delivered the opinion of the court, in which CLAY, J., joined. SILER, J. (pp. 607-09), delivered a separate opinion concurring in part and dissenting in part.
OPINION
SUTTON, Circuit Judge.A jury convicted Rayshawn Johnson of the murder of Shanon Marks and sentenced him to death. The Ohio courts upheld his conviction and sentence on direct review. In reviewing Johnson’s habe-as petition, the district court rejected all of Johnson’s claims but one: It granted the writ on the ground that he received ineffective assistance of counsel at the penalty phase of his trial. We agree and affirm.
I.
On Wednesday, November 12, 1997, Norman Marks prepared for work, roused his wife Shanon from bed at 6:50 a.m. and left their Cincinnati home without locking the back door. While Shanon was still in her bathrobe, Rayshawn Johnson, wearing gloves and carrying a baseball bat, jumped the fence that separated the Marks’ residence from his grandmother’s home (where he lived), entered the back door of the Marks’ house, climbed the stairs to the second floor bathroom and murdered Sha-non by repeatedly striking her with the bat. When Norman returned home at approximately 8:00 p.m. that evening, he found Shanon lying face down on the bathroom floor with the contents of her purse strewn across the bed.
It did not take long to link Johnson to the crime. After Johnson appeared on a number of local news broadcasts expressing his dismay at the crime, officers interviewed him and, during the interview, noticed that the soles of his Nike Air Jordans resembled a shoe print lifted from the crime scene. After the broadcasts aired, another woman, Nicole Sroufe, called the police and told the officers that Johnson had robbed her two months earlier. The investigation closed when the officers called Johnson to the police station, where *595he waived his Miranda rights and confessed to the murder.
An Ohio grand jury indicted Johnson on several counts, including aggravated murder with two capital specifications — that the murder occurred during the commission of an aggravated robbery and an aggravated burglary. In May 1998, a jury convicted Johnson on all counts and recommended the death penalty. The trial judge accepted the recommendation and sentenced Johnson to death.
Johnson appealed his conviction to the Ohio Supreme Court, arguing (among other things) that misconduct by the judge and the prosecutor denied him a fair trial and that his counsel conducted a constitutionally ineffective penalty-phase investigation. See State v. Johnson, 88 Ohio St.3d 95, 723 N.E.2d 1054, 1072-73, 1076 (Ohio 2000), cert. denied, 531 U.S. 889, 121 S.Ct. 212, 148 L.Ed.2d 149 (2000). The court rejected each claim and independently determined that Johnson’s death penalty was appropriate. Id. at 1077-78. In state post-conviction proceedings, Johnson raised a claim of judicial bias and renewed his ineffective-assistance claim, submitting additional evidence and affidavits. State v. Johnson, No. C-000090, 2000 WL 1760225, at *3-9, *11 (Ohio Ct.App. Dec.1, 2000). The Ohio Court of Appeals denied each claim, id., and the Ohio Supreme Court denied review, 91 Ohio St.3d 1481, 744 N.E.2d 1194 (Ohio 2001).
Johnson petitioned for a writ of habeas corpus. The district court granted relief on Johnson’s ineffective-assistance claim but denied his other claims. When the warden appealed the district court’s grant of the writ, Johnson (after receiving a certificate of appealability) cross-appealed the denial of his judicial-misconduct, judicial-bias and prosecutorial-misconduct claims as well as the denial of his motion to amend the petition.
II.
Because Johnson filed his habeas petition after the effective date of the Antiter-rorism and Effective Death Penalty Act (AEDPA), we may grant the writ with respect to claims “adjudicated on the merits in State court proceedings” only if the state-court 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). We start by addressing the claims denied by the district court because they affect Johnson’s conviction and sentence.
A.
1.
Johnson premises his judicial-misconduct claim on the ground that the state-court trial judge gave a ride to a juror who had missed her bus. After the first day of trial, Johnson returned to his cell and watched (from his window) as Judge Ruehlman stopped at the bus stop and picked up a female juror who had been waiting there. The next morning, Johnson told his attorneys about the encounter, after which they raised the issue with the judge. The judge explained that the juror had missed her bus, that she was alone in the downtown area and that he offered her a ride because it was about to start raining. After the judge stopped his car, he explained, the juror confirmed with another bus driver that she had missed the last bus heading in her direction, and the judge said, “Get in, and we won’t talk about the case.” JA 2688. At defense counsel’s request, the judge brought the juror into his chambers and asked her if they had *596“talk[ed] about the case.” JA 2696. “No, not at all,” the juror insisted. Id. After giving the defense attorneys a chance to question the juror, the judge denied their request to excuse the juror, agreeing with the prosecution that the gesture was “a common act of courtesy” and caused no harm because they “didn’t talk about the case.” JA 2696, 2698.
While we think it odd, unwise, maybe even improper, for the trial judge to have given the juror a ride, the Ohio Supreme Court did not unreasonably apply federal law in holding that Johnson still received a fair trial. Not every “ex parte conversation between a trial judge and a juror” violates the Constitution, United States v. Gagnon, 470 U.S. 522, 526, 105 S.Ct. 1482, 84 L.Ed.2d 486 (1985), and communications with a juror are presumptively prejudicial only when they concern “a matter pending before the jury,” Remmer v. United States, 347 U.S. 227, 229, 74 S.Ct. 450, 98 L.Ed. 654 (1954). The trial judge gave Johnson a hearing on the impact of the ex parte contact, ef. Remmer, 347 U.S. at 229, 74 S.Ct. 450, and during that hearing both the judge and the juror stated several times that they did not talk about the case, see, e.g., JA 2689 (Judge Ruehlman stating “[w]e didn’t talk about the case at all”); JA 2694 (Judge Ruehlman saying “we didn’t talk about anything”); JA 2696 (“I’m saying as an officer of the Court I didn’t talk about the case.”); JA 2696-97 (Juror Miller answering “No” repeatedly when asked whether she and the judge discussed the case). On top of all this, the judge explained that he merely wanted to help a juror in need.
In certain circumstances, we recognize, ex parte communications with jurors have special risks, such as when the jury is deliberating. See, e.g., United States v. U.S. Gypsum Co., 438 U.S. 422, 460, 98 S.Ct. 2864, 57 L.Ed.2d 854 (1978) (“Any ex parte meeting or communication between the judge and the foreman of a deliberating jury is pregnant with possibilities for error.”). But no one suggests that this contact occurred at such a sensitive stage of the case.
The trial judge did not help matters, as Johnson points out, when he failed to report the contact first and when he failed to assign another judge to investigate the propriety of, and prejudice resulting from, the juror contact. But the uncontradicted fact remains that the judge and juror did not talk about the case, and these after-the-fact actions of the judge remain consistent with his underlying belief, naive though it might have been, that this kind of helping-hand juror contact need not be disclosed or investigated seriously. And while the assignment of another judge to preside over the Remmer hearing undoubtedly would have helped clear the air of even a hint of impropriety, the reality is that Johnson made no such request and, at any rate, the Ohio Supreme Court reviewed the trial judge’s actions. As to both claims, moreover, Johnson remains in the rut of failing to identify a Supreme Court precedent that requires reversal or that the state courts unreasonably applied. We affirm the district court’s rejection of this claim.
2.
Johnson next claims that the trial judge held a bias against him and prejudged his guilt. In support, he relies on an affidavit in which a juror says that at times he “felt that the defense’s argument wasn’t going anywhere” and the judge’s “facial expressions” sometimes “seemed to confirm” that belief, JA 3786, as well as on the following statements by the judge: (1) a statement by the judge that it was “amazing” that Johnson continued to watch females from his cell window in a manner that resembled his observation of Shanon Marks and *597Nicole Sroufe before he committed the crimes, JA 2693; (2) the judge’s comment that Johnson’s alleged assault of Sroufe was “chillingly similar” to the Marks murder, JA 1561; (3) a statement by the judge that Johnson’s story about a potential co-murderer named Dante “doesn’t have much credibility,” JA 3044; (4) the judge’s use of the term “we” to describe what Johnson claims were the prosecutor’s positions, JA 2898, 3228; (5) an observation that Johnson “was clearly enjoying himself’ during taped statements, JA 1562; (6) the fact that in ruling on Johnson’s Batson challenge the judge came up with his own reasons to justify the strike never offered by the prosecution; (7) the judge’s complaints about the federal courts’ tendency to overturn or delay capital sentences; and (8) the judge’s statement that certain objections would “go up to the Supreme Court on appeal,” JA 2520.
The state court of appeals did not unreasonably apply federal law when it held that the juror’s affidavit was improper evidence of a juror’s mental processes under Ohio Rule of Evidence 606(B). See Johnson, 2000 WL 1760225, at *11; cf. Tanner v. United States, 483 U.S. 107, 121-26, 107 S.Ct. 2739, 97 L.Ed.2d 90 (1987) (describing and applying a functionally similar federal rule). As for the judge’s statements, none of them establishes that the judge held an actual bias against Johnson or evinces a predisposition “so extreme as to display clear inability to render fair judgment.” Liteky v. United States, 510 U.S. 540, 551, 114 S.Ct. 1147, 127 L.Ed.2d 474 (1994); see also Getsy v. Mitchell, 495 F.3d 295, 311 (6th Cir.2007) (en bane). Most of the comments were innocuous and made in the course of ruling on contested issues. All of the statements occurred outside the presence of the jury. And some were not even directed at Johnson’s case. See, e.g., JA 3413 (complaining about federal courts’ death-penalty habeas practices).
One of the judge’s comments may well have been inappropriate because it denigrated the fact that Johnson reported an incident in which the judge was involved. See JA 2693 (comparing Johnson’s observation of the female juror getting into the judge’s car to his actions leading up to the crimes of which he was accused). But, by itself, the statement does not display the type of bias that prevents a fair adjudication.
3.
Johnson next maintains that several alleged instances of prosecutorial misconduct require a new trial on due-process grounds. “Even if the prosecution’s conduct was improper or even universally condemned, this court can only reverse if the statements were so flagrant as to render the entire trial fundamentally unfair.” Gillard v. Mitchell, 445 F.3d 883, 897 (6th Cir.2006). Some of the targeted statements are quite wide of this mark, as they amount to legitimate arguments premised on properly admitted evidence. The prosecutor’s statement during voir dire, for example, that the jury should recommend a life sentence in the event “that the mitigation [evidence] outweighs the aggravation” evidence, JA 2122, did not misstate the law. The statement simply conveyed to the jurors what their duty would be if the balancing played out that way, and it did not tell them one way or another what to do if the factors remained in equipoise. The same is true of other statements made by the prosecution during closing arguments, as they too were premised on fair inferences from the evidence. See, e.g., JA 3335 (saying that nothing “in any way mitigates or explains or justifies what the defendant did”); JA 3355 (calling Johnson a “cold-blooded, vicious, unremorseful killer”).
*598The prosecutor, we acknowledge, would have done well to refrain from making certain other statements. See JA 3070 (calling Johnson’s story passing blame to a co-conspirator named Dante “bull crap”); JA 3083 (noting that Johnson, who did not testify, “never once showed any remorse” during the trial and “didn’t even flinch” when pictures of Shanon Marks were shown); JA 3331 (“doubt[ing that] anyone will ever hear a case where you have a more evil defendant than Rayshawn Johnson or more innocent victim than Shanon Marks”); JA 3355 (describing a human tendency to assume that individuals who commit outrageous crimes “ha[ve] to be crazy”); JA 3353 (downplaying the mitigation testimony of Johnson’s foster mother by saying that “people that knew Adol[f] Hitler when he was three years old [probably] ... would say nice things” about him). The potential problem with several of these statements is not that the topic falls outside the range of argument; it is the manner in which the prosecutor made his points. Still, the comments were not egregious and were isolated, and any risk flowing from them was accounted for by instructions from the trial judge. See, e.g., JA 3047 (instructing the jury that they could not hold Johnson’s decision not to testify against him); JA 3364 (instructing the jury that the “arguments of counsel are not evidence”). These statements thus did not so “infect[ ] the trial with unfairness as to make the resulting conviction a denial of due process.” Donnelly v. De-Christoforo, 416 U.S. 637, 643, 94 S.Ct. 1868, 40 L.Ed.2d 431 (1974). In the end, in holding that “[t]he instances of alleged misconduct, taken singly or together, did not substantially prejudice appellant or deny him a fair trial,” Johnson, 723 N.E.2d at 1073, the Ohio Supreme Court did not unreasonably apply federal law.
4.
Johnson next argues that the district court abused its discretion, see Coe v. Bell, 161 F.3d 320, 342 (6th Cir.1998), in denying his motion to amend his petition to add a claim that he was denied an impartial jury. The premise of the new claim was Janet Miller’s deposition testimony to the effect that she became nervous when Johnson watched her from his jail window. Yet Johnson waited nearly four years from the filing of his petition to file his motion to amend, even though all of the evidence on this point had been available for years. Granting the motion also would have prejudiced the State by prolonging the case and mandating new rounds of briefing, all when Johnson acknowledges he did not have a “compelling reason” for the delay. Br. at 135. No abuse of discretion occurred.
B.
That brings us to Johnson’s sentencing claim: that his attorneys failed adequately to investigate his childhood, violating his right to effective assistance of counsel at the penalty phase. To prevail, Johnson must show that his representation was constitutionally deficient and that he was prejudiced by the poor representation. Strickland v. Washington, 466 U.S. 668, 687, 104 S.Ct. 2052, 80 L.Ed.2d 674 (1984). The district court held that he satisfied both prongs, and so do we.
1.
Under the deficient-performance prong, we “indulge a strong presumption that counsel’s conduct falls within the wide range of reasonable professional assistance.” Id. at 689, 104 S.Ct. 2052. Strong though that presumption is, “strategic choices made after less than complete investigation are reasonable” only “to the extent that reasonable professional *599judgments support the limitations on investigation.” Id. at 690-91, 104 S.Ct. 2052. We review the decision of the last state court to reach this point, the state court of appeals, for compliance with AEDPA. See Garcia v. Andrews, 488 F.3d 370, 374 (6th Cir.2007). And in judging the reasonableness of the adjudication, we look to Wiggins v. Smith, 539 U.S. 510, 123 S.Ct. 2527, 156 L.Ed.2d 471 (2003), and Rompilla v. Beard, 545 U.S. 374, 125 S.Ct. 2456, 162 L.Ed.2d 360 (2005) — even though the state court’s decision predated both opinions-because they did not rest on “new law” but instead “applied the same ‘clearly established’ precedent of Strickland.” Wiggins, 539 U.S. at 522, 123 S.Ct. 2527; see also Rompilla, 545 U.S. at 380-81, 125 S.Ct. 2456; cf. Jells v. Mitchell, 538 F.3d 478, 491 n. 2 (6th Cir.2008).
In our view, the state court of appeals unreasonably applied these precedents in rejecting this claim. But before explaining why Johnson’s lawyers committed legal malpractice — or, what is worse, legal representation that amounts to constitutionally ineffective assistance — it is well to put in context the options they faced after the jury came back with a guilty verdict against their client on the underlying murder charges and capital specifications. According to the ABA Guidelines on death-penalty representation, upon which the Supreme Court has relied before, see Rom-pilla, 545 U.S. at 387 & n. 7, 125 S.Ct. 2456, counsel should consider the following in deciding what witnesses and evidence to introduce at the penalty phase of a capital case:
1.Witnesses familiar with and evidence relating to the client’s life and development, from conception to the time of sentencing, that would be explanatory of the offense(s) for which the client is being sentenced, would rebut or explain evidence presented by the prosecutor, would present positive aspects of the client’s life, or would otherwise support a sentence less than death;
2. Expert and lay witnesses along with supporting documentation (e.g., school records, military records) to provide medical, psychological, sociological, cultural or other insights into the client’s mental and/or emotional state and life history that may explain or lessen the client’s culpability for the underlying offense(s); to give a favorable opinion as to the client’s capacity for rehabilitation, or adaptation to prison; to explain possible treatment programs; or otherwise support a sentence less than death; and/or to rebut or explain evidence presented by the prosecutor;
3. Witnesses who can testify about the applicable alternative to a death sentence and/or the conditions under which the alternative sentence would be served;
4. Witnesses who can testify about the adverse impact of the client’s execution on the client’s family and loved ones;
5. Demonstrative evidence, such as photos, videos, and physical objects (e.g., trophies, artwork, military medals), and documents that humanize the client or portray him positively, such as certificates of earned awards, favorable press accounts, and letters of praise or reference.
ABA Guidelines for the Appointment and Performance of Defense Counsel in Death Penalty Cases 10.11(F) (rev. ed.2003).
At a surface level, it appears that Johnson’s counsel considered all of these options, performed some investigation with respect to each option and deployed some of these strategies. A central theme of their ultimate penalty-phase strategy was to feature Marian Faulkner, Johnson’s grandmother, as a pivotal figure in his life, one who “did everything that one could reasonably expect to do to try [to] help” *600Johnson. JA 3287. The apparent goal was not only to humanize Johnson with Faulkner’s testimony, see ABA Guideline 10.11(F)(1), but also to present a compelling witness who would suffer from a jury decision to impose a death sentence, see Guideline 10.11(F)(4). In the abstract, this might well have been a legitimate strategic decision, one about which the Constitution would have nothing to say. But in Johnson’s case, his counsel pursued this strategy after what can only be described as an anemic and leaderless investigation that suffered from at least three conspicuous flaws.
First, Johnson’s counsel never interviewed Johnson’s mother, Demeatra, who not only could have explained the precise role that Faulkner played in Johnson’s (and her) life but also could have provided other information about Johnson’s childhood. Even though the defense team knew Demeatra’s “whereabouts,” Pete Pandilidis, the lead defense attorney, testified that the defense team chose not to interview her because she “had a very bad background ... being a prostitute and a drug addict” and therefore would be a “bad mitigation witness.” JA 219. But Demeatra’s “bad background” is precisely what should have prompted the defense team to interview her — both to see what that background entailed and to learn more fully how her prostitution and drug addiction affected Johnson’s childhood. That someone may make a bad witness is no explanation for not interviewing her first. And that is particularly true with respect to this witness, who was Johnson’s mother, not a distant aunt or neighbor. No “reasonable professional judgment,” Strickland, 466 U.S. at 690, 104 S.Ct. 2052; cf. Wiggins, 539 U.S. at 536, 123 S.Ct. 2527, could have supported a decision not to interview Demeatra, and neither the State, see First Br. at 40, nor the state court of appeals, see Johnson, 2000 WL 1760225, at *9, even attempted to justify such a decision.
Second, in attempting to investigate Johnson’s background, his counsel obtained a large number of files from the Ohio Department of Human Services but apparently never read them. Instead, they simply submitted them to the jury— unorganized and without knowing whether they hurt Johnson’s strategy or helped it. Three days before the penalty phase began, Chuck Stidham, the defense’s “mitigation specialist,” acquired the files from the agency. Yet he never looked at the records but simply handed them over to defense counsel. Pandilidis testified that he and Dixon “reviewed them,” but, because they were “about 12 inches thick,” they “didn’t read [them] word for word.” JA 4534. That, it turns out, is an understatement. So far as the record shows, it is doubtful whether they read the records sentence for sentence, even page for page, before placing them before the jury. The records included many pages that were not even relevant to Johnson or his family, something Pandilidis acknowledges he did not realize at the time. Had Johnson’s attorneys read all of the records before placing them in front of the jury, surely something even modestly competent counsel would do, they not only would have removed the records that had no bearing on the proceeding but they also would have learned more about Faulkner. Social workers at Human Services, the records show, were concerned about placing Johnson in Faulkner’s custody because of her abusive history, information that would have prompted reasonable mitigation counsel to investigate Faulkner’s past further. A review of the records, in short, not only would have tipped them off to a different mitigation strategy but also would have avoided the pitfall of submitting records to the jury that directly contradicted their *601theory that Faulkner was a positive force for change in his life.
Third, the record suggests that these investigative blunders occurred because no one who participated in Johnson’s penalty-phase defense made any deliberate decisions about the scope of the investigation, let alone the “reasonable” ones Strickland requires. Pandilidis provided the two alleged mitigation specialists, Stidham and Steve Vonderhaar, with an initial set of names, but he admitted that he provided no significant guidance, saying, “I don’t plan the investigation. We get the mitigation experts out to do that.” JA 4524. Pandilidis’s representations to the trial court show that he and the other lead attorney, Dixon, were not involved in the investigation. On May 4, one week before the trial began, Pandilidis said that he did not know what witnesses they would call for mitigation. On May 12, the day before the prosecution’s case in chief began, he said that he did not know what investigation Vonderhaar had done, and told the State that they had “Absolutely nothing” in the way of mitigation documents, JA 2369. And on May 15, midway through the trial, defense counsel still did not “even know what records [they were] going to have.” JA 2961. Johnson’s attorneys admitted in their post-conviction testimony that they began thinking about a mitigation strategy only when “the verdict was back and [the jury] found [Johnson] guilty.” JA 4525.
Chuck Stidham, who ostensibly served as the lead mitigation specialist in this case, was no better. He did not personally conduct any of the initial investigation; rather, he “obtain[ed] whatever records were necessary,” JA 4619, and handled “most of the interaction [with] the attorneys, [while Vonderhaar] would do [the] ... background investigation stuff,” JA 4618. Stidham “ha[d] discussions” with Vonderhaar regarding “whatever information [they] had,” JA 4618-19, but he never gave Vonderhaar any direction and never gave him potential mitigation theories to pursue. Vonderhaar confirmed that “[t]here was no theory” of mitigation that Stidham “directed [him] to investigate.” JA 4487.
Stidham, moreover, was in the midst of a debilitating bout of depression — indeed at one point his secretary found him under his desk “crying [his] eyes out.” JA 4637. His health problems, as he has acknowledged, led him to “neglect ... client matters,” JA 4633, “had an impact on everything that [he] did,” JA 4634, and ultimately resulted in his suspension from the practice of law, see Cincinnati Bar Ass’n v. Stidham, 87 Ohio St.3d 455, 721 N.E.2d 977, 983-84 (Ohio 2000). Stidham, it seems clear, was almost certainly not making any significant decisions, reasonable or otherwise, regarding the scope of Johnson’s investigation.
At the end of this chain of delegation was Vonderhaar, a well-meaning graduate student who had done some work on two other capital cases and who was working part time while taking classes for his psychology degree. But Vonderhaar was apparently not making any independent investigative decisions either: He interviewed only the witnesses on the initial list given to him, and he spent only 41 hours on the entire investigation. While none of the people he interviewed gave him names of other people to talk to, the record does not show that he bothered to ask. Vonderhaar, moreover, had no contact with Dr. Hawkins, Johnson’s mitigation expert; he simply dropped off his interview notes at Dr. Hawkins’s office. One could certainly question whether a graduate student with little experience, little time and little guidance could make reasonable judgments about the scope of a capital-sentencing investigation. Even *602Pandilidis recognized that Vonderhaar needed “a little more experience, if he ... really wanted to get into the [mitigation] field.” JA 241.
This lack of structure and supervision over the investigation led to significant and costly delays, including missed appointments. See JA 1631 (Pandilidis telling the court on May 4 that “[o]ur people were supposed to meet with [Johnson] over the weekend and they have not”). Stidham also inexplicably waited until May 14 to issue subpoenas for the critical Human Services records, leaving Johnson’s attorneys insufficient time to review them. These delays “suggest that [the defense’s] incomplete investigation was the result of inattention, not reasoned strategic judgment.” Wiggins, 539 U.S. at 534, 123 S.Ct. 2527; cf. Powell v. Collins, 332 F.3d 376, 399 (6th Cir.2003) (holding in a preAEDPA case that a capital defense attorney has a duty to investigate “the circumstances of [his client’s] case and to explore all avenues relevant to the merits of the case and the penalty in the event of a conviction”) (quoting 1 ABA Standards for Criminal Justice, Standard 4-4.1 (1982 Supp.)).
The state court of appeals unreasonably applied Strickland when it held otherwise. In determining that Johnson’s counsel was constitutionally adequate, the court principally concluded that counsel had discovered the relevant evidence, as shown by the fact that “[m]ost of the additional evidence ... could be found in some form either in the testimony or in the documents admitted at the mitigation hearing.” Johnson, 2000 WL 1760225, at *9. But the testimony only scratched the surface of Johnson’s horrific childhood. And even if it is true that some aspects of Faulkner’s and Demeatra’s problematic roles in Johnson’s life could be gleaned from reviewing the 12-inch stack of files that defense counsel obtained from Human Services and admitted into evidence, that does not mean defense counsel performed a reasonable investigation or for that matter reasonably used the evidence. As we have shown, defense counsel was not familiar with the records; some of the records contradicted their mitigation strategy (e.g., those showing that Faulkner had contributed to Johnson’s traumatic childhood rather than worked to improve it); and it hardly constitutes a reasonable investigation and mitigation strategy simply to obtain Human Services records from the State, then dump the whole file in front of the jury without organizing the files, reading them, eliminating irrelevant files or explaining to the jury how or why they are relevant.
The court of appeals also unreasonably concluded that Johnson’s attorneys “presented a meaningful concept of mitigation,” id., without looking to the reasonableness of the investigation’s scope. Johnson’s defense team, to be sure, interviewed some witnesses and submitted some testimony regarding Johnson’s troubled past: Vonderhaar interviewed five people, including Johnson, Faulkner, Norma Berry (Johnson’s foster mother), Ronnie Johnson (Johnson’s brother) and Abby (Johnson’s girlfriend). And Dr. Hawkins testified at the mitigation hearing that Johnson’s family unit was “someplace between terrible and chaotic,” JA 3286, that his neighborhood was “less than optimal,” JA 3287, and that Demeatra “was abusing drugs heavily at the time of [Johnson’s] birth,” JA 3286. But an unreasonably truncated mitigation investigation is not cured simply because some steps were taken prior to the penalty-phase hearing and because some evidence was placed before the jury. See Rompilla, 545 U.S. at 382-83, 125 S.Ct. 2456 (holding an investigation objectively unreasonable where the lawyers spoke to the defendant, five family *603members and three mental health witnesses).
Buttressed by a reasonably adequate investigation, the defense team’s ultimate presentation to the jury might have been justified as the product of strategic choice. But that is not what happened. Johnson’s attorneys “were not in a position to make ... reasonable strategic choice[s] ... because the investigation supporting their choice[s] was unreasonable.” Wiggins, 539 U.S. at 536, 123 S.Ct. 2527; see also Strickland, 466 U.S. at 690-91, 104 S.Ct. 2052.
Instead of trying to defend Johnson’s attorneys’ decision not to interview De-meatra, the State points to the federal-court testimony of Dixon, who said that the defense team did interview Demeatra and that she “wasn’t any help.” JA 314. But Dixon’s testimony contradicts Demea-tra’s statement that she was not interviewed and Pandilidis’s testimony that they never interviewed her. Dixon’s memory, moreover, was so flawed that it is difficult to understand how it could be worthy of credence: He did not recognize the name of one of their two investigators, Vonderhaar, and he did not remember whether the defense called Demeatra as a defense witness when they clearly did not. Demeatra said that she would have testified at his trial if asked, and medical records show that she was seeking professional treatment for her psychological problems at that time and therefore quite probably could have testified. Even though the district court did not make a specific factual finding regarding whether the defense interviewed Demeatra, it accepted the allegation that they did not. See JA 453 (noting Demeatra’s testimony that she was not interviewed without discussing Dixon’s testimony).
The State argues that Johnson’s attorneys cannot be held responsible for the evidence not presented to the jury because none of their interviewees “provide[d] them with” evidence of Faulkner’s abusive history or the extent of Demea-tra’s negative influence on Johnson. First Br. at 42. Uncooperative defendants and family members, however, do not shield a mitigation investigation (even under AED-PA’s deferential standards) if the attorneys unreasonably failed to utilize other available sources that would have undermined or contradicted information received. See Rompilla, 545 U.S. at 381-87, 125 S.Ct. 2456 (holding that the state court unreasonably applied Strickland even where the defendant and other witnesses told his attorneys that his childhood was “normal”).
2.
Turning to the prejudice prong of Strickland, we must determine whether Johnson has shown “a reasonable probability that, but for counsel’s unprofessional errors, the result of the proceeding would have been different.” 466 U.S. at 694, 104 S.Ct. 2052. Because the state court ended its Strickland inquiry at the deficient-performance prong for all categories of evidence relevant to our holding, see Johnson, 2000 WL 1760225, at *9 (determining only that the evidence that Johnson “did well in structured environments ... [did not] demonstrate[ ] that Johnson was prejudiced”), we “examine this element of the Strickland claim de novo,” Rompilla, 545 U.S. at 390, 125 S.Ct. 2456; see also K & T Enters., Inc. v. Zurich Ins. Co., 97 F.3d 171, 175 (6th Cir.1996).
The errors of Johnson’s attorneys, particularly their lack of investigation, had a serious impact on the mitigation theory presented to the jury. Competent counsel could have put on evidence that “differed] in a substantial way — in strength and subject matter — from the evidence actually presented at sentencing.” Hill v. Mitch*604ell, 400 F.3d 308, 319 (6th Cir.2005). As the district court found, “not one witness testified about the abuse that [Johiison] and his brother suffered as a way of life,” and the jury “was misled into believing that [Faulkner] had raised [Johnson] properly and provided for his needs.” JA 468.
First, had Johnson’s attorneys interviewed Demeatra or investigated Faulkner’s history, they would have unearthed a goldmine of mitigating evidence showing that Faulkner was anything but a saint when it came to raising Johnson. Hospital records show that doctors diagnosed Faulkner with schizoid personality disorder, which prevented her from forming intimate relationships, and Dr. Smith, Johnson’s post-conviction expert, testified that Faulkner “had no maternal instincts,” JA 4411. After trying to abort Demeatra, Faulkner “neglected her [and] rejected her.” JA 4412. Demeatra testified at an evidentiary hearing that Faulkner beat her regularly with her fists, extension cords and broom handles and that Faulkner once shot at her with a gun. Faulkner regularly held uninhibited parties at her house when Demeatra was growing up, and one night, while Faulkner was passed out naked on the bed, one of Faulkner’s friends forced his way into Demeatra’s bedroom and raped her.
As Demeatra’s son, Johnson was an indirect victim of Faulkner’s neglect and abuse, but he was a direct victim as well. Dr. Smith testified that Faulkner addressed Johnson “with the same schizoid personality traits ... that she demonstrated with Demeatra,” that she went through periods of rejecting Johnson and that her heavy substance abuse and mental illness made her even more unavailable in his life. JA 4405. Faulkner herself acknowledged at the district court’s evidentiary hearing that, in accordance with her husband’s wishes, she never showed Johnson any affection “because [otherwise he would] grow up to be a sissy.” JA 4561. Faulkner went through stages of rejecting Johnson outright — from threatening to kick Johnson and his brother out of her house when they cried too much to telling the police that he could not stay with her. She also engaged in “physically abusive” forms of discipline (by hitting Johnson with belts and switches). JA 4405. But because Johnson’s counsel did not present this evidence to the jury, as the district court found, “[t]he jury never learned of Mrs. Faulkner’s blemishes, and was misled into believing that Mrs. Faulkner was a stable influence in petitioner’s life who did everything she could to help him” and who “raised him properly and provided for his needs.” JA 467-68.
Second, had the defense team interviewed Demeatra, they could have presented a detailed and horrific picture of De-meatra’s role in Johnson’s life. At the mitigation hearing, Johnson’s attorneys portrayed Demeatra as a mostly absent mother, when the truth is that her early abuse and on-again-off-again presence in his life had an irreparable and devastating impact on Johnson. A reasonably competent mitigation counsel could have shown that: (1) Demeatra was a prostitute who sold herself to buy drugs; (2) she often fed Johnson only sugar water in a bottle; (3) when Johnson cried, she would put him in a closet (sometimes leaving him there all day) and give him beer or Percocet (a Schedule II pain killer) to make him stop crying; (4) she hit Johnson regularly and at times threatened to kill him; (5) she once put a cigarette out on his eye; (6) she once hit Johnson’s brother with a glass bottle and told the hospital that Johnson had done it; (7) she was involved in many abusive relationships; (8) Johnson watched his father beat Demeatra; (9) Demeatra tried to set fire to Johnson’s father; (10) when Johnson was an adolescent, Demea-tra taught Johnson how to cut cocaine, *605cook it into crack and sell it; and (11) she killed one of her abusive boyfriends and bragged to Johnson about doing so.
Yet, due to counsel’s bungling or sheer laziness, the jury heard none of this. “I and the public know [wjhat all schoolchildren learn,” it has been said, “[tjhose to whom evil is done [d]o evil in return.” W.H. Auden, “September 1, 1939.” While these words may not capture a satisfactory theory of morality, they assuredly suggest a plausible theory for sparing a life at a mitigation hearing, see ABA Guideline 10.11(F)(l)-(2), one that on this remarkable record could well have affected a juror’s vote in the case.
Third, the deficiencies in the attorneys’ investigation prevented them from using their mitigation expert, Dr. Hawkins, properly and ultimately led to damaging testimony from him. At the mitigation hearing, Dr. Hawkins testified that Johnson’s statements had a “poor degree of reliability,” JA 3286, that Johnson “thought pretty highly of himself,” that Johnson “thought that he could probably do and get out of most everything” and that Johnson’s “self-image is based on how well he cons the world,” JA 3294. Hawkins diagnosed Johnson as having an antisocial personality disorder, which can lead to and reflect a “history of lying ... and a lack of remorse for anything you’ve done.” JA 3302-03. Hawkins testified that Johnson was not remorseful “about the murder,” only “about being incarcerated,” and that Johnson “more than likely” would consider himself the “victim in this offense.” JA 3315.
This unhelpful testimony could have been prevented if the attorneys had conducted a competent investigation and given Hawkins a more complete picture of Johnson’s background — namely, the remarkably traumatic childhood Johnson suffered at the hands of his mother and grandmother. Hawkins then might have been able to say, as Dr. Smith did in post-conviction proceedings, that Johnson’s “chaotic, abusive, neglectful” family “play[ed] a significant role in the development of Ra-yshawn’s personality and his addiction to alcohol and drugs and later mental illness.” JA 4405-06. He could have described the cycles of generational “abusive and neglectful” parenting that “repeat[ ] the same behaviors” and lead to “the same outcomes.” JA 4408. He could have described the fact that Johnson’s psychological profile is “almost identical” to that of his mother; they both had “strong fantasy li[ves]” which is “very common in children who have suffered from abuse and whose reality is not acceptable to them.” JA 4411. He might have said, as Smith did, that “having a history of neglect, abandonment [and] abuse ... set[ ] [Johnson] up to be frightened [and] vulnerable,” JA 4410, and had a “pervasive impact” on him, giving him “little understanding of what is good or right and how to form attachments to others,” JA 4417. With a more detailed picture of Johnson’s history, Dr. Hawkins might have realized that Johnson was capable of “relatively respectful and caring relationships” in the right circumstances, JA 4434, that with the proper attention “Rayshawn did fairly well,” JA 4409, and that if Johnson had not been abused he could have learned “to act in ways that are more appropriate,” JA 4410.
Defense attorneys, we recognize, are not obligated to shop for “the ‘best’ experts” who will testify in the most advantageous way possible. Reynolds v. Bagley, 498 F.3d 549, 557 (6th Cir.2007). But it is unreasonable, after an incomplete investigation, to put an expert on the stand who will “directly contradict! ] the sole defense theory” and “render worthless” other helpful testimony. Combs v. Coyle, 205 F.3d 269, 288 (6th Cir.2000). Even the prosecution called Hawkins’s testimony *606“inflammatory,” JA 3356, and said to the jury, “I can’t believe it, but [Hawkins] testified for the defendant allegedly,” JA 3354-55. Indeed.
In all three respects, this post-conviction evidence differs from that heard by the jury not only in degree but also in kind, see Hill, 400 F.3d at 319, and forms a mitigation story that “bears no relation” to the story the jury heard, Rompilla, 545 U.S. at 393, 125 S.Ct. 2456. Strickland is based in part on the need to preserve “a reliable adversarial testing process,” 466 U.S. at 688, 104 S.Ct. 2052, but here the defense team’s incompetent performance served only to help the prosecutor’s case. Under these circumstances, there is “a reasonable probability that, but for counsel’s unprofessional errors, the result of the proceeding would have been different.” Strickland, 466 U.S. at 694, 104 S.Ct. 2052.
Nor can the State sidestep this conclusion on the ground that some of this information — particularly Faulkner’s deficiencies — could have been gleaned from the documents submitted to the jury. Even though we generally assume that the jury will consider the evidence admitted, this jury was given no basis for construing and digesting this information: (1) the Human Services records alone were “about 12 inches thick,” JA 4533, (2) Johnson’s own attorneys failed to read the records before admitting them into evidence, (3) no mitigation witness ever referred to the records, (4) Johnson’s attorneys simply told the jury to look at them if they “want[ed] to leaf through” them, JA 3341, and (5) “leaf[ing]” through the records likely would have discouraged the jury from reading more closely because the references to Faulkner’s deficiencies were few and significant portions of the records were not even relevant.
While our dissenting colleague agrees that de novo review applies to our application of the prejudice prong to Johnson’s case, see Rompilla, 545 U.S. at 390, 125 S.Ct. 2456, he ultimately concludes that Johnson has not satisfied this requirement. Judge Siler’s thoughtful disagreement with us proves that the case is a close one, but in our respectful view it does not prove that we should reverse. By failing to interview Johnson’s mother, Demeatra, Johnson’s attorneys failed to uncover two critical categories of information that establish “a reasonable probability that, but for counsel’s unprofessional errors, the result of the proceeding would have been different.” Strickland, 466 U.S. at 694, 104 S.Ct. 2052. While Johnson’s attorney featured Faulkner as a force for good in his life, later-discovered evidence showed just the opposite: A jury that imposed death under the Faulkner-as-saint approach to this case might well have reached a different conclusion had it learned that in truth she represented one more bad card that Johnson had been dealt. And by interviewing Demeatra, Johnson’s attorneys also could have discovered not just that his mother was a neglectful drug addict but also that she committed atrocities on him that the evidence introduced at trial does not begin to convey: When Johnson cried, she put him in a closet (sometimes for the whole day) and gave him beer and Percocet to stop the crying; she put a cigarette out in his eye; when he was still an adolescent, she taught him how to prepare crack cocaine and sell it; and when she hit Johnson’s brother with a glass bottle, she told the police that Johnson had done it. In our view, the failure to uncover these pieces of evidence, and others already mentioned, establish prejudice. Cf. Jells, 538 F.3d at 498-501.
III.
For these reasons, we affirm.