficiency of the evidence test — the Pennsylvania Supreme Court applied the proper standard. However, unreasonable application of federal law under AEDPA occurs when the state court identifies the proper principle, but “unreasonably applies that principle to the facts of the prisoner’s case.” Williams, 529 U.S. at 413, 120 S.Ct. 1495. Indeed, in Lafler, the state court had identified the two Strickland prongs — prejudice and performance — yet the United States Supreme Court concluded that the state court had unreasonably used the “knowing and voluntary” standard and disregarded Strickland. 132 S.Ct. at 1390.
Here, the Commonwealth’s argument that the Pennsylvania Supreme Court knew the proper standard for materiality does little to demonstrate that it actually applied it reasonably. Instead of engaging in a holistic materiality inquiry per Kyles, the Pennsylvania Supreme Court proceeded down an analytical path that hinged the activity sheet’s Brady materiality on the sufficiency of the evidence, namely, the strength of Bertha and Cameron’s eyewitness testimony, in direct contravention of how the Supreme Court has defined materiality.
Judge Fisher’s Dissent relies on the Supreme Court’s decision in Strickler to support the Pennsylvania Supreme Court’s approach to materiality in Dennis IV. Like the activity sheet, the exculpatory materials at issue in Strickler would have cast doubt on the testimony of a key prosecution, Anne Stoltzfus. In Strickler, the Court of Appeals for the Fourth Circuit below had identified the Kyles standard for materiality and had concluded that “without considering Stoltzfus’ testimony, the record contained ample, independent evidence of guilt, as well as evidence sufficient to support the findings of vileness and future dangerousness that warranted *304the imposition of the death penalty.” Strickler, 527 U.S. at 290, 119 S.Ct. 1936. The United States Supreme Court soundly rejected the Fourth Circuit’s approach upon review in Strickler. It instructed that “[t]he standard used by [the Fourth Circuit] was incorrect” and reiterated that “the materiality inquiry is not just a matter of determining whether, after discounting the inculpatory evidence in light of the undisclosed evidence, the remaining evidence is sufficient to support the jury’s conclusions.” Id. (“[T]he question is whether ‘the favorable evidence could reasonably be taken to put the whole case in such a different light as to undermine confidence in the verdict.’ ” (quoting Kyles, 514 U.S. at 435, 115 S.Ct. 1555)). The Pennsylvania Supreme Court did precisely what the Strickler Court rejected — it evaluated whether, after considering Howard’s testimony, the remaining eyewitness testimony was sufficient for Dennis’s conviction. Dennis IV, 17 A.3d at 309 (“[T]here were two eyewitnesses other than Howard who observed the shooting at close range. ... The disclosure of the activity sheet would have had no impact upon these eyewitnesses’ testimony.”).
Further, the materiality of the impeachment evidence in Strickler is distinguishable from the police activity sheet at issue here because the evidence against petitioner in Strickler was far more extensive and varied than the Commonwealth’s case against Dennis. As Judge Fisher recognizes, there was “considerable forensic and other physical evidence” linking the petitioner to the crime in Strickler. 527 U.S. at 293, 119 S.Ct. 1936. The Supreme Court ultimately concluded that “[t]he record provide[d] strong support for the conclusion that petitioner would have been convicted of capital murder and sentenced to death, even if Stoltzfus had been severely impeached.” Id. at 294, 119 S.Ct. 1936. Thus, the Strickler Court held that petitioner had not shown materiality under Brady.
The record laid by the Commonwealth in Dennis’s case pales in comparison to the one mounted by the government in Strick-ler. For instance, the police in Strickler recovered hairs from clothing found with the victim that were microscopically akin to petitioner’s, and petitioner’s fingerprints were found on the inside and outside of the victim’s car. 527 U.S. at 293 n.41, 119 S.Ct. 1936. No similar physical evidence exists on the record in Dennis’s case. The Supreme Court recognized the importance of Stoltzfus’s testimony, as it was the only disinterested narrative account provided at trial, but ultimately concluded in its holistic materiality inquiry that petitioner failed to show that there was “a reasonable probability that his conviction or sentence would have been different had these materials been disclosed.” Id. at 296, 119 S.Ct. 1936. The conclusion that petitioner failed to show materiality against the variety and extensiveness of the evidence against petitioner in Strickler differs from the Pennsylvania Supreme Court indication that two other eyewitness accounts were sufficient for a jury to convict Dennis.
In sum, the Pennsylvania Supreme Court unreasonably applied Brady and its progeny in denying Dennis’s Brady claim based on the Howard inconsistent statement. It unreasonably disregarded the impeachment value of the evidence in discrediting the Commonwealth’s key eyewitness and the adequacy of the investigation. It unreasonably applied a sufficiency of the evidence test by tying the materiality of the activity sheet to the sufficiency of the remaining inculpatory eyewitness testimony. And finally, the Pennsylvania Supreme Court failed to consider that the activity sheet would have enabled defense counsel to raise a defense he was otherwise unable to present — that a student at *305Olney High School committed the murder. There is a reasonable probability that, had the activity sheet been disclosed, the jury would have had a reasonable doubt as to Dennis’s guilt.
C. Frazier Documents
1. Facts
Prior to Dennis’ arrest, Philadelphia police received a lead .from Montgomery County Detectives that someone other than Dennis may have murdered Williams. William Frazier, an initiate at the Montgomery County Correctional Facility called police and told them that Tony Brown “shot ... [a] female in the middle of the street near the Fern Rock station” after the girl resisted his efforts to take her earrings, which Brown sold at a pawn shop for $400. App. 1689-90.
Frazier heard Brown’s confession during a three way call facilitated by his aunt, Angela Frazier. Frazier recounted the conversation in a signed statement given to Philadelphia Police less than two weeks after Williams’s murder. Brown admitted that he — along with Frazier’s cousin, Ricky Walker, and a man called “Skeet”— had “fucked up” and killed Chedell Williams. App 1692. Frazier told police that Brown knew Williams, and identified her as “Kev with the blue pathfinder ... his girl.” App. 1694.
During the call, Brown asked Frazier if he heard about “the incident on the news about the girl that [was] killed over a pair of earrings,” and Brown confessed “that was us.” App. 1692. Frazier reported “[Tony] said that he and Ricky got out of the car and Skeet was driving. They approached the girl, Tony pulled his gun out and told her to give up the earrings she refused. So he put-the gun to her neck ... [and] it accidentally went off.” Id. Walker briefly joined the call and reported that they were scared, and that they left Frazier’s apartment, where they sought cover after the murder, in the middle of the night so that no one would see them. Frazier reported that Brown and Walker sounded “extremely nervous and upset.” App. 1694. Frazier described Tony as 5'7", two inches taller than Dennis, with light brown skin. Like the assailant, Tony “like[d] to wear sweat suits;” he had also committed robberies in the past and owned “a collection of guns.” App. 1693-95.
Frazier gave 'detectives addresses for Brown and Walker, the address where Skeet used to live. Frazier also gave police Angela Frazier’s address and phone number, Brown’s mother’s address, and an address of the pawn shop, along with a description of the proprietor. Frazier agreed to go on a ride along to show detectives the addresses he reported. The Philadelphia police, including Jastrzembski, spoke with Frazier’s landlord, who confirmed that Frazier rented the apartment located at the address he provided. Although the landlord reported that nobody had been in the apartment since Frazier’s arrest, the men used unconventional means to enter Frazier’s apartment the night of the murder — they climbed through Frazier’s right window.26
Detectives interviewed Walker, who told them that he “c[ouldn’t] stand” his cousin, *306Frazier. App. 1703. Walker denied knowing Tony Brown and Skeet and denied any involvement with, or knowledge about, Williams’s murder. He told detectives that he was at his house with his mother on the day of the murder. Police did not conduct an investigation into Walker’s alibi or alert defense counsel to any of the information on Frazier’s tip.
2. State court decision
The Pennsylvania Supreme Court affirmed the PCRA court’s denial of Dennis’s Brady claim as to the Frazier documents on the grounds that Dennis failed to demonstrate that the documents were material and admissible. The Pennsylvania Supreme Court relied on its decision in Commonwealth v. Lambert, 584 Pa. 461, 884 A.2d 848 (2005), in which it emphasized that the prosecution need not “disclose to the defense every fruitless lead followed by investigators of a crime” and asserted that “inadmissible evidence cannot be the basis for a Brady violation.” Lambert, 884 A.2d at 857 (citation omitted). The court concluded: “In the absence of any argument regarding the gravamen of Lambert ... [Dennis] has failed to establish a basis for relief’ regarding the Frazier documents. Dennis III, 950 A.2d at 968. However, as Dennis points out, the Pennsylvania Supreme Court retreated from its decision in Lambert in a later opinion so as to comport with Supreme Court precedent regarding the need for admissibility. Commonwealth v. Willis, 616 Pa. 48, 46 A.3d 648, 670 (2012) (“[W]e hold that admissibility at trial is not a prerequisite to a determination of materiality under Brady.... Therefore, nondisclosed favorable evidence which is not admissible at trial may nonetheless be considered material for Brady purposes[.]”).
3. AEDPA Review
The state court addressed the merits of the Frazier claim and, as a result, Dennis may obtain habeas relief only if he can demonstrate that the decision was an unreasonable application of, or contrary to, clearly established law, or an unreasonable determination of the facts. 28 U.S.C. § 2254(d). It is undisputed that the first two elements of Brady are met. The Frazier documents indicated that someone other than Dennis committed the crime, and were thus exculpatory, and there is no question that the state did not disclose the documents until PCRA discovery. However, the Pennsylvania Supreme Court unreasonably applied Brady and its progeny in concluding that the Frazier documents were immaterial. Also, in appending an admissibility requirement onto Brady, the Pennsylvania Supreme Court acted contrary to clearly established law, as defined by the United States Supreme Court.
The Pennsylvania Supreme Court’s justification that the Frazier documents were a “fruitless lead” was unreasonable. There is no requirement that leads be fruitful to trigger disclosure under Brady, and it cannot be that if the Commonwealth fails to pursue a lead, or deems it fruitless, that it is absolved of its responsibility to turn over to defense counsel Brady material. The rationale behind Brady itself rests on the principle that prosecutors bear an obligation to structure a fair trial for defendants:
Society wins not only when the guilty are convicted but when criminal trials are fair; our system of the administration of justice suffers when any accused is treated unfairly. ... A prosecution that withholds evidence ... which, if made available, would tend to exculpate him or reduce the penalty helps shape a trial that bears heavily on the defendant. That casts the prosecutor in the role of an architect of a proceeding that does not comport with standards of justice, *307even [if] ... his action is not the result of guile[.]
Brady, 373 U.S. at 87-88, 83 S.Ct. 1194 (internal quotation marks and footnote omitted). Structuring a fair trial for defendants demands that prosecutors freely disclose material that is helpful to the defense. Consequently, making Brady disclosure depend on a prosecutor’s own assessment of evidentiary value, as opposed to the benefit to defense counsel, is anathema to the goals of fairness and justice motivating Brady.
The lead was not fruitless, it was simply not rigorously pursued. Detectives did not interview Angela Frazier, who facilitated the three-way call and was on the phone when Brown confessed to the murder. Detectives did not question Walker again— who admitted to having a bias against Frazier — after he stated that he did not know any Brown or Skeet, nor did they attempt to verify Walker’s alibi on the day of the murder. Detectives did not investigate the owner of the pawn shop where Brown purportedly sold Williams’s earrings. Detectives did not obtain the photos of Brown, Skeet, and Walker that were in Frazier’s apartment. Detectives went to an incorrect address seeking information about Skeet and spoke with a woman named Janice Edelen, who said she did not know any man called Skeet. Finally, detectives did not visit the addresses Frazier provided until ten years after the murder. Armed with the Frazier documents, Dennis’s counsel would have been prepared to pursue the lead himself or at least informed the jury of the police’s misguided focus on Dennis and failure to pursue the lead.
The Pennsylvania Supreme Court grafted an admissibility requirement onto the traditional three-prong Brady inquiry when it rejected Dennis’s Brady claim as to the Frazier documents on the ground that he failed to affirmatively show that the documents were admissible. The Pennsylvania Supreme Court’s characterization of admissibility as dispositive under Brady was an unreasonable application of, and contrary to, clearly established law as defined by the United States Supreme Court.
The Commonwealth articulates the Pennsylvania Supreme Court’s decision somewhat differently. It argues that our role on habeas review is determining “whether, under Supreme Court precedent, it was objectively unreasonable for the Pennsylvania Supreme Court to reject Dennis’s claim that he only had to argue or allege that disclosure ‘might’ have affected his investigation or preparation for trial.” Appellants Br. 74. This framing incorrectly states what the Pennsylvania Supreme Court did in Dennis III. It did not simply discount Dennis’s argument that defense counsel could have prepared differently had the documents been disclosed — it appended an admissibility requirement to Brady in contravention of clearly established law.
The Pennsylvania Supreme- Court cited Wood v. Bartholomew, 516 U.S. 1, 116 S.Ct. 7, 133 L.Ed.2d 1 (1995), as attaching an admissibility requirement to Brady. The United States Supreme Court’s holding in Wood compels the opposite conclusion, however. The Supreme Court held in Wood that there was no Brady violation when the prosecution did not disclose the results of two polygraph examinations that were inadmissible at trial. Wood, 516 U.S. at 6, 116 S.Ct. 7. The Wood Court noted that Brady governs “evidence,” and that the polygraph results, since they were inadmissible under state.law, were “not ‘evidence’ at all.” Id. at 5-6, 116 S.Ct. 7. However, under Washington law, poly-graphic examinations cannot be admissible for any purpose at trial, even for impeachment purposes. Id. at 5, 116 S.Ct. 7. At most, the Court’s holding in Wood could *308support the proposition that evidence that cannot be used in any manner at trial under state law may be immaterial under Brady. The holding does not reach so far as to allow state courts to attach a general admissibility requirement onto the Brady inquiry as the Pennsylvania Supreme Court did in Dennis III.
Further, the Wood Court analyzed the effect of suppressing the polygraph results, despite their uncontroverted inadmissibility. After acknowledging their inadmissibility, the Wood Court proceeded to examine whether, if disclosed, the results would have led to the discovery of evidence that would have influenced the course of trial, including pre-trial preparations. See Wood, 516 U.S. at 7, 116 S.Ct. 7 (considering whether trial counsel would have prepared differently given the results, though ultimately concluding that disclosure would not have resulted in a different outcome). The Supreme Court’s decision to continue its inquiry in light of wholly inadmissible alleged Brady material is telling. As the District Court aptly observed, “[i]f inadmissible evidence could never form the basis of a Brady claim, the Court’s examination of the issue would have ended when it noted that the test results were inadmissible.” Dennis V, 966 F.Supp.2d at 503.
The Supreme Court’s choice in Wood to consider the way in which suppression of the polygraph results affected preparation and trial aligns with the way in which materiality is discussed in Kyles. Kyles makes clear that evidence is material under Brady when the defense could have used it to “attack the reliability of the investigation.” 514 U.S. at 446, 115 S.Ct. 1555. As noted by the District Court, in Kyles, defense counsel could have used the information at issue “to throw the reliability of the investigation into doubt and to sully the credibility” of the lead detective. Id. at 447, 115 S.Ct. 1555. The proper inquiry for the Pennsylvania Supreme Court was to consider whether disclosure of the Frazier documents would have impacted the course of trial, which includes investigative activities. Here, disclosure of the Frazier documents would have empowered defense counsel to pursue strategies and preparations he was otherwise unequipped to pursue.
Imposition of an admissibility requirement does not comport with the United States Supreme Court’s longstanding recognition that impeachment evidence may be favorable and material, and if so, is unquestionably subject to Brady disclosure. The Court stated definitively in Strickler that “[o]ur cases make clear that Brady’s, disclosure requirements extend to materials that, whatever their other characteristics, may be used to impeach a witness.” 527 U.S. at 282 n.21, 119 S.Ct. 1936 (emphasis added). As to both the first Brady prong, favorability, and the third Brady prong, materiality, the Supreme Court has held that impeachment evidence falls under Brady’s purview. Id. at 281-82, 119 S.Ct. 1936 (the evidence “must be favorable to the accused, either because it is exculpatory, or because it is impeaching.”); Kyles, 514 U.S. at 445, 115 S.Ct. 1555 (concluding that evidence was material because “the effective impeachment of one eyewitness can call for a new trial even though the attack does not extend directly to others”). Further, nearly all of the cases decided by the United States Supreme Court since Brady have dealt with impeachment evidence. See Wearry v. Cain, — U.S.-, 136 S.Ct. 1002, 194 L.Ed.2d 78 (2016) (per curiam), Wetzel v. Lambert, -U.S.-, 132 S.Ct. 1195, 182 L.Ed.2d 35 (2012); Smith v. Cain, — U.S. -, 132 S.Ct. 627, 181 L.Ed.2d 571 (2012); Cone v. Bell, 556 U.S. 449, 129 S.Ct. 1769, 173 L.Ed.2d 701 (2009); Banks v. Dretke, 540 U.S. 668, 124 S.Ct. 1256, 157 L.Ed.2d 1166 (2004); Strickler v. Greene, 527 U.S. 263, 119 S.Ct. 1936, 144 L.Ed.2d 286 *309(1999); United States v. Bagley, 473 U.S. 667, 105 S.Ct. 3375, 87 L.Ed.2d 481 (1985); United States v. Agurs, 427 U.S. 97, 96 S.Ct. 2392, 49 L.Ed.2d 342 (1976); Moore v. Illinois, 408 U.S. 786, 92 S.Ct. 2562, 33 L.Ed.2d 706 (1972); Giglio v. United States, 405 U.S. 150, 92 S.Ct. 763, 31 L.Ed.2d 104 (1972); Giles v. Maryland, 386 U.S. 66, 87 S.Ct. 793, 17 L.Ed.2d 737 (1967). It would be difficult to find stronger support for the proposition that admissibility is not a requirement under Brady, and the Supreme Court’s repeated consideration of impeachment material in Brady cases — without any reservation whatsoever — compels us to conclude that it is unreasonable to graft an admissibility requirement onto Brady’s traditional three-pronged inquiry.
Beyond the recognition that impeachment evidence is covered by Brady, the essence of the United States Supreme Court’s Brady jurisprudence focuses on the benefits of disclosure to the defense, not admissibility. This is evidenced by the definition of materiality itself. Kyles provides that evidence is material “if there is a reasonable probability that, had the evidence been disclosed to the defense, the result of the proceeding would have been different.” 514 U.S. at 433-34, 115 S.Ct. 1555 (1995) (quoting Bagley, 473 U.S. at 682, 105 S.Ct. 3375 (opinion of Blackmun, J.)) (emphasis added). Quite simply, under Brady, the focus of the inquiry is on whether the information had “been disclosed to the defense,” not whether it was admissible at trial. See id. An admissibility requirement improperly shifts that focus.
The United States Supreme Court’s focus on disclosure is mirrored in the way in which it has applied the “reasonable probability” standard used to assess materiality under Brady. When the Court has reviewed applications of the “reasonable probability” standard, it has weighed the strength of the suppressed evidence against the strength of disclosed evidence to evaluate its. impact, not critiqued the character of the evidence itself. See Strickler, 527 U.S. at 290-94, 119 S.Ct. 1936. In Strickler, the Court denied a Brady claim on materiality grounds because “the record provides strong support for the conclusion that petitioner would have been convicted of capital murder and sentenced to death, even if [an eyewitness] had been severely impeached.” Id. at 294, 119 S.Ct. 1936. Thus, the focus was on disclosure, given the effect of other available material, not the character of the material itself.
The Supreme Court’s later decision in Cone v. Bell similarly affirmed its longstanding focus on disclosure regardless of admissibility at trial. There, the Court considered impeachment evidence including police bulletins, statements contained in official reports, and FBI reports to be Brady material. Cone, 556 U.S. at 470-71, 129 S.Ct. 1769. Neither the Sixth Circuit nor the District Court below fully considered whether the suppressed documents would have persuaded the jury to impose a lesser sentence. Id. at 475, 129 S.Ct. 1769 (“It is possible that the suppressed evidence, viewed cumulatively, may have persuaded the jury that Cone had a far more serious drug problem than the prosecution was prepared to acknowledge, and that Cone’s drug use played a mitigating, though not exculpating, role in the crimes he committed.”). Cone held that the courts below had failed to “thoroughly review the suppressed evidence or consider what its cumulative effect on the jury would have been” regarding Cone’s sentence. Id. at 472,129 S.Ct. 1769. By remanding the case for full consideration of the Brady claim despite the fact that the suppressed evidence was not necessarily admissible, the Court indicated that the admissibility of suppressed evidence ought not to change the materiality inquiry itself, which is understood as “a reasonable probability that, had the evidence been disclosed, the result *310of the proceeding would have been different.” Id. at 470, 476, 129 S.Ct. 1769.
Our recent decision in Johnson v. Folino, 705 F.3d 117 (3d Cir. 2013) further affirms the view that inadmissible evidence is often very material:
[Inadmissible evidence may be material if it could have led to the discovery of admissible evidence. Furthermore ... we think that inadmissible evidence may be material if it could have been used effectively to impeach or corral witnesses during cross-examination. Thus, the admissibility of the evidence itself is not dispositive for Brody purposes. Rather, the inquiry is whether the undisclosed evidence is admissible itself or could have led to the discovery of admissible evidence that could have made a difference in the outcome of the trial sufficient to establish a “reasonable probability” of a different result.
Id. at 130 (citations omitted). Here, however, the Pennsylvania Supreme Court ignored how the United States Supreme Court has evaluated materiality and instead made inadmissibility a determinative factor, indeed, the determinative factor.
The Pennsylvania Supreme Court’s characterization of admissibility as a separate, independent prong of Brady effectively added admissibility as a requirement. This runs afoul of Supreme Court precedent. The Pennsylvania Supreme Court required “evidence sought under Brady be material and admissible.” Dennis III, 950 A.2d at 968 (emphasis added). The Supreme Court has never added a fourth “admissibility” prong to Brady analysis. Like the imposition of a due diligence prong, adding an admissibility prong would alter Brady’s, traditional three-prong inquiry in a manner that the Supreme Court rejected in Williams. See Williams, 529 U.S. at 393, 120 S.Ct. 1495.
Most federal courts have concluded that suppressed evidence may be material for Brady purposes even where it is not admissible. See United States v. Morales, 746 F.3d 310, 314 (7th Cir. 2014) (listing cases). However, the Seventh and Fourth Circuits have indicated that inadmissible evidence cannot be material. Morales, 746 F.3d at 314; see also Jardine v. Dittmann, 658 F.3d 772, 777 (7th Cir. 2011) (“Logically, inadmissible evidence is immaterial under [the Brady] rule”); Hoke v. Netherland, 92 F.3d 1350, 1356 n.3 (4th Cir. 1996). Jar-dine and Hoke involved evidence that was prohibited from being used under state evidence laws and their assertions regarding an admissibility requirement were not determinative to their holdings. Jardine, 658 F.3d at 777 (noting that the undisclosed material was inadmissible under state law and could not be used to impeach, but concluding that no Brady violation occurred only after evaluating other avenues through with the material could be used); Hoke, 92 F.3d at 1355-56 (holding that the undisclosed information about the murder victim’s sexual history would not have been material in light of overwhelming physical and other evidence and resolving the case on grounds other than admissibility). Morales is similarly unpersuasive, as it observed that the Courts of Appeals for the First, Second, Third, Sixth, and Eleventh Circuits have read Brady to include material but inadmissible evidence. 746 F.3d at 314. The Morales court even conceded that “[w]e find the Court’s methodology in Wood to be more consistent with the majority view in the courts of appeals than with a rule that restricts Brady to formally admissible evidence.” Id. at 315.27
*311The Frazier documents were material under Brady. Dennis’s counsel could have used the information contained in the Frazier documents to challenge detectives at trial regarding their paltry investigation of the lead. As we previously noted, the lead was “fruitless” because the Commonwealth failed to take sufficient action to determine if it was fruitful — the Commonwealth essentially abandoned it. The Commonwealth does not dispute that trial counsel could have used the information in the suppressed documents to question the detectives.
Further, had the Commonwealth not suppressed the Frazier documents, Dennis could have presented an “other person” defense at trial, which he was otherwise not able to do. The Frazier documents bring to light that Walker admitted to going to Olney High School — -the school Williams and Howard attended — and he recognized Williams from school. Thus, the documents not only support an alternative shooter theory, but the very same alternative shooter theory that defense counsel could have been prepared to raise had the Howard activity sheet also been disclosed. Alterations in defense preparation and cross-examination at trial are precisely the types of qualities that make evidence material under Brady. Consequently, it was, unreasonable for the Pennsylvania Supreme Court to conclude that the Frazier documents were not material. There is a reasonable probability that had the jury heard an “other person” defense, the result of the proceeding would have been different.
The Pennsylvania Supreme Court unreasonably applied federal law and applied law in a manner contrary to Supreme Court precedent. The Commonwealth’s suppression of the Frazier documents violated Brady as they were favorable to the defense, and could have been used by defense counsel as exculpatory and impeachment evidence. Dennis is entitled to a new trial.
D. Cumulative Materiality
While the suppression of the Ca-son receipt, the Howard police activity sheet, and the Frazier documents support ordering a new trial, the cumulative effect of their suppression commands it. Had the Brady material been disclosed, there is a reasonable probability that the outcome of the trial would have been different, and its *312suppression undermines confidence in the verdict.
The District Court engaged in a cumulative materiality analysis in addition to granting each individual Brady claim. Dennis V, 966 F.Supp.2d at 517-18. This analysis was proper. When the issue ripened in Dennis IV and the Pennsylvania Supreme Court could have assessed the cumulative prejudice of withholding the Cason receipt, Frazier documents, and police activity sheet containing Howard’s statements, it declined to do so explicitly. We are required to presume that the state court considered and rejected Dennis’s cumulative materiality argument. Johnson v. Williams, — U.S. -, 133 S.Ct. 1088, 1097, 185 L.Ed.2d 105 (2013). Just as the Pennsylvania Supreme Court’s rejections of Dennis’s Brady claims constituted unreasonable application of federal law, its rejection of the cumulative materiality of the suppressed evidence, though not done explicitly, was an unreasonable application of Brady and its progeny.
The Supreme Court in Kyles instructed that the materiality of withheld evidence must be “considered collectively, not item by item.” 514 U.S. at 436, 115 S.Ct. 1555. The importance of cumulative prejudice cannot be overstated, as it stems from the inherent power held by the prosecution, which motivated Brady. Id. at 437, 115 S.Ct. 1555 (“[T]he prosecution ... alone can know what is undisclosed[ ] [and] must be assigned the consequent responsibility to gauge the likely net effect of all such evidence and make disclosure when the point of ‘reasonable probability1 is reached.”). The Supreme Court recently reiterated that state courts are required to evaluate the materiality of suppressed evidence cumulatively. Wearry, 136 S.Ct. at 1007 (“[T]he state postconviction court improperly evaluated the materiality of each piece of evidence in isolation rather than cumulatively.”)
As acknowledged by the District Court, the cumulative impeachment value of the suppressed evidence would have undermined the Commonwealth’s case. The Ca-son receipt would have impeached the Commonwealth’s primary response to Dennis’s alibi by providing documentary proof that Cason testified falsely and would have transformed her into a witness for the defense. The inconsistent statement contained in the police activity sheet would have impeached Howard’s credibility, undoubtedly the Commonwealth’s most important eyewitness. Her impeachment by the Pugh statement would challenge her credibility, not simply the reliability of her identification during the photo array and lineup, which was what defense counsel was limited to at trial. Discrediting Cason and Howard may very well have raised sufficient doubt among the jury to acquit Dennis. Moreover, the Frazier documents could have supported the existence of another suspect who attended Howard’s high school, and the significance of this becomes even more pronounced when considered with Howard’s statements to the Pughs that the suspect attended her high school.
Together, the suppressed documents provided ample material to challenge the Commonwealth’s investigation following the murder. As the District Court stated:
Defense would have had a strong case to make that the Commonwealth abandoned promising leads: Police failed to meet with Frazier’s aunt, to verify Walker’s alibi, or to include Walker and Brown in photo arrays or line-ups; police also failed to follow up with Howard about the statement she allegedly made to the Pughs, to take a formal statement from the Pughs, or to interview Quinton. The Commonwealth allowed Cason to testify incorrectly that she worked until 2 p.m., and failed to investigate Den-nises] alibi given the actual timing of *313Cason’s activities. Discrediting the investigation is a crucial corollary to presenting an innocence/alibi defense: If the defense could lead the jury to believe that the Commonwealth conducted a shoddy investigation, the jury would have been more likely to listen to and believe Dennis’fs] alibi.
Dennis V, 966 F.Supp.2d at 518. The withholding of the Brady material would have given defense counsel unique ability to discredit the Commonwealth’s primary witnesses, bolster his alibi defense using objective documentary support from a disinterested party, highlight the shoddiness of the Commonwealth’s investigation, and perhaps point to another perpetrator. The cumulative effect of the suppression of these documents requires habeas relief.
IV. Conclusion
For the foregoing reasons, we will affirm the judgment of the District Court and grant Dennis a conditional writ of habeas corpus. Petitioner shall be released unless the Commonwealth commences a new trial against him within ninety days after issuance of the mandate.
. The Fern Rock SEPTA station is located in North Philadelphia. The Abbottsford projects are located in Northwest Philadelphia.
. Detective Jastrzembski testified at trial that neither the alleged second individual nor the person in the car were ever arrested, although the case was ongoing.
. Chief Judge McKee's masterful concurrence summarizes with great detail the photo array, line up, and the bystanders' identifications. As Chief Judge McKee notes, a majority of the nine eyewitnesses who viewed the photo array were unable to identify Dennis. Anthony Overstreet was installing stone facing on a nearby garage with Bertha at the time of the incident. Overstreet told police that he recognized the shooter from around Broad and Olney Streets in North Philadelphia. Although Overstreet stated that Dennis looked like the shooter when he reviewed the photo array, he identified a different individual as the shooter during a later in-person lineup — not Dennis. George Ritchie, who was across the street from Bertha and Overstreet, was unable to identify anyone as the shooter among photos provided by the police, despite initially asserting that he would be able to identify the perpetrators again. The two fruit vendors Howard ran toward, David Leroy, a hot dog vendor near the station, and Clarence Verdell, a bystander on the SEPTA steps, did not identify Dennis from the photo array. None of these bystanders were called to testify at trial.
. The District Court reasoned that the eyewitnesses' memory may have been supplanted by photos from the array: "That some (but notably not all) of the witnesses went on to identify Mr. Dennis in a life [sic] lineup two months after providing only tentative photo array identification indicates that their memories of the photo array may have 'replaced' their memories of the actual event. Or, more simply, that Mr. Dennis was familiar to them because they had seen his photo previously, and had no prior exposure to the other members of the lineup." Dennis v. Wetzel, 966 F.Supp.2d 489, 492 n.4 (E.D. Pa. 2013) (internal quotation marks and citation omitted), vacated and remanded sub nom. Dennis v. Sec'y, Pa. Dep’t of Corr., 777 F.3d 642 (3d Cir. 2015), reh’g en banc granted, opinion vacated (May 6, 2015) ("Dennis V”).
Chief Judge McKee's concurrence expands on this concern. He observes that “[ajllowing a witness to view a suspect more than once during an investigation can have a powerful corrupting effect on that witness’ memory.” J. McKee Concurring Op. at 328. Research shows "that while fifteen percent of witnesses who mistakenly identify an innocent person during the first viewing of a lineup, that percentage jumps to thirty-seven percent if the witness previously viewed that innocent person's mug shot.” Id. Here, "[t]he witnesses who identified Dennis at trial'were given not two, but three, opportunities to view Dennis. These multiple views could help explain why initially tentative guesses became certain identifications by the time the witnesses took the stand.” Id. at 329
. Detectives Manuel Santiago and William Wynn testified at trial about the eyewitnesses' prior identifications. Detective Santiago supervised the activities at the crime scene on the day of the murder and compiled a photo array to show to Howard, Bertha, and Cameron, which included eight photographs with Dennis’s photo in the first position. Dennis looked different in the photograph than at the time of arrest. Santiago did not ask Howard why she could not be sure that it was the shooter. Detective Wynn, the lineup supervisor for the Philadelphia Police, conducted the in-person lineups for Howard, Bertha, and Cameron. Defense counsel placed Dennis as number three in the lineup. All participants dressed similarly and carried large numbers for identification.
. The Commonwealth's other witnesses did not testify as to Dennis’s connection to the murder. Rather, they spoke to the emergency response to the crime (Fireman Oakes), the scene of the crime (Sergeant Fetscher), Williams's body chart (Detective Brown), and the projectile removed from her body (Detective Reinhold). Williams’s ex-boyfriend recounted a prior incident where Williams had been robbed at gunpoint for the same earrings she wore on the day of the murder. Officer Jachimowicz, a firearms expert, testified as to the type of gun that was likely used in the murder, and although he acknowledged that there were thousands of models of .32 caliber handguns, he asserted with certainty that the nickel finish Harrington Richardson 733 was probably used in the murder. Detective Dominic Mangoni transported Howard and Bertha to the lineup. Detective Thomas Perks participated in Dennis’s arrest. Williams’s mother and father, Barbara and Barry, identified their daughter and testified to her future. Dr. Sekula-Perlman, a medical examiner, ruled Williams’s death a homicide by a shot at close range. Sergeant Fetscher took information from witnesses at the scene, including Howard, Bertha, and Cameron. None of these witnesses testified substantively as to Dennis's alleged involvement in the murder.
. Defense counsel sought to discredit eyewitness testimony put forth by the Commonwealth, primarily that of Zahra Howard. However, counsel's cross-examination was confined to highlighting Howard's prior hesitation in identifying Dennis. Similarly, defense counsel’s cross-examination of Cason focused on shakiness in her " recollection; counsel had nothing to indicate that her time-line was incorrect, or that she was mistaken or testifying falsely.
. Lawrence Merriweather also testified to seeing Dennis on the day in question. Merri-weather testified that he saw Dennis between 3:00 and 3:30 p.m.
. The Commonwealth responded with character witnesses that disputed the testimony of Dennis's character witnesses.
. Dennis testified that Helen Everett, his girlfriend, told him about the rumor that he, Derrick, and Rodney, committed the murder. He testified that Derrick and Rodney spoke with the police about the murder. Neither testified at trial.
.Anthony Sheridan, a SEPTA employee called by the Commonwealth, testified that there was a K bus that left the stop near Dennis's father's house at approximately 1:56 p.m. and that it would take approximately half an hour to arrive at Henry and Midvale.
. It is not clear how counsel would have been able to obtain Cason's receipt on appeal because DPW regulations placed strict limitations on the type of information it would disclose and to whom. See 55 Pa. Code § 105.4(a)(1). Presumably, counsel would have sought permission from Cason, or assistance from Cason herself, in obtaining the receipt.
. Williams, the victim, previously dated a man named Kevin Williams.
. The Pennsylvania Supreme Court's 2004 decision, Commonwealth v. Dennis, 580 Pa. 95, 859 A.2d 1270 (2004) (“Dennis II"), is not relevant to this appeal. On December 12, 2000, Dennis filed a motion for discovery, seeking the prosecutor’s jury selection notes, and the PCRA court granted Dennis’s motion. After granting the Commonwealth’s request for reconsideration of the order, the PCRA court reinstated the discovery order on July 10, 2001. In Dennis II, the Pennsylvania Supreme Court reversed the order granting discovery of the prosecutor’s jury selection notes and remanded the case for completion of PCRA review.
. The parenthetical language here is a direct quote from the parenthetical used by the District Court in its description of Banks. See Dennis V, 966 F.Supp.2d at 514-15.
. This framing of Kyles was taken from Lambert v. Beard, 537 Fed.Appx. 78, 86 (3d Cir. 2013).
. The Commonwealth argues on appeal that the Pennsylvania Supreme Court did not make a factual finding and that the statement that the police had the receipt was merely framing for the later substantive discussion. In Bobby v. Bies, 556 U.S. 825, 129 S.Ct. 2145, 173 L.Ed.2d 1173 (2009), cited in support by the Commonwealth, the Supreme Court held that a state court’s alleged factual finding could not support issue preclusion because there was no evidence that the alleged state court finding was supported by the record at trial or on appeal and further was not necessary to the judgments made by the state court. Bies bears no relation to our case where there is ample evidence in the record that the police took possession of the receipt, as attested by Cason herself.
. Dennis’s trial counsel asserted in an affidavit he "did not specifically request a copy of the welfare check receipt from the Commonwealth, because [he] did not know of its existence,” but he had “[b]y formal motion ... requested] all exculpatory evidence be produced.” App. 1725.
. The Tenth Circuit and the D.C. Circuit agree that defense counsel’s knowledge is not at issue in Brady. Banks v. Reynolds, 54 F.3d 1508, 1517 (10th Cir. 1995) ("[T]he prosecution's obligation to turn over the evidence in the first instance stands independent of the defendant’s knowledge.... The only relevant inquiry is whether the information was exculpatory.” (internal quotation marks omitted)); accord In re Sealed Case, 185 F.3d 887, 896-97 (D.C. Cir. 1999).
. Surprisingly, several courts of appeals have endorsed some form of a due diligence requirement. For a comprehensive overview of common features of the diligence rule and where it emerged, see Kate Weisburd, Prosecutors Hide, Defendants Seek: The Erosion of Brady Through the Defendant Due Diligence Rule, 60 UCLA L. Rev. 138, 141, 147-56 (2012). Common features include that the evidence was equally available to the prosecution and the defense, that the evidence was known by the defendant, and that the relevant facts were accessible by the defendant. Id. at 153-56.
. The Second Circuit also recently recognized in a habeas case that "[t]he [United States] Supreme Court has never required a defendant to exercise due diligence to obtain Brady material.” See Lewis v. Conn. Comm'r *292of Corr., 790 F.3d 109, 121 (2d Cir. 2015). It retained its test for when evidence is not “suppressed” for Brady purposes, however. Id.
. The Commonwealth concedes that if it had the receipt, Cason would have provided little value to the prosecution and they would not have called her. Indeed, Dennis probably would have.
. The Commonwealth argues that Cason’s testimony would be duplicative of Willis Meredith’s non-alibi testimony. Willis Meredith, a friend of Dennis’s, testified that he saw Dennis at Abbottsford Homes around 2:30 p.m. Cason’s testimony is not cumulative for two reasons: (1) Willis, like Dennis’s other witnesses, was a friend and open to accusations of bias from the prosecution; and (2) Cason’s testimony was corroborated by independent documentary evidence. So, even if her testimony simply placed Dennis at Abbottsford Homes around 2:30, it did so with more evi-dentiary weight than Meredith's.
. Howard's testimony undoubtedly bore more emotional weight with the jury than the other eyewitness testimony presented at trial due to Howard’s close friendship with the victim. Because of Howard's personal connection with, and physical proximity to, Williams at the time of her murder, stress may have played a particularly damaging role in the strength of her identification. Chief Judge McKee explains in his concurrence that that stress may impair a witness's identifications. J. McKee Concurring Op. at 329-30. Here, the identification that the Commonwealth so confidently framed as sufficient to support Dennis's conviction may have suffered the greatest from the effect of stress.
. Judge Fisher concedes that Bertha and Cameron may not have been paying attention during the incident, but urges that "the gunshot focused their view and spurred them into action.” J. Fisher Dissent Op. at 366. As Chief Judge McKee's concurrence highlights, however, the presence of a weapon at a crime scene "has a consistently negative impact on both feature recall accuracy and identification accuracy.” J. McKee Concurring Op. at 331. Here, the gunshot may have startled Bertha and Cameron to attention, but research demonstrates that the accuracy of their recollection of the perpetrators would have been reduced, not amplified, by the presence of the silver handgun.
. While this matter was pending before the panel, the government located Frazier in federal prison and interviewed him. During this interview, Frazier admitted the story he told police in 1991 was, in his words, "bullshit,” that the "three-way” phone call with his aunt and "Tony Brown” "never happened,” and that he did not know anyone named “Tony Brown” or "Skeet.” Response to Pet. Rh’g at 17 n.13. Ultimately, Frazier's admission many years post-trial does not change our analysis of whether, given the information the Commonwealth had at the time of the tip, they were required to disclose the lead documents pursuant to Brady.
. Although the United States Supreme Court recently recognized that circuit splits may indicate a possibility of fairminded disagreement under AEDPA, it did so where the circuit split emerged out of an express reservation left by the Supreme Court on the *311precise question decided by the state court. In White v. Woodall, the Kentucky Supreme Court decided that a no-adverse inference instruction, required by the Fifth Amendment to protect a non-testifying defendant at the guilt phase, is not required at the penalty phase. - U.S. -, 134 S.Ct. 1697, 1701, 188 L.Ed.2d 698 (2014) reh’g denied, - U.S. -•, 134 S.Ct. 2835, 189 L.Ed.2d 799 (2014). In so doing, the Kentucky Supreme Court relied on the Supreme Court’s decision in Mitchell v. United States, 526 U.S. 314, 119 S.Ct. 1307, 143 L.Ed.2d 424 (1999), to support its denial. Mitchell included an express reservation on the question the state court decided. See White, 134 S.Ct. at 1703. In the wake of reservation in Mitchell, "[t]he Courts of Appeals ... recognized that Mitchell left [the sentencing question] unresolved; their diverging approaches to the question illustrate the possibility of fairminded disagreement.” White, 134 S.Ct. at 1703 n.3. Thus, the United States Supreme Court opined that the Kentucky Supreme Court’s rejection of respondent's Fifth Amendment claim was not objectively unreasonable because there was an intentional lack of guidance from the Court. The United States Supreme Court has made no such express reservations when it comes to Brady materiality or an admissibility requirement. Consequently, to the extent that language from our sister circuits might be read to recognize a general admissibility requirement in Brady, we respectfully conclude that they have erred. Discrepancies as to the interpretation of Wood ought not to substantiate the Pennsylvania Supreme Court’s erroneous application of the Brady materiality standard in this case.
.Watkins v. Sowders, 449 U.S. 341, 352, 101 S.Ct. 654, 66 L.Ed.2d 549 (1981) (Brennan, J., dissenting) (alterations and emphasis in original) (quoting Elizabeth Loftus, Eyewitness Testimony 19 (1979)).
. Int'l Ass'n of Chiefs of Police, Training Key No. 600: Eyewitness Identification 5 (2006), available at http://www.ripd.org/Documents/ APPENDIX/2/Supporting% 20Materials/IP% 20113% 20IACP% 202006.pdf.