dissenting.
Timothy Rosencrantz was convicted of sexual assault and sentenced to twenty-two-and-one-half to fifty years in prison based on the testimony of the victim who materially perjured herself at trial regarding her collaboration with the prosecution, her sobriety at the time of the attack, and her certainty as to the identity of the assailant. The prosecution knew the victim’s testimony was false at the time it was given, but did nothing to correct it. The prosecution’s inaction violated Rosencrantz’s due process rights under the Fourteenth Amendment, and Rosencrantz is entitled to postconviction relief under Brady v. Maryland, 373 U.S. 83, 86-87, 83 S.Ct. 1194, 10 L.Ed.2d 215 (1963). The majority, however, refuses to grant Rosencrantz such relief, and alters the straightforward rule of Brady/Giglio by holding that the prosecutorial misconduct constituted “harmless error” under Brecht v. Abrahamson, 507 U.S. 619, 629-30, 113 S.Ct. 1710, 123 L.Ed.2d 353 (1993). See Brady, 373 U.S. at 86-87, 83 S.Ct. 1194; Giglio v. United States, 405 U.S. 150, 153-154, 92 S.Ct. 763, 31 L.Ed.2d 104 (1972). The majority’s position misconstrues both Brady and Brecht and fails to appreciate that a Brady violation necessarily brings into question the substantive fairness of the trial, and, therefore, material Brady violations are never “harmless” under Brecht. Moreover, even if Brecht’s harmless-error analysis applied, the prosecution’s inaction was not harmless in this case. In short, because the majority lacks any sound reason for denying postconviction relief, I respectfully dissent.
I.
Brady is rooted in a line of cases requiring postconviction relief for “material” prosecutorial misconduct. See Brady, 373 U.S. at 86-87, 83 S.Ct. 1194; Giglio, 405 U.S. at 153-154, 92 S.Ct. 763.1 These cases demonstrate that certain prosecutorial misconduct can so undermine confidence in a verdict and impact the fairness of trial that a new trial is required. Giglio, for example, plainly states: “A new trial is required if ‘the false testimony could ... in any reasonable likelihood have affected the judgment of the jury.’ ” 405 U.S. at 154, 92 S.Ct. 763 (quoting Napue v. Illinois, 360 U.S. 264, 271, 79 S.Ct. 1173, 3 L.Ed.2d 1217 (1959)); see also Kyles v. Whitley, 514 U.S. 419, 433, 115 S.Ct. 1555, 131 L.Ed.2d 490 (1995). This Court has recognized that same rule, stating that a new trial follows automatically from the “knowing use of perjured testimony” given “any reasonable likelihood that the false testimony could have affected the judgment of the jury.” Byrd v. Collins, 209 F.3d 486, 517 (6th Cir.2000) (citations omitted). The majority all but ignores this line of precedent and its per se requirement *593for habeas relief in cases of “material” prosecutorial misconduct.
In so doing, the majority treats Brady and its associated line of cases as providing merely a threshold test for Brecht’s harmless-error analysis. (Majority Opinion (“Maj.Op.”) at 583-85.); Brecht, 507 U.S. at 629-30, 113 S.Ct. 1710. But the majority’s application of Brecht is misplaced. Brecht sets forth a two-prong inquiry for the grant of habeas relief for trial errors: first, the habeas petitioner must establish the existence of a constitutional “trial error,” and second, the court must ascertain that the error was not harmless. Id. The majority reads Brady as simply identifying the existence of a particular constitutional “trial error”; that is, the majority reads Brady as being relevant only to the first prong of Brecht. (Maj. Op. at 583-85.) But this interpretation effectively breaks the link between the existence of a Brady/Giglio violation and the grant of postconviction relief. The whole purpose of the Giglio-materiality test is to identify those due process harms requiring post-conviction relief. 405 U.S. at 153-54, 92 S.Ct. 763. If Giglio is still good law today, then the idea that a petitioner’s claims could satisfy the Giglio-materiality test — as the majority concedes Rosencrantz has done — but not be entitled to a new trial is insupportable. (Maj. Op. at 589-92.); see Giglio, 405 U.S. at 153-54, 92 S.Ct. 763.
The majority’s error may lie in its failure to appreciate that Brecht’s use of “trial error” as a term of art. Not every trial defect constitutes a Brecht “trial error,” and certain defects necessarily remain wholly outside Brecht’s purview. For example, “trial error” as used by Brecht excludes “structural” trial defects, and harmless-error analysis does not apply to such defects. 507 U.S. at 629-30, 113 S.Ct. 1710. “Structural” defects are defects that, by their very nature, undermine confidence in the jury’s verdict or otherwise call the substantive fairness of the entire trial into question. Kyles v. Whitley provides indirect support for this view. 514 U.S. at 453, 115 S.Ct. 1555. In Kyles, the Supreme Court applied the Bagley-materiality standard to the non-disclosure of exculpatory evidence at trial, stressing that satisfaction of Bagley’s standard necessarily implied satisfaction of Brecht, thereby rendering moot a Brecht analysis. Id. at 436, 115 S.Ct. 1555; United States v. Bagley, 473 U.S. 667, 682, 105 S.Ct. 3375, 87 L.Ed.2d 481 (1985). Thus, although Bagley does not itself apply in the present case, the Court’s discussion of Bagley in Kyles sheds light on the sort of case in which Brecht’s analysis is not required. The Kyles Court stated:
[T]he question is not whether the State would have had a case to go to the jury if it had disclosed the favorable evidence, but whether we can be confident that the jury’s verdict would have been the same.
514 U.S. at 453, 115 S.Ct. 1555. The implication of this statement is that Brecht’s analysis is moot where we cannot be “confident that the jury’s verdict would have been the same.” Id.
Brady/Giglio harm by its nature goes to the substantive fairness of trial. The genius of Brady-and perhaps the reason that it has lent its name to a whole category of constitutional claims — lies in its recognition that certain material prosecutorial misconduct renders a trial ipso facto substantively unfair. See Brady, 373 U.S. at 87-88, 83 S.Ct. 1194. Where this misconduct is present, post-conviction relief is mandatory. Id. An individual may not be imprisoned when the fairness of his trial is in question. See id. 86-87, 83 S.Ct. 1194 (describing prosecutorial-abuse cases lead*594ing up to Brady). And Brecht does not alter this fundamental rule. See Brecht, 507 U.S. at 627-39, 113 S.Ct. 1710. As the Supreme Court stated in the post-Brecht case of Kyles: “a conviction 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.” 514 U.S. at 433, 115 S.Ct. 1555 (internal quotations omitted).
Given the conceptual framework of Brady and Brecht, we should not shoehorn Brady into Brecht’s harmless-error analysis. Rather, Brady and Brecht remain consistent only so long as they stand apart. The Court’s only task in the present case, then, is to apply the Giglio test. If the test is satisfied, “a new trial is required.” Giglio, 405 U.S. at 154, 92 S.Ct. 763.
II.
The majority does not dispute that the prosecutor’s knowing failure to correct the perjured testimony regarding meetings occurring between the prosecution and the victim meets the Giglio-materiality test. (Maj. Op. at 587-89.) Such material misconduct plainly had a “reasonable likelihood” of “affect[ing] the judgment of the jury.” Giglio, 405 U.S. at 154, 92 S.Ct. 763. Because Giglio requires a “new trial” under these conditions, I would reverse the district court and order a new trial. Id.
III.
Assuming arguendo that Brecht has overruled Giglio-as well as our prior precedent, see Byrd v. Collins, 209 F.3d 486 (6th Cir.2000) (following the Giglio rule) — I still maintain that the prosecution’s misconduct requires a new trial because those errors were not harmless under Brecht.
In Brecht, the Supreme Court resolved a dispute over the proper “harmless error” analysis applicable to certain types of “trial error.” Brecht, 507 U.S. at 637-38. The Brecht Court held that trial errors do not require postconviction relief unless the error had a “substantial and injurious effect or influence in determining the jury’s verdict.” Id. Stated differently, Brecht’s harmless-error analysis requires “actual prejudice” before a “trial error” can result in habeas relief. Id.
The majority considers four allegations of prosecutorial misconduct. Separately, three of these errors satisfy Brecht’s harmless-error analysis. Considered jointly, therefore, there is little doubt that the facts of this case call for a new trial.
First, Elaine Lasky, the victim, testified at trial that she did not meet with prosecutors prior to trial. The prosecutor did nothing to correct Lasky’s false testimony, even though the prosecutor knew it was false. The majority properly concludes that this constitutes a material Brady violation but fails to recognize that this fact satisfies more than the first prong of Brecht. Specifically, the majority rejects the idea that prosecutorial misconduct of this kind had a “substantial and injurious effect or influence [on] the jury’s verdict” under Brecht. (Maj. Op. at 591.) I strongly disagree with this conclusion. The prosecution’s inaction is troubling in several respects. First, Lasky’s testimony eliminates any opportunity for the defense to examine Lasky’s interaction or communication with the police and investigating officers. Second, her testimony forecloses questioning into her dealing with the prosecution, which in this case may have been relevant to the strength of her identification of Rosencrantz. Third, and perhaps most troubling, is the fact that had the prosecution corrected Lasky’s testimony it *595may have raised credibility concerns with the jury. Because of the prosecution’s inaction, the jury was unaware that the key witness repeatedly and unequivocally lied on the stand. This is precisely the sort of information that would affect any careful jury deliberation. See United States v. Serido, 705 F.2d 459, 1982 U.S.App. LEXIS 11672, *5 (6th Cir.1982) (stating that “due process is violated and a new trial is required when a prosecutor intentionally or inadvertently fails to correct materially false testimony relevant to the credibility of a key government witness”) (citing Giglio, 405 U.S. at 150, 92 S.Ct. 763). Under these conditions, where the prosecution knew of the perjury, the process of fair jury deliberation is undermined such as to call into question society’s confidence in the verdict itself. This constitutes “actual prejudice,” and satisfies the second prong of Brecht.
Second, at the post-trial evidentiary hearing before the district court, Rosencrantz presented uncontroverted evidence that the victim was intoxicated at the time of the assault and that the police were aware of this fact. The record is clear that the prosecution allowed Lasky to testify at trial that she was sober at the time of the assault: “I wasn’t intoxicated at all when I entered into that truck.” (JA 386.) At the subsequent evidentiary hearing, La-sky repudiated her prior testimony under direct examination by Rosencrantz’s attorney, Nancy McGunn:
Ms. McGunn: Okay. This assault took place — at trial, you testified, rather, that this assault took place after you left the hotel room and entered a truck, is that correct?
Ms. Lasky: Yes.
Ms. McGunn: Okay. And then at that point in the truck you were sexually assaulted?
Ms. Lasky: Yes.
Ms. McGunn: Okay. When you left the hotel room that night, were you intoxicated?
Ms. Lasky: Yes.
Ms. McGunn: Okay. You were under the influence of alcohol, at a minimum?
Ms. Lasky: Yes.
Ms. McGunn: Okay. Do you recall testifying at trial that you were not intoxicated when you entered the truck?
Ms. Lasky: No, I don’t. I do not recall that.
Ms. McGunn: Okay. If you had testified in that manner, would that have been incorrect?
Ms. Lasky: Yes.
(JA 1837-38.) Moreover, Jack Pascoe— Lasky’s companion at the hotel the night of the incident — submitted an affidavit for the post-trial evidentiary hearing corroborating Lasky’s testimony that she was intoxicated at the time of the attack: “At the point at which [Lasky] left the room [shortly before the assault], she was well on her way to being intoxicated.” (Affidavit of Jack Pascoe, Oct. 30, 2006 (“Pascoe Aff.”) ¶ 1.) Pascoe’s affidavit states that he and Lasky were both drinking and using “a lot of powder cocaine,” and “[he] remember[s] talking to the prosecutor and some police officers about the night ... [and he is] confident he told them that [Lasky] had been drinking and doing drugs before she left the hotel room to get some ice, because she definitely had been.” (Pascoe Aff. ¶¶ 2, 4.)
While we have previously stated in Williams v. Coyle, 260 F.3d 684, 708 (6th Cir.2001) that “recanting affidavits are always viewed with extreme suspicion,” the reasoning of our decision provides guidance regarding the acceptance or rejection of such affidavits. In cases such as Williams, where the prosecutor provides testimony contrary to an affidavit at the *596evidentiary hearing, we require a greater showing to support the recanting affidavit. Id. However, in those cases where the prosecutor does not testify at the evidentiary hearing or offer any contradictory evidence arguing for the rejection of the affidavit, the sworn testimony of the affiant may not be summarily dismissed. In the present case, there was no counter-evidence offered by the prosecution. Moreover, Pascoe’s affidavit corroborated, and was corroborated by, Lasky’s testimony. The district court’s summary rejection of Jack Pascoe’s affidavit was thus improper, and should be rejected as clearly erroneous. (JA 1810); see Girts v. Yanai, 501 F.3d 743, 752 (6th Cir.2007) (Factual findings are reviewed for “clear error.”).
Once Pascoe’s affidavit is given due weight, it is evident that Lasky lied on the stand about an issue materially relevant to her ability to have reliably identified Rosencrantz as the perpetrator — namely, whether or not she was intoxicated (and how severely) at the time of the assault. Because prosecutors are charged with knowledge of facts known to the police, the prosecution’s failure to correct Lasky’s testimony constitutes a material Brady violation that meets the first prong of Brecht. See Kyles, 514 U.S. at 437-38, 115 S.Ct. 1555 (holding that prosecutors have a duty to learn of evidence “known to others acting on the government’s behalf in the case, including the police,” and that such constructive knowledge is sufficient to implicate Brady). Brecht’s second prong is satisfied because withholding the fact of Lasky’s intoxication at the time of the assault — information relating to Lasky’s ability to identify her assailant — seriously harmed the jury’s deliberative process. In short, this was a case where the jury weighed the credibility and testimony of the accused against that of the accuser. Lasky’s sobriety or intoxication at the time of the assault is precisely the sort of information that would affect the determination of the jury’s verdict. See Brecht, 507 U.S. at 637-38, 113 S.Ct. 1710. Because withholding this information taints jury deliberation and undermines confidence in the verdict, Rosenerantz must be afforded a new trial.
Third, Rosenerantz also offered uncontroverted evidence at the district court’s evidentiary hearing that the victim had reservations about her identification of Rosencrantz as her assailant. The record is also clear that the prosecution nevertheless allowed Lasky to lie at trial about her confidence that she had correctly identified her assailant. On direct examination by the Assistant Prosecuting Attorney, Garner Train, Lasky testified about the photo-lineup identification:
Mr. Train: Okay. Based on your seeing the man that did this to you, were you able to pick out the person in the line-up that did it?
Ms. Lasky: Immediately.
(JA 375-76.) The prosecutor emphasized this point himself, repeating, “[immediately,” and then proceeded to have Lasky identify Rosenerantz as her attacker. Id. At the evidentiary hearing, Lasky dramatically altered her account of the photo-lineup identification:
Ms. McGunn: Okay. Did you select a photo from the lineup that day? Ms. Lasky: Yes, I did.
Ms. McGunn: Okay. In selecting that photo were you absolutely sure that was the person who had assaulted you?
Ms. Lasky: No.
Ms. McGunn: ... Did you tell the police that you were not sure that Mr. Rosencrantz was the person who had assaulted you?
*597Ms. Lasky: Yes.
(JA 1838-40.)
The district court concluded that the victim’s own post-trial testimony was “insufficient to establish that the prosecutor knowingly presented false testimony when he permitted Lasky to testify that she identified Petitioner.” (JA 1811.) By “insufficient,” the district court apparently meant legally insufficient, because the court immediately offered its opinion on the obligation of the defense counsel to attack the photo-lineup-identification evidence at trial. Id. The district court’s reasoning appears to be that Lasky’s recanting photo-lineup-identification testimony may be disregarded because defense counsel, at trial, had a fair chance to “explore the certainty of her identification.” Id. This line is echoed by today’s majority. (Maj. Op. at 586.)
This legal analysis is seriously flawed. When considering Brady violations, the obligations and actions of defense counsel are irrelevant, and the analysis should focus on the actions of the prosecutor in contributing to a substantively unfair trial. See Giglio, 405 U.S. at 153-54, 92 S.Ct. 763 (noting that satisfying Brady “is the responsibility of the prosecutor”). It is prosecutors who have minimum obligations under Brady, not defense counsel. Id. Consequently, Lasky’s photo-lineup testimony at the evidentiary hearing is legally sufficient to constitute a Brady claim.
Thus, Lasky’s uncorrected false testimony regarding her intoxication and her photo-lineup identification account for two more violations of Rosencrantz’s due process rights, either of which requires a new trial. Giglio, 405 U.S. at 153-54, 92 S.Ct. 763; Kyles, 514 U.S. at 433 n. 7, 115 S.Ct. 1555; Byrd, 209 F.3d at 517. This same result is also required under the harmless-error analysis of Brecht. Brecht, 507 U.S. at 637-638, 113 S.Ct. 1710.
Here, Rosencrantz’s conviction depended on the ability of the victim to have confidently identified him as the perpetrator. The fact that Lasky was intoxicated at the time of the assault and the fact that Lasky did not have confidence in her identification of Rosencrantz at the photo lineup would, if known to the jury, have had a “substantial ... effect or influence” on its deliberations. But for the prosecutor’s inaction, in violation of Brady, the jury would have known these facts. As such, we cannot with confidence accept the jury’s verdict or the substantive fairness of the trial itself, and a new trial is warranted.
IV.
For all of the above reasons, I respectfully dissent.
. In chronological order, this line of cases includes: Mooney v. Holohan, 294 U.S. 103, 55 S.Ct. 340, 79 L.Ed. 791 (1935); Pyle v. Kansas, 317 U.S. 213, 63 S.Ct. 177, 87 L.Ed. 214 (1942); Napue v. Illinois, 360 U.S. 264, 79 S.Ct. 1173, 3 L.Ed.2d 1217 (1959); Brady v. Maryland, 373 U.S. 83, 86-87, 83 S.Ct 1194, 10 L.Ed.2d 215 (1963); Giglio v. United States, 405 U.S. 150, 92 S.Ct. 763, 31 L.Ed.2d 104 (1972); Kyles v. Whitley, 514 U.S. 419, 115 S.Ct. 1555, 131 L.Ed.2d 490 (1995).