COOK, J., delivered the opinion of the court, in which BOGGS, C.J., joined. COLE, J. (pp. 592-97), delivered a separate dissenting opinion.
OPINION
COOK, Circuit Judge.A jury found Timothy Rosencrantz guilty of sexually assaulting Elaine Lasky. He seeks habeas relief on the ground that the prosecution, by countenancing false testimony from Lasky, violated his due process rights. As explained here, we affirm the district court’s denial of Rosencrantz’s petition because, even assuming the materiality of the testimony at issue, the prosecutorial misconduct qualifies as harmless under Brecht v. Abrahamson, 507 U.S. 619, 113 S.Ct. 1710, 123 L.Ed.2d 353 (1993).
I.
A. State Court Proceedings
The controversy in this case centers on Rosencrantz’s trial. The prosecution began by calling Sergeant William Gooch, who testified that Lasky reported a sexual assault to the police at 2:01 A.M. on August 17, 1995. Gooch stated that he met with Lasky by 2:06 A.M, and she told him that around 1:45 A.M. a clean-shaven, shirtless, white male with a car tattoo on his chest and driving a black pick-up truck assaulted her.
Lasky testified next, giving her story that she spent the evening of August 16th at a motel in Burton, Michigan with her friend Jack Pascoe. In contrast to Gooch’s testimony, she told the jury that around 10:30 or 10:45 P.M., she left her room and walked to the motel’s parking lot, where a black pick-up truck — driven by a man she identified at trial as Rosencrantz — pulled alongside her. Lasky entered the truck intending to steal Rosencrantz’s money, but instead Rosencrantz, armed with a knife, sexually assaulted her, inflicting cuts to her arms, legs, and chest in the struggle. Lasky then begged him to take her back to the motel — which he did, but not before Lasky noticed the image of a car tattooed on his chest.
According to Lasky, Rosencrantz dropped her off at the motel thirty-five to sixty minutes after she left. Back in her room, she initially hesitated to involve police because she was on probation and, she said, the motel had no phone. But Lasky went on to say that she changed her mind after an hour to an hour and fifteen minutes, during which time she consumed two bottles of wine and took “one or two” Xanax pills, an anti-anxiety medication. Another twenty minutes passed before she found a phone down the street and allowed Pascoe to call the police on her behalf.
Lasky also testified that when detectives showed her a photo line-up of six faces, she “immediately” picked Rosencrantz’s face as her rapist. Her direct examination concluded when the prosecutor asked Rosencrantz to bare his chest. Seeing a tattoo of a car, Lasky confirmed that it matched that of her assailant.
On cross-examination, defense counsel opened by questioning Lasky about her pretrial preparation:
*581Q. Did you have an opportunity to discuss your testimony with anybody prior to coming in here today?
A. No, Sir.
Q. You never talked to anybody?
A. No, sir.
Q. You didn’t go over — You weren’t over to the prosecutor’s office a few days ago, being brought in for any interviews? You weren’t around anything like that?
A. No, sir.
Q. You didn’t talk to anybody about it prior to coming in here today? You weren’t in any rooms up on the second floor talking to anybody about it?
A. No, sir, I was not.
The prosecutor remained silent during this colloquy and never reopened the topic during the trial.
Defense counsel continued cross-examining Lasky, questioning her about how her account of when the assault occurred changed between her reporting to the police (approximately 1:45 A.M.) and testifying at the preliminary hearing (approximately 1:30 A.M.), and the day of trial (approximately 10:30 P.M.). Likewise, the defense undermined Laslfys credibility by eliciting admissions that: she had a retail-fraud conviction; she lied under oath during the preliminary investigation in claiming Rosencrantz forced her into the truck at knife-point; and her story varied on whether her assailant had a beard or whether he wore a shirt. Defense counsel also impeached Lasky by contrasting part of Lasky’s preliminary-examination testimony, where she admitted drinking during the afternoon of the assault, with her trial testimony, where she insisted she was sober at the time she met Rosencrantz. And Lasky also conceded during cross-examination that she drank heavily after the assault, and suffered from a history of alcohol problems.
Aside from Gooch and Lasky, the state’s other witnesses included Jack Pascoe, whose testimony supported much of La-sky’s account, and Ellen Rogers, a pool attendant in Flint Township, Michigan, who testified that about a month before the alleged assault, she saw Rosencrantz with a car tattoo on his chest driving a dark pick-up.
Rosencrantz’s defense aimed at establishing an alibi. His girlfriend claimed to be with him in Fairview, Missouri — 822 miles away from the assault — on both August 16 and 17. Fairview restaurant owners, Linda and Gary Yanderlinden, also said they saw Rosencrantz in mid-to-late morning on the 17th. And his Fairview landlord testified to meeting with Rosencrantz about 2:00 P.M. on the 17th.
The jury found Rosencrantz guilty of first-degree sexual assault and felonious assault, and the court sentenced him to 22-1/2 to 50 years imprisonment. He succeeded in his state court appeal to the extent that the Michigan Court of Appeals vacated the felonious assault conviction. On the remaining conviction, Rosencrantz persisted in the Michigan Supreme Court, eventually exhausting his state remedies.
B. District Court Proceedings
Rosencrantz argued to the district court, and maintains on appeal, that the prosecutor failed to disclose exculpatory evidence or knowingly presented false testimony on four material points: (1) that the assault occurred shortly after 10:30 P.M., on August 16, 1995, rather than shortly after 1:30 A.M. on August 17, 1995, (2) that Lasky felt certain in identifying Rosencrantz as the attacker, (3) that she was sober at the time of the assault, and (4) that Lasky met with the prosecution before trial.
*582Rosencrantz asserts that he discovered that Lasky testified falsely through an affidavit submitted by Jan Burgess, who stated that while working in the Genesee County Sheriffs Department from August 1993 through July 1996, interviewing detainees, she had contact with Lasky in January 1995 and May 1996 (Rosencrantz’s trial took place in June 1996). With regard to her May 1996 encounters with Lasky, Burgess’s affidavit states in pertinent part:
May 1996: Re-interviewed Lasky and re-enrolled her. She said she was back in jail as a witness. She said she told the police she had been raped. She said she gave a description of the man but that she had no idea what he really looked like or the details of what happened because she had been “cracked up.”
May 1996: While working with Lasky in the 3rd floor activity room at approximately noon, a group of 5 or 6 men arrived on the floor. The activity deputy brought the men into the room and told me they were there to meet with Lasky and that I would have to leave. After leaving, I asked the deputy what was going on and who these men were. He pointed out [prosecutor] Garner Train and someone else from the prosecutor’s office. He said the others were detectives. This group surrounded La-sky and was often quite loud, although with the room closed, the deputy and I couldn’t understand what was being said. I left and returned twice more than afternoon to continue working with Lasky. Each time I returned, this group was still with Lasky (for a total time of at least 3-4 hours).
The next day, I called Lasky out to continue working with her. She was very agitated and afraid. She said she had to do what the men wanted or she felt her husband would kill her. (Note: at some time during Elaine Lasky’s incarceration, I tested and interviewed John Lasky [Elaine’s Husband], who was serving a one-year sentence. He told me he would not be in jail that long because his “old lady” was working with the cops and would get him out.) ... She said she worked with the police because her husband “made her” and she was so afraid of him she did whatever he told her to do.
Burgess Affidavit, at 2.
The district court conducted an evidentiary hearing to develop the factual record on Rosenerantz’s claim that the prosecutor knowingly presented false testimony or allowed such testimony to stand uncorrected. Burgess and Lasky both testified at the hearing.
Lasky testified, in pertinent part, that:
• she did not use drugs the day of the assault, but had been drinking;
• she was intoxicated at the time she entered the truck;
• on the day she selected Rosencrantz’s photo from a photo lineup, she was “probably still half drunk” from a night of drinking;
• she was not certain that Rosencrantz was the person who assaulted her;
• she met with the prosecutor Garner Train and several police officers approximately three times prior to trial;
• she did not recall whether she informed the police or prosecutor that she was intoxicated at the time of the assault;
• she informed police that she was not certain Rosencrantz was the person who assaulted her;
• she had no conversations with the prosecutors or investigators regarding any other cases besides the one in *583which she was the complainant against Rosenerantz.
Tr., Evid. Hearing, 11/7/2006, pp. 4-15.
Burgess testified consistent with her affidavit about the meeting she observed and her impressions from Lasky. Noting that both Burgess and Lasky testified that La-sky met with the prosecution team prior to trial, the district court found “their testimony in this regard to be credible,” and in the absence of evidence to rebut, the court concluded that the prosecution allowed false testimony to stand uncorrected when Lasky denied any pretrial meetings.
As for the other three instances, the district court found that the prosecutor did not present false testimony on those — either because the testimony failed to qualify as indisputably false, or because, if La-sky did testify falsely, the prosecutor did not know it.
This left the court with the finding that the prosecution potentially violated Rosencrantz’s due process rights by knowingly countenancing Lasky’s false denial of pretrial meetings. The court examined whether that false testimony was material, and determined that it was not. Then the court went on to decide that even if material, the false testimony meant little when viewed in the context of all the other evidence presented to the jury and therefore could be excused as harmless. In deciding to apply harmless-error review, the district court looked for guidance to our decision in Carter v. Mitchell, 443 F.3d 517, 537 (6th Cir.2006) (applying harmless error to a knowing-presentation-of-false-testimony case), and the First Circuit’s decision in Gilday v. Callahan, 59 F.3d 257, 268 (1st Cir.1995) (same). Carter, however, summarily cited harmless error as an alternative basis for denying a false-testimony claim, and this circuit has yet to explicitly hold that a knowing-presentation-of-false-testimony due process violation should be reviewed for harmless error.
The district court denied Rosenerantz habeas relief but granted a certificate of appealability on the question whether La-sky’s false denial of the pretrial meeting with the prosecution counted as material. This court expanded the certificate of appealability to include the other alleged due process violations discussed below. Rosencrantz timely appealed.
II.
A conviction obtained by the knowing use of perjured testimony must be set aside if “the false testimony could ... in any reasonable likelihood have affected the judgment of the jury....” Giglio v. United States, 405 U.S. 150, 154, 92 S.Ct. 763, 31 L.Ed.2d 104 (1972) (internal quotation marks omitted); see also United States v. Agurs, 427 U.S. 97, 103, 96 S.Ct. 2392, 49 L.Ed.2d 342 (1976). Rosenerantz relies on Giglio’s rule that a prosecutor may not deliberately deceive “a court and jurors by [presenting] known false evidence,” 405 U.S. at 153, 92 S.Ct. 763, and on the command in Brady v. Maryland, 373 U.S. 83, 83 S.Ct. 1194, 10 L.Ed.2d 215 (1963), that a prosecutor must disclose evidence favorable to the accused. By relying on both Brady and Giglio, Rosenerantz implicitly asserts a specific type of Brady violation: one where the prosecutor failed to correct false testimony that he knew, or should have known, to be false (a “knowing-presentation-of-false-testimony claim”).
The difference between Brady/Giglio false-testimony claims and traditional Brady withholding claims drives the analysis here. See Agurs, 427 U.S. at 104, 96 S.Ct. 2392. To prove that the prosecutor’s failure to correct false testimony violated due process rights, a petitioner must demonstrate that: (1) the statement was actually false; (2) the statement was mate*584rial; and (3) the prosecution knew it was false. Coe v. Bell, 161 F.3d 320, 343 (6th Cir.1998); Abdus-Samad v. Bell, 420 F.3d 614, 625-26 (6th Cir.2005); Carter v. Bell, 218 F.3d 581, 601 (6th Cir.2000) (placing the burden on the habeas petitioner). But in these Brady/Giglio claims, the materiality assessment is less stringent than that for more general Brady withholding of evidence claims. We weigh the materiality of Brady withholding claims by asking whether “there is a reasonable probability that, had the evidence been disclosed to the defense, the result of the proceeding would have been different.” Kyles v. Whitley, 514 U.S. 419, 433, 115 S.Ct. 1555, 131 L.Ed.2d 490 (1995) (quoting United States v. Bagley, 473 U.S. 667, 682, 105 S.Ct. 3375, 87 L.Ed.2d 481 (1985)). By contrast, for Brady/Giglio claims, we ask only “if there is any reasonable likelihood that the false testimony could have affected the judgment of the jury.” Agurs, 427 U.S. at 104, 96 S.Ct. 2392 (citing Giglio, 405 U.S. at 154, 92 S.Ct. 763); see also Carter, 443 F.3d at 535. The distinction matters here, because while a traditional Brady materiality analysis obviates a later harmless-error review1 under Brecht v. Abrahamson,2 courts may excuse Brady/Giglio violations involving known and materially false statements as harmless error. See Carter, 443 F.3d at 537; Gilday, 59 F.3d at 268.
In reaching the merits of the case, we note that the deferential standard supplied by AEDPA does not apply here because no state court addressed the merits of Rosencrantz’s knowing-presentation-of-false-testimony claims.3 Williams v. Haviland, 467 F.3d 527, 530 (6th Cir.2006). We thus apply de novo review to the four instances of alleged false testimony — (1) that the assault occurred shortly after 10:30 P.M., on August 16, 1995, rather than shortly after 1:30 A.M. on August 17, 1995, (2) that Lasky felt certain in picking Rosencrantz as the assailant, (3) that she was sober at the time of the assault, and (4) that Lasky had not met with the prosecution before trial.
As for the first three instances of alleged false testimony, we dispose of the them quickly. They fail under the Coe test, either because Rosencrantz fails to demonstrate that Lasky testified in an indisputably false manner, or because he fails to establish that the prosecution knew the statements were false, or both. Al*585though the final instance of false testimony — Lasky’s statements concerning her pretrial interaction with the prosecution— requires closer treatment, we conclude that even if Rosencrantz satisfies the Coe test, harmless-error review excuses the violation.
A. Lasky’s Time Line
According to Gooch, Lasky told him that the assault occurred at 1:45 A.M. on August 17. Lasky testified similarly at the preliminary examination, saying that the assault occurred at 1:30 A.M. But at trial she placed the assault closer to 10:45 P.M. the previous night. Rosencrantz speculates that the prosecutor coaxed this change to weaken his alibi defense. At the habeas evidentiary hearing, however, Rosencrantz failed to question Lasky about the substance of her pretrial discussion with the prosecutor, and the district court correctly declined to make any inferences favoring Rosencrantz on that subject.4 The district court then reasoned that Rosencrantz failed to prove “the time-line testimony provided at trial was false.” We agree. While Lasky vacillated in her reporting of the time, that does not establish “indisputably false” testimony at trial in the presence of prosecutors. Coe, 161 F.3d at 343 (“mere inconsistencies” do not show indisputable falsity).5
Moreover, cross-examination on Lasky’s time line highlighted the inconsistencies between her trial testimony, her statements to Officer Gooch at the time of the assault, and her testimony at the preliminary hearing. The defense focused the *586jury on these inconsistencies, and it was up to the jury — not a federal court conducting collateral review — to sort this out. See Hayes v. Brown, 399 F.3d 972, 992 (9th Cir.2005) (en banc) (Talhnan, J. dissenting) (“An appellate judge’s place is not in the jury box, post hoc.”); United States v. Bortnovsky, 879 F.2d 30, 33 (2d Cir.1989) (“Presentation of a witness who recants or contradicts his prior testimony is not to be confused with ... perjury. It was for the jury to decide whether or not to credit the witness.”). And wherever the truth lies— we still do not know because even at the evidentiary hearing Lasky said she could not “pin down the exact time of day that [she was] assaulted” — nothing about the time-line issue was unknown to the defense at the time of trial. Though Rosencrantz hypothesizes that the prosecution procured this time change in Lasky’s testimony during the pretrial meeting, Rosencrantz’s inferences do not carry the burden of proving indisputable falsity.
B. Lasky’s Certainty in Identifying Rosencrantz as the Assailant
The indisputably-false requirement dooms another of Rosencrantz’s claims-that the prosecution allowed Lasky to identify Rosencrantz with a level of certainty that she did not actually possess. According to Rosencrantz, the prosecution knew Lasky harbored doubt about whether he assaulted her. As support, Rosencrantz relies on Lasky’s testimony at the evidentiary hearing, where she testified that at the time of the photo line-up she “was hung over, and still, and shaking and probably still half drunk” and that she told police she was “in no condition” to make an identification. Lasky also claimed at the evidentiary hearing that she was never certain in her identification of Rosencrantz and that she said as much to the police. But at trial, Lasky said that she “immediately” picked Rosencrantz’s face from the photo line-up. Lasky also confirmed that the tattoo on Rosencrantz’s chest matched her assailant’s tattoo. And when the prosecutor pressed Lasky with “[t]here’s no doubt in your mind that [Rosencrantz] is the man that [assaulted] you?,” she answered “yes.” As the district court observed, this situation fails to establish that the prosecutor presented false testimony because “[o]nce the prosecutor elicited testimony from Lasky identifying [Rosencrantz] as her assailant, it fell to defense counsel to explore the certainty of her identification,” which defense counsel did on cross examination. Indisputable falsehood is not established by a simple swearing contest. The district court was not convinced that Lasky’s testimony at the evidentiary hearing established indisputably her earlier falsehood, let alone the prosecutor’s knowledge of it, and the court did not commit clear error in so deciding.
Without a showing of the testimony’s falsity and that the prosecutor knew it to be false — by Lasky saying that she told the prosecution of her doubt at the time— Rosencrantz fails to show a due process violation in presenting the identification testimony. Though the evidentiary hearing suggested that Lasky either was not as certain at trial as she let on, or that she came to second guess her identification, that does not establish her trial testimony as indisputably false, particularly when Lasky consistently identified her assailant as having a car tattoo on his chest and Rosencrantz bears just such a tattoo.
C. Lasky’s Sobriety
Finally, Rosencrantz complains that the prosecutor allowed Lasky to falsely testify that she was sober at the time of the assault. To show that this testimony qualifies as indisputably false, Rosencrantz points to Pascoe’s post-trial affidavit where he claims that, before the trial, he tipped *587prosecutors to Lasky’s pre-assault insobriety. Rosencrantz also stresses that at the evidentiary hearing, Lasky herself said she was drunk at the time of the assault. Nevertheless, both Lasky and Pascoe stated at trial that she was not under the influence of alcohol at the time of the assault. These contradictory accounts amount only to “mere inconsistencies,” and therefore Rosencrantz’s claim fails to satisfy the indisputably-false requirement. See Coe, 161 F.3d at 343. Nor does Rosencrantz show that, if Lasky did lie, the prosecution knew it. At the evidentiary hearing, defense counsel asked Lasky whether she told “the police and/or the prosecutor in the case that you were drunk, intoxicated, at the time of the assault?” Lasky responded: “I’m not positive of that.”
By failing to satisfy the Coe test, Rosencrantz does not state a due process violation for any of the first three instances of allegedly false testimony.
D. Lasky’s Pretrial Meeting with the Prosecution
We approach this fourth alleged instance of false testimony differently: the district court found that the prosecutor did indeed knowingly allow Lasky’s false denial of the pretrial meeting and we see no clear error in that factual finding. It then assessed whether the prosecutor’s countenancing Lasky’s lying about the fact of a meeting was constitutionally material, and decided to excuse the lie as immaterial because its impeachment value would have been cumulative and because it would not have undermined Lasky’s consistent testimony that her assailant, like Rosencrantz, had the image of a car tattooed on his chest. We review the district court’s materiality analysis de novo as a legal conclusion. See Carter, 443 F.3d 517 at 535.
1. Materiality
In Brady/Giglio cases, false testimony qualifies as material “if there is any reasonable likelihood that the false testimony could have affected the judgment of the jury.’ ” Agurs, 427 U.S. at 104, 96 S.Ct. 2392 (citing Giglio, 405 U.S. at 154, 92 S.Ct. 763). In other words, we will only excuse the perjury as immaterial if we can “say that no reasonable jury could have been affected by the undisclosed information.” Gilday, 59 F.3d at 269. The First Circuit, in contrasting Brady/Giglio (false testimony) materiality with Brady/Bagley (withholding) materiality, observed that the Brady/Giglio standard is “lower,” “more favorable to the defendant,” and “hostile to the prosecution.” Id. at 267-68; see also Ventura, 419 F.3d at 1284 (Brady/Bagley materiality standard “is appreciably more stringent than” the Brady/Giglio standard). And under Giglio’s friendly-to-the-accused standard, “in most cases involving perjury or its equivalent [the result will likely be] a finding of constitutional error.” Gilday, 59 F.3d at 268.
Applying this standard here, we first stress that our review turns on the uncorrected false evidence that no meeting took place — not the other three claimed falsehoods, and not on the four claimed falsehoods viewed cumulatively. Although Rosencrantz wants us to infer from the evidentiary-hearing testimony that the meeting itself precipitated the other instances of allegedly false testimony we lack a valid premise for such an inference. First, the evidentiary-hearing testimony failed to convince the district court (and fails to convince us) of the indisputable falsity of the other three claimed instances. Second, we cannot fill the evidentiary gap here by speculating that the prosecution used the pretrial meetings to coach or coerce Lasky to testify falsely' — de*588fense counsel failed to pursue this topic at the evidentiary hearing, and thus no evidence exists on the substance of the meeting. Rosencrantz’s theory relies on Burgess’s impressions of the tenor of the meetings as intimidating (though she could not hear the conversations) and her claimed conversations with Lasky. But the district court did not find any of that Burgess evidence to be credible. What we do have from the district court regarding the evidentiary hearing is its finding that Lasky’s and Burgess’s testimony (that Lasky met with the prosecution team prior to trial) was credible “in this regard.” Focusing, then, only on La-sky’s denial of the pretrial meeting, we analyze whether any reasonable jury could have been affected by knowing about that false testimony.
Given the implausibility of the untruthful answer Lasky gave — jurors would expect the prosecuting witness to meet with the prosecution before trial — we might assess the impact as minimal. But turning to the impact on the jury had the prosecutor corrected Lasky, or the defense counsel confronted Lasky with her false denial, it is reasonable to infer that exposing La-sky as untruthful — thereby tipping the jury to another of Lasky’s inconsistencies and her willingness to lie under oath— would have affected the jury’s view of La-sky’s credibility. The State argues that, viewed in the context of the whole trial and with the advantage of hindsight, the likely impact of suppressing the fact of the meeting would be minor because it would have been cumulative of her many inconsistencies. Brown v. Smith, 551 F.3d 424, 433-34 (6th Cir.2008) (“Evidence is cumulative when it supports a fact already established by existing evidence; adds very little to the probative force of the other evidence in the case; is merely a repetition of previous testimony; or — in the case of undisclosed impeachment evidence — when the witness has already been sufficiently impeached at trial.”) (internal citations omitted); Mastracchio v. Vose, 274 F.3d 590, 604 (1st Cir.2001) (concluding that, because perjury came from an already “sullied witness,” it “was unlikely to have tipped the balance and changed the minds of those who credited his testimony”). That argument may succeed in many cases. But here, we will assume that Lasky’s lie about the pretrial meeting is material because, as argued by Rosencrantz, Lasky served as the prosecution’s star witness, essentially providing the only evidence against Rosencrantz (though the pool attendant added to the identification evidence). Compare Shih Wei Su v. Filion, 335 F.3d 119, 129 (2d Cir.2003) (viewing false testimony as material where lying witness was “prosecution’s chief witness” and the “conviction depended significantly on [lying witness’s] testimony”), with Foley v. Parker, 488 F.3d 377, 392 (6th Cir.2007) (denying habeas relief where, among other reasons, perjured testimony was immaterial because lying witness “was not a crucial link in the case against [the Defendant].”); see also Napue v. Illinois, 360 U.S. 264, 269, 79 S.Ct. 1173, 3 L.Ed.2d 1217 (1959) (“The jury’s estimate of the truthfulness and reliability of a given witness may well be determinative of guilt or innocence.... ”).
2. Harmless Error
Having assumed materiality and therefore assumed a constitutional error, we now consider the harmlessness of that error. For the reasons given, we find the error harmless because the uncorrected testimony would likely fail to “substantially influence” the jury’s verdict given Lasky’s uncontradicted testimony about the assailant’s tattoo and given the many ways that defense counsel impeached Lasky. Brecht, 507 U.S. at 637, 113 S.Ct. 1710.
*589Our decision to review for harmlessness recognizes “that most constitutional errors can be harmless,” Arizona v. Fulminante, 499 U.S. 279, 306, 111 S.Ct. 1246, 113 L.Ed.2d 302 (1991), and follows the First Circuit’s Gilday analysis in holding that courts should review Brady/Giglio claims for harmless error but not Brady/Bagley habeas claims, given the different materiality standards for establishing the two types of constitutional breaches. Gilday, 59 F.3d at 267-68.
Moreover, the error here is a trial error, not a structural one. Certainly, the prosecutor’s behavior in this case constitutes misconduct that we condemn. But “[despite the fundamental nature of the injury to the justice system caused by the knowing use of perjured testimony by the state, the Supreme Court has not deemed [Brady/Giglio] errors to be ‘structural’ in the sense that they ‘affect[] the framework within which the trial proceeds.’ ” Shih Wei Su, 335 F.3d at 126 (citing United States v. Feliciano, 223 F.3d 102, 111 (2d Cir.2000) (quoting Fulminante, 499 U.S. at 307-10, 111 S.Ct. 1246) (brackets in original)). True enough, harmless-error review under Brecht did not “foreclose the possibility that in an unusual case, a deliberate and especially egregious error of the trial type, or one that is combined with a pattern of prosecutorial misconduct, might so infect the integrity of the proceeding as to warrant the grant of habeas relief, even if it did not substantially influence the jury’s verdict,” 507 U.S. at 638 n. 9, 113 S.Ct. 1710, but we do not view this case as the unusual, especially egregious instance of prosecutorial misconduct, or one that reveals any “pattern of prosecutorial misconduct.”
Governing Supreme Court case law finds “structural errors only in a very limited class of cases,” including: total deprivation of the right to counsel; judicial bias; the unlawful exclusion of grand jurors of defendant’s race; denial of the right to self-representation at trial; the denial of the right to a public trial; and erroneous reasonable-doubt instruction to jury. Johnson v. United States, 520 U.S. 461, 468-69, 117 S.Ct. 1544, 137 L.Ed.2d 718 (1997) (collecting cases). The error in this case, in contrast, is a trial error — it “occurred during the presentation of the case to the jury, and [is one] which may ... be quantitatively assessed in the context of other evidence presented in order to determine whether its admission was harmless.... ” Fulminante, 499 U.S. at 307-08, 111 S.Ct. 1246; see also Hayes, 399 F.3d at 984 (commenting that knowing-presentation-of-false-testimony error is not structural). Although the dissent makes a valid point in arguing that a literal reading of Giglio’s rule suggests that countenancing false testimony implicates structural concerns, the Supreme Court has yet to explicitly hold Brady/Giglio errors as structural. The dissent relies on a footnote in Kyles, but read in its entirety the footnote confirms that Kyles does not settle Brecht’s impact on Brady/Giglio error. 514 U.S. at 433 n. 7, 115 S.Ct. 1555 (“[W]e do not consider the question whether Kyles’s conviction was obtained by the knowing use of perjured testimony and our decision today does not address any claim under the first Agurs category.”).
Reviewing for harmless error under Brecht-instead of reviewing only for materiality and instead of reviewing under Chapman’s harmless-beyond-a-reasonable-doubt standard — is especially appropriate on collateral review, and doing so here honors Brecht’s weighty concerns:
Overturning final and presumptively correct convictions on collateral review because the State cannot prove that an error is harmless under Chapman undermines the States’ interest in finality *590and infringes upon their sovereignty over criminal matters. Moreover, granting habeas relief merely because there is a ‘reasonable possibility’ that trial error contributed to the verdict, see Chapman v. California, 386 U.S. at 24, 87 S.Ct. 824, is at odds with the historic meaning of habeas corpus-to afford relief to those whom society has ‘grievously wronged.’ Retrying defendants whose convictions are set aside also imposes significant ‘social costs,’ including the expenditure of additional time and resources for all the parties involved, the ‘erosion of memory’ and ‘dispersion of witnesses’ that accompany the passage of time and make obtaining convictions on retrial more difficult, and the frustration of ‘society’s interest in the prompt administration of justice.’
507 U.S. at 637, 113 S.Ct. 1710 (internal citations and quotation marks omitted). Testing Rosencrantz’s claim only for Brady/Giglio materiality would be tantamount to applying only Chapman’s harmless error (the standards are equivalent) and would therefore brush aside Brecht’s policy concerns.
Our dissenting colleague argues that harmless-error review is inappropriate here. Yet at oral argument, counsel for Rosencrantz, in ably arguing for reversal and in responding to Chief Judge Boggs’s question, agreed that this court should review for harmless error under Brecht. Rosencrantz’s brief likewise concedes the point. Appellant’s Br. at 29 (“[I]f the court finds the testimony material, it must then determine whether the error was harmless”). The Ninth Circuit adopted the opposite view in Hayes v. Brown, but we think it erred in failing to distinguish false-testimony claims from Brady withholding claims. See Hayes, 399 F.3d at 984-85.
Under Brecht, a knowing-presentation error harms the accused when the error had a “substantial and injurious effect or influence in determining the jury’s verdict.” Brecht, 507 U.S. at 637, 113 S.Ct. 1710. The state bears responsibility for showing that the error had no effect on the verdict. Gilday, 59 F.3d at 268 n. 11 (citing O’Neal v. McAninch, 513 U.S. 432, 436, 115 S.Ct. 992, 130 L.Ed.2d 947 (1995)). If “the matter is so evenly balanced that [the judge] feels himself in virtual equipoise as to the harmlessness of the error,” then courts may not excuse the error as harmless. O’Neal, 513 U.S. at 435, 115 S.Ct. 992.
The district court explained:
The substance of the case against [Rosencrantz] would not have changed had Lasky’s false testimony been corrected. Ascribing a motive to Lasky’s change in testimony or showing that this portion of her testimony was untruthful would not have made a measurable difference in the defense’s ability to challenge her credibility. Lasky’s testimony that her assailant had a car tattoo on his chest remained constant from her initial statement to Officer Gooch through her trial testimony. In light of the cumulative nature of the impeachment evidence and the car tattoo testimony, the court is not in grave doubt about whether the false testimony had substantial and injurious effect or influence in determining the jury’s verdict. Therefore, the court finds any error harmless.
Other comments from the district court bolster this conclusion:
[T]he jury had a compelling piece of incriminating evidence to consider: before there had arisen any time-line disputes or discrepancies about talking to the assistant prosecutor before trial, La-sky told the responding police officer that the perpetrator, whoever he was, *591possessed something immutable, unusual and memorable: the tattoo of a car on his chest. The parties do not dispute that, as of the summer of 1995, [Rosencrantz] bore just such a tattoo on his chest.
We agree with the district court and harbor no “grave doubt” that Lasky’s denial of the pretrial meeting had a “substantial and injurious effect or influence in determining the jury’s verdict.’ ” Id. at 436, 115 S.Ct. 992 (quoting Brecht, 507 U.S. at 637, 113 S.Ct. 1710). Instead, even if the prosecutor had contradicted Lasky’s denial of the meeting, or defense counsel confronted Lasky with her lie, that would not have substantially influenced the jury’s deliberation given the following: (1) the independent and compelling evidence that Lasky identified her assailant as having the tattoo on his chest — an identification she consistently made from the time she reported the assault to the police; (2) the absence of any evidence suggesting that Lasky could have been persuaded or coerced into offering the tattoo testimony; (3) the absence of evidence that the police knew Rosencrantz had a tattoo before putting his photo in the line up; (4) the many ways that defense counsel impeached La-sky and impugned her character; (5) the modest value to have been reasonably gained from the jury learning that Lasky lied about meeting with the prosecutor; (6) the absence of evidence of coaching and coercing despite the opportunity to explore the interaction between Lasky and the prosecution at the evidentiary hearing; and (7) a Michigan jury found Rosencrantz guilty after a trial where Lasky’s credibility and character had been thoroughly impugned by the defense, even without knowing she lied about the meeting.6
Our decision to excuse the error here as harmless simply analyzes the effect on the verdict of the prosecutorial lapse — it in no way countenances it. The “deliberate deception of a court and jurors by the presentation of known false evidence is incompatible with rudimentary demands of justice.... [T]he same result obtains when the State, although not soliciting false evidence, allows it to go uncorrected when it appears.” Giglio, 405 U.S. at 153, 92 S.Ct. 763 (internal citations and quotation marks omitted); Agurs, 427 U.S. at 104, 96 S.Ct. 2392 (condoned perjury “involve[s] a corruption of the truth-seeking function of the trial process”); Jones v. Kentucky, 97 F.2d 335, 338 (6th Cir.1938) (“[Fundamental conceptions of justice which lie at the base of our civil and political institutions must ... condemn as a travesty a conviction upon perjured testimony if later, but fortunately not too late, its falseness is discovered.... ”). Prosecutorial culpability cannot, however, outweigh Brechts policy concerns. See Smith v. Phillips, 455 U.S. 209, 219 n. 10, 102 S.Ct. 940, 71 L.Ed.2d 78 (1982) (in analyzing materiality, noting that the prosecutorial “misconduct’s effect on the trial, not the blameworthiness of the prosecutor, is the crucial inquiry for due process purposes.”). Rosencrantz’s case, although it *592presents a close call, falls short. We excuse the error here as harmless.
III.
Viewing Rosencrantz’s claim through the habeas lens, we are convinced that any constitutional error fails the Brecht standard. We affirm the judgment of the district court dismissing Rosencrantz’s habeas petition.
. The materiality standard in traditional Brady claims supplants harmless-error review because practically speaking, the two analyses are the same. See Kyles, 514 U.S. at 435, 436, 115 S.Ct. 1555 ("[OJnce there has been Bagley error ... it cannot subsequently be found harmless under Brecht" because " 'a reasonable probability that, had the evidence been disclosed to the defense, the result of the proceeding would have been different’ necessarily entails the conclusion that the suppression must have had 'a substantial and injurious effect or influence in determining the jury's verdict.’ ”) (internal citations omitted).
. We do not apply the harmless-error standard supplied by Chapman v. California, 386 U.S. 18, 87 S.Ct. 824, 17 L.Ed.2d 705 (1967), because — as Justice Blackmun tells us in a portion of his Bagley opinion not joined by the majority — the Chapman harmless error standard "is equivalent to" the materiality standard already applicable to Brady/Giglio claims. 473 U.S. at 679 n. 9, 105 S.Ct. 3375 (1985); see also Ventura v. Attorney General, 419 F.3d 1269, 1279 n. 4 (11th Cir.2005).
. An issue exists as to whether AEDPA deference applies to Rosencrantz’s time-line claim because the Michigan Court of Appeals addressed that point, but it did not have the benefit of Lasky's later testimony at the evidentiary hearing. The parties ignore this issue, and we need not decide because Rosencrantz's claim fails even under de novo review. Smith v. Jones, No. 07-2275, 2009 WL 973349, at *5 (6th Cir. April 10, 2009) (citing Baze v. Parker, 371 F.3d 310, 320 (6th Cir.2004)).
. Aside from Burgess’s affidavit, all that we know about the pretrial discussions between Lasky and the prosecution comes from these questions, asked by counsel for the state at the evidentiary hearing:
Ms. Gagliardi: You indicated in your ... responses to the questioning that you were incarcerated prior to the preliminary hearing?
Elaine Lasky: Yes, ma'am.
Ms. Gagliardi: At Genesee County Jail. Could you tell the Court why you were incarcerated at that time?
Elaine Lasky: I was charged with a shoplifting charge in Mount Pleasant and was under their jurisdiction....
Ms. Gagliardi: Did the police officers or the prosecutors ever discuss with you the shoplifting charge that you were incarcerated for at that time? Did you ever meet with them on those charges?
Elaine Lasky: No. Because I was already sentenced on that. I was sentenced to one year at the county, the Mount Pleasant Jail. And Genesee County just held me in their jail.
Ms. Gagliardi: Were there any discussions with the prosecutors or the investigators regarding any other case beside the one where you were the complainant against Mr. Rosencrantz?
Elaine Lasky: No.
. Rosencrantz’s argument suggests that La-sky’s testimony at the evidentiary hearing represents the truth because it occurred on a later date than her trial and preliminary-examination testimony. But our review reveals reason to doubt the credibility of Lasky’s evidentiaiy-hearing testimony because that testimony — made a decade after Rosencrantz’s trial — evinces signs of a failing memory. For example, the trial transcript shows Lasky insisting that she was sober at the time of the assault. But at the evidentiary hearing, Lasky could not recall having testified in that way. Also during the evidentiary hearing, Lasky had trouble distinguishing between the preliminary examination and the trial. Lasky voiced her confusion, saying: "when you’re saying preliminary, I’m trying to get it in my head which one was the preliminary hearing, which one was which. I can’t say positively." (Admittedly, later in the evidentiary hearing, Lasky seemed to appreciate the timing of these hearings.) Finally, even at the evidentiary hearing, Lasky still could not “pin down” the assault’s exact time. Rosencrantz misapprehends not only our indisputably-false requirement, but also the nature of our appellate review. The responsibility for factual accuracy primarily rests with trial judges and juries, and state appellate courts. See Kyles, 514 U.S. at 458, 115 S.Ct. 1555 (Scalia, J. dissenting).
. And as stated above in our materiality discussion, Rosencrantz also urges this court to consider how defense counsel might have used the admission of a pretrial meeting to further impeach Lasky on her changed time line. But at the evidentiary hearing, Rosencrantz's counsel did not question Lasky about what the prosecutor said to her during the pretrial meeting. Rosencrantz's counsel instead relied solely on what Burgess observed from outside the room. And that evidence failed to impress the district court; it credited Burgess only with establishing that a meeting took place. Our harmless-error review cannot include credibility findings beyond those of the district court. See Hayes, 399 F.3d at 992 (en banc) (Tallman, J. dissenting) ("An appellate judge's place is not in the jury box, post hoc.").