KELLY, Circuit Judge.
Clifton Gabaree, a Missouri inmate, petitioned the district court1 for a writ of habeas corpus, 28 U.S.C. § 2254. The court concluded that Gabaree’s trial counsel had rendered ineffective assistance and granted Gabaree’s petition in part and denied it in part. The State appeals the partial grant of relief, and Gabaree appeals the partial denial. We agree with the district court’s conclusions and affirm the judgment.2
I. Background
In 1992, Gabaree began living with a woman who had two daughters, B.S. and A.S. Gabaree and the woman later had two children together, M.G. and B.G. In December 1996, B.S., A.S., and M.G. told an employee of the Missouri Department of Social Services that Gabaree was physically abusing them and B.G. The children “had marks, burns, and bruises consistent with the physical abuse they described; they were dirty and covered with untreated scratches and scrapes.” Gabaree v. State, 290 S.W.3d 175, 177 (Mo.Ct.App.2009). Dr. James Kelly, a pediatric physician, examined the children and determined that the marks had been inflicted intentionally.
Six-year-old B.S. also told a social worker that Gabaree had “put his ‘thingy’ and finger inside of her bottom ‘a long time ago,’ when she was three years old.” Gabaree, 290 S.W.3d at 177. B.S. said she saw Gabaree doing the same thing to A.S.; but A.S., who then was five years old, told the social worker that Gabaree had not “given her or her sister a bad touch.” Id. Yet A.S. had told her mother that Gabaree had “played with her privates with Q-tips when she and the other children were playing hide-and-go-seek with him.” Id. The daughters later gave videotaped interviews at the Child Protection Center in Missouri. B.S. again told the interviewer that Gabaree had “put his ‘private’ in A.S.’s ‘back private’ ” and had “put his finger in her front private and made her ‘suck his private.’” Id. A.S. this time said Gabaree had “touched her ‘butt’ and ‘pee pee’ with a belt and his hand.” Id. Both girls were examined for physical evidence of sexual abuse, but no evidence was found.
Gabaree was arrested and charged with three counts • of statutory sodomy, three counts of first-degree child molestation, and six counts of child abuse. A jury found him guilty of all 12 counts, and the Missouri Court of Appeals upheld his convictions. On post-conviction review, however, the trial court granted Gabaree relief on his claim of ineffective assistance of [994]*994trial counsel. The post-conviction court concluded that Gabaree’s trial attorney had failed to impeach the girls by introducing statements they had made recanting their allegations of sexual abuse. The court concluded that counsel’s performance prejudiced Gabaree and ordered the State to retry him.
Gabaree’s second trial took place in 2003. At the trial, B.S. testified that Ga-baree had sex with her when she was between three- and seven years old and that it “ ‘felt like being in hell.’ ” Gabaree, 290 S.W.Bd at 179. B.S. also said she saw Gabaree having sex with her sister, A.S., who testified that Gabaree had “touched her ‘private part’ with his hands.” Id. Gabaree testified in his defense that he did physically discipline the children but explained that he did not do so in an abusive manner. He denied any sexual abuse but testified that he, his ex-wife (the mother of all the children), and the girls had bathed naked together and that during those baths, he may have had an erection that touched the girls. He suggested that the contact made during the baths may have led the girls to believe that they had been molested. The attorney for the State asked Gabaree on cross-examination if he had “leaned back and smiled ... kind of reliving this experience” when describing the possible sexual contact; counsel did not object to that suggestion. Also presented into evidence was the statement of an unavailable witness who had testified at the first trial that B.S. and A.S. had recanted their allegations of sexual abuse.
Two doctors also testified regarding their meetings with Gabaree and his stepdaughters. Dr. Kelly described the marks on the children’s bodies as “angry” and suggested they had been caused by belt buckles and cigarettes. He told the jury that, in his opinion, the wounds had been intentionally inflicted. Dr. Kelly also was permitted to testify about another doctor’s finding that there was no physical evidence of sexual abuse, which he said was not surprising based on the passage of time and the body’s ability to heal.3 Dr. Kelly also testified, without objection, that although he had not interviewed B.S. and A.S. regarding any allegations of sexual abuse and was not present when they were interviewed on that subject, he believed that their disclosure of the details of the sexual abuse was “a very consistent ... very credible disclosure from both of them.” On cross-examination, counsel asked few questions and made no attempt to impeach Dr. Kelly’s assessment of the girls’ credibility.
Dr. Gregory Sisk, a clinical psychologist, testified that he had interviewed Gabaree and administered psychological questionnaires designed to uncover abusive or neglectful attitudes. According to Dr. Sisk, Gabaree’s answers suggested that he was using his children “to meet his needs, to give him comfort or love,” and also possibly to meet his sexual needs. Dr. Sisk opined that Gabaree “was holding onto parenting beliefs that contributed to abuse and neglect.” Gabaree’s trial counsel did not object to Dr. Sisk’s testimony. During cross-examination, in response to counsel’s questions, Dr. Sisk further testified that if a person with children received the same scores on the questionnaires as Gabaree, “they’re probably abusing or neglecting the children if they’re acting on those beliefs.” When asked if a person necessarily acted in accordance with the beliefs documented by the questionnaire, Dr. Sisk responded, “I don’t know how someone would approach the questionnaire and start writing down beliefs that were abusive but, yet, in practice, were not doing those.”
[995]*995The State relied on the testimony of the doctors in its closing argument, reminding the jury that Dr. Kelly believed the girls were “credible, believable, and consistent” when talking about the sexual abuse and that Dr. Sisk had suggested that Gabaree had a “propensity toward violence” and “would use children to satisfy his own needs, including sexual needs.” In her closing argument, Gabaree’s counsel mentioned neither doctor nor their testimony. Gabaree again was convicted of all 12 counts. He was sentenced to life imprisonment on the sodomy convictions, 7 years’ imprisonment for each of the three child-molestation convictions, and a total of 38 years’ imprisonment on the six counts of child abuse, with all terms to be served consecutively. His convictions and sentences again were upheld on direct appeal.
Gabaree then petitioned the state trial court for post-conviction relief, which was denied without a hearing. The Missouri Court of Appeals, however, reversed the trial court’s decision and remanded for an evidentiary hearing. During that hearing, Gabaree’s trial counsel testified that she had no memory whatsoever of the case except Gabaree’s “unusual name” and an interview she had conducted with the two girls. Nor did counsel have any memory of what her trial strategy may have been, though she insisted that she has “never gone to trial without a theory or defense.”
Counsel also was questioned about her strategy regarding Drs. Kelly and Sisk. Although she did not remember any of the testimony, she agreed that a witness, even an expert witness, is not allowed to comment on the truth or veracity of another’s statement. She stated that, as a general matter, she tries to object to testimony “that is really bad,” but she sometimes does not hear it and fails to object. In other circumstances, she added, she intentionally refrains from objecting because “it’s easier just to let it go than to make a big deal and make the jury obvious that this is a problem.” Counsel reviewed the testimony of the doctors from Gabaree’s second trial. She characterized Dr. Kelly’s testimony as “non-responsive” and suggested that she “could also very possibly have just missed it.” Counsel described Dr. Sisk’s opinion that Gabaree’s tests suggested he was abusing his children merely as suggesting that Gabaree “needed parenting classes, in a sort of complicated T am a doctor’ sort of way.” Counsel was less sure that she did not object to avoid drawing attention to Dr. Sisk’s testimony and suggested that she “could have been [sic] easily been thinking I could argue this way. Then, of course, the question is, did I argue it away or did I not get that far?” Counsel acknowledged that, regarding arguing away Dr. Sisk’s conclusions, “it could very easily not have been in the closing, because I didn’t say it because I didn’t think to put it in.”
After the hearing, the trial court again denied Gabaree’s petition. Though it noted that the questionable excerpts of both doctors’ testimony were inadmissible, the court of appeals affirmed the trial court’s decision, concluding that counsel’s stated possible reasons for not objecting to the doctors’ testimony were not necessarily unreasonable. The Supreme Court of Missouri denied Gabaree’s petition to transfer.
Gabaree then petitioned for a writ of habeas corpus under 28 U.S.C. § 2254. He maintained that his trial attorney had been ineffective for failing to object and request a mistrial in response to Dr. Kelly’s and Dr: Sisk’s testimony.4
[996]*996The district court granted Gabaree’s petition in part and denied it in part. The court first determined that the state court of appeals’ conclusion that Gabaree’s counsel had adequately performed was both an unreasonable application of Strickland v. Washington, 466 U.S. 668, 104 S.Ct. 2052, 80 L.Ed.2d 674 (1984), and an unreasonable determination of the facts. 28 U.S.C. § 2254(d)(1), (2). The court first noted that, though counsel had no memory of the case, the strategy that during the eviden-tiary hearing she suggested she had followed was “invalid.” Although in some instances, the court acknowledged, it would be reasonable not to call attention to harmful testimony or improper evidence, in this case there was little other evidence to support the convictions. Given the context, the court rejected any notion that counsel’s purported strategy not to object to Dr. Kelly’s testimony was reasonable or valid. The court next determined that permitting Dr. Sisk’s inadmissible testimony evinced deficient performance. Moreover, the court continued, counsel had not employed her purported strategy of “arguing away” Dr. Sisk’s conclusions: She did not mention Dr. Sisk or his testimony during her closing argument. Allowing in the doctors’ testimony out of fear of highlighting it, the court concluded, “was a patently unreasonable trial strategy.”
The court also concluded that Gabaree was prejudiced by counsél’s failures. Because the state court had not addressed this question at all, the court first determined, there was no decision to which deference had to be given, and review was de novo. The court pointed to the weaknesses in the State’s case, described the doctors’ testimony as improperly “shor[ing] up those weaknesses,” and noted that the State relied on that evidence in its closing argument. Thus, the court concluded, there is a reasonable probability that, but for counsel’s errors, the outcome of the trial would have been different with respect to the sodomy and child-molestation counts. Because physical- and medical evidence supported the child-abuse convictions, however, the court concluded that Gabaree was not prejudiced with respect to those convictions.
II. Discussion
On appeal from a district court’s grant of relief under § 2254, we review de novo the district court’s conclusions of law, including its application of the standards of review under § 2254; factual findings are reviewed for clear error. Escobedo v. Lund, 760 F.3d 863, 868 (8th Cir.2014). Under § 2254, we must give deference to the state court’s decision unless it is “contrary to, or involved an unreasonable application of, clearly established Federal law” or rests 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 presume the factual findings of the state court are correct unless rebutted by “clear and convincing evidence.” 28 U.S.C. § 2254(e)(1).
We review Gabaree’s claims of ineffective assistance under the two-part test from Strickland, which requires him to show that counsel’s performance fell below professional norms and that, as a result of that deficient performance, he was prejudiced. Forrest v. Steele, 764 F.3d 848, 853 (8th Cir.2014). To grant relief under § 2254, therefore, we must conclude that the state court unreasonably applied the Strickland test or that, in reaching its conclusion regarding the performance of Gabaree’s attorney, it made unreasonable factual conclusions. See Escobedo, 760 F.3d at 869-70 (quoting Harrington v. Richter, 562 U.S. 86, 131 S.Ct. 770, 785, 178 L.Ed.2d 624 (2011)).
A. Sex-abuse Convictions
First, the State argues that the district court improperly granted Gabaree relief [997]*997on his claim of ineffective assistance of counsel with respect to his sex-abuse convictions. The State asserts that the state court reasonably concluded that counsel’s purported reasons not to object to the testimony of Drs. Kelly and Sisk were part of a reasonable trial strategy. Thus, the State concludes, the district court improperly rejected the state court’s conclusions and failed to defer to the state court’s decision under § 2254. We separately will discuss each Strickland, factor.
i. Counsel’s performance
Under the first part of the Strickland test, we review whether the performance of Gabaree’s attorney “fell below an objective standard of reasonableness,” keeping in mind that “counsel’s function ... is to make the adversarial testing process work.” Strickland, 466 U.S. at 688, 690, 104 S.Ct. 2052. Our review is “highly deferential” to counsel’s performance, and we determine whether counsel’s conduct was reasonable based on “the facts of the particular case, viewed as of the time of counsel’s conduct.” Id. at 689-90, 104 S.Ct. 2052. We are in an awkward posture in this case because counsel was unable to remember her trial strategy or her rationale for any decision made during the trial. Thus, as we have done in this situation in the past, “we will examine'counsel’s trial tactics and strategy as revealed by the state court record because that record best reflects ‘counsel’s perspective at the time.’ ” Fretwell v. Norris, 133 F.3d 621, 624 (8th Cir.1998) (quoting Strickland, 466 U.S. at 689, 104 S.Ct. 2052); see Paul v. United States, 534 F.3d 832, 837 (8th Cir.2008). Though counsel in this case repeated that she had no memory of the case, she gave possible reasons for her decisions regarding the testimony of the two doctors. Thus, as the district court did, we will consider her proposed tactics and strategy, in conjunction with the state court record, so that we may best perceive “counsel’s perspective at the time.” Fretwell, 133 F.3d at 624 (quotation omitted).5
The state court first concluded that counsel gave sufficient possible reasons why she might not have objected to the testimony of Dr. Kelly, crediting her belief that she did not want to call attention to the testimony. The court also recognized, as counsel had testified in the post-conviction hearing, that she simply may have missed the objectionable testimony. Nonetheless, the state court ruled that, because counsel “identified reasons why she might have chosen not to object as a matter of trial strategy[,] and the record does not conclusively establish that such a strategic decision would have been unreasonable” (emphasis added), Gabaree could not overcome the presumption that counsel’s purported strategic reasons for not objecting were reasonable. The district court rejected the state court’s rationale and concluded that “in this context,” in which the only inculpating evidence was the inconsistent testimony of the girls, counsel’s putative strategy not to draw attention to Dr. Kelly’s corroboration of the girls was unreasonable.
We agree with the district court that, in the context of this case, the state court’s [998]*998application of Strickland is unreasonable. The main evidence against Gabaree on these counts was the testimony of the girls; there was no medical' evidence nor eyewitness accounts to support their testimony. And the girls’ testimony was inconsistent: The second trial was ordered because, the state post-conviction court decided, the State’s case was weak and based predominately on the believability of the two young girls, who had recanted their allegations. Faced with this situation, we do not think it reasonable not to object when an expert witness tells the jury he believed the girls were truthful when speaking with a different doctor. Dr. Kelly’s testimony, in fact, was inadmissible bolstering of other witnesses (the girls) and, as the state court recognized, should have been excluded after a proper objection. See Olesen v. Class, 164 F.3d 1096, 1102 (8th Cir.1999) (concluding that counsel rendered deficient performance by failing to object to psychologist’s opinion that child-victim would not have fabricated sexual-assault allegations); State v. Churchill, 98 S.W.3d 536, 539 (Mo. banc 2003) (noting that trial courts must reject expert testimony “concerning a specific victim’s credibility as to whether they have been abused ... because it usurps the decision-making function of the jury”). Yet counsel did not object to this statement “very possibly” because she “just missed it.” Given the weak nature of the case, and Dr. Kelly’s improper opinion on the contested issue of the girls’ credibility, not objecting to avoid highlighting this inadmissible testimony, if counsel heard the testimony at all, would not be objectively reasonable trial strategy.
The state court also concluded that Dr. Sisk’s testimony was inadmissable propensity evidence but ruled that counsel’s purported strategy not to highlight the damning testimony was not unreasonable. The state court also found reasonable counsel’s other possible strategy of “arguing away” Dr. Sisk’s testimony by characterizing his opinion as only a suggestion, in a “complicated, T am a doctor’ sort of way,” that Gabaree needed parenting classes and counsel’s worry that objecting to or impeaching Dr. Sisk’s testimony would reveal that there had been an earlier trial. The district court again concluded that counsel’s trial strategy was unreasonable. The court noted counsel’s failure to recognize, at the evidentiary hearing, that Dr. Sisk’s testimony was inadmissible and her failure to interpret the testimony as harmful. Moreover, the court pointed out, counsel did not employ her supposed strategy of arguing away Dr. Sisk’s opinion. Last, the court found no reason why objecting would have revealed to the jury that there had been an earlier trial.
We again agree with the district court. Dr. Sisk testified that Gabaree, if he was following the beliefs he expressed to Dr. Sisk, was “probably abusing or neglecting the children,” including by using them to meet his sexual needs. Like Dr. Kelly’s .testimony, Dr. Sisk’s opinion improperly concluded for the jury that Gabaree had acted on his expressed beliefs and sexually abused his children. His testimony, with proper objection, would have been excluded. See United States v. Johnson, 439 F.3d 884, 887 (8th Cir.2006) (concluding that sexually explicit stories written by defendant should have been excluded from trial for possession of child pornography because stories improperly suggested his propensity to possess child pornography); Gabaree, 290 S.W.3d at 181 (“Because Dr. Sisk’s testimony about Mr. Gabaree’s beliefs constituted propensity evidence, it was not admissible.”)
Moreover, the state court’s conclusions regarding Dr. Sisk have no support in the record. First, counsel said she did not want to highlight Dr. Sisk’s testimony. Yet the record from the state court reveals [999]*999that she did just that: On cross-examination she elicited additional statements from Dr. Sisk that bolstered his testimony, including that he did not believe someone would answer his questionnaire as Gabaree had and yet not abuse his children. This is not in line with a strategy to ignore the testimony — even if that strategy might be reasonable under some circumstances— and by itself could establish deficient performance. See United States v. Villalpando, 259 F.3d 934, 939 (8th Cir.2001) (affirming district court’s conclusion that counsel rendered deficient performance by eliciting from government witness harmful testimony that defendant was “threatening and murderous”). Nor did counsel follow her other purported strategy of recharac-terizing Dr. Sisk’s testimony as.mere doctor’s advice. Counsel mentioned neither doctor nor their testimony during her closing argument and said nothing to suggest that Dr. Sisk’s testimony related only to Gabaree’s need for parenting classes. We cannot impute to counsel a trial strategy that the record reveals she did not follow. See Alcala v. Woodford, 334 F.3d 862, 871 (9th Cir.2003); Griffin v. Warden, 970 F.2d 1355, 1358-59 (4th Cir.1992). Last, counsel’s ancillary worry about revealing the earlier trial also is unreasonable. Nothing in the record suggests she could not have objected to Dr. Sisk’s testimony without mentioning the first trial. The state court’s determination of counsel’s strategic decisions thus “resembles more a post hoc rationalization of counsel’s conduct than an accurate description of [her] deliberations.” Wiggins v. Smith, 539 U.S. 510, 526-27, 123 S.Ct. 2527, 156 L.Ed.2d 471 (2003). Because the state court’s conclusions rely on this unreasonable determination of the facts, § 2254(d) does not require that we adhere to that court’s decision on this issue, and our review is de novo. See Rice v. White, 660 F.3d 242, 257 (6th Cir.2011); Jones v. Walker, 540 F.3d 1277, 1288 & n. 5 (11th Cir.2008); Taylor v. Maddox, 366 F.3d 992, 1008 (9th Cir.2004). We conclude that, unlike in Fretwell, and based on the state-court record in this ease, there was no reasonable strategic reason for counsel not to object to the testimony of Dr. Sisk.
Though we apply a strong presumption that counsel acted reasonably, the state court record belies that presumption and reveals no reasonable strategic reason not to object to either doctor’s testimony. Nor do we believe that counsel’s possible “trial tactics and strategy,” even if in line with what counsel proposed they might have been, were reasonable. We thus conclude that the state court’s determination that counsel’s performance met professional norms involves an unreasonable application of Strickland and is based on an unreasonable determination of the facts. Counsel’s performance, we conclude, was objectively unreasonable.
ii. Prejudice
Under the second factor of Strickland, we review whether counsel’s deficient performance prejudiced Gabaree; that is, we must determine whether Gabar-ee has shown “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. The state court, however, did not conduct this analysis and instead determined only that counsel had not performed deficiently. Thus, our review is not constrained by § 2254(d), and we review the question of prejudice de novo. Rompilla v. Beard, 545 U.S. 374, 390, 125 S.Ct. 2456, 162 L.Ed.2d 360 (2005); Worthington v. Roper, 631 F.3d 487, 495 (8th Cir.2011).
Based on our review of the doctors’ testimony, we agree with the district court that Gabaree suffered prejudice. The doctors’ inadmissible testimony strengthened [1000]*1000an otherwise very weak case. Dr. Kelly told the jury that he believed the girls were credible despite never having discussed with them their allegations of sexual abuse. That statement improperly bolstered the credibility of the girls, which was a central issue in the trial. Dr. Sisk testified that, because Gabaree had answered questionnaires in a certain way, he likely acted in accordance with those answers and sexually abused his children, a conclusion that counsel helped bolster with her cross-examination. This testimony improperly drew a conclusion about Gabar-ee’s actions that was the jury’s to make. Because counsel did not object to these inadmissible statements, the jury entered deliberations with the impression that the doctors’ testimony was credible, correct, and incontestable.
Absent the doctors’ testimony, there was little evidence against Gabaree with respect to the sex-abuse charges. The State presented no medical evidence and no independent eyewitness accounts of sexual abuse. The testimony of the young girls was conflicting and inconsistent. As the district court noted, the doctors’ testimony improperly “shored up” the weaknesses in the State’s case. By not challenging that testimony, counsel failed “to make the adversarial testing process work.” Strickland, 466 U.S. at 690,104 S.Ct. 2052. Had counsel performed effectively, we believe, there is a reasonable chance that the outcome of the trial would have been different. We thus conclude that trial counsel rendered ineffective assistance regarding the charges of sodomy and child molestation.
B. Child-abuse Convictions
In the second issue on appeal, Ga-baree argues that the district court should have concluded that he also was prejudiced by counsel’s performance with regard to his convictions for child abuse. He asserts that the testimony of Drs. Kelly and Sisk improperly bolstered the girls’ testimony about physical abuse and insists that, but for counsel’s performance, he would have been acquitted of these charges.
The district court concluded that Gabar-ee was not' prejudiced by counsel’s performance with respect to the child-abuse charges. The court noted that other evidence at trial corroborated the testimony of the girls and the doctors that he had physically abused them: Bruises and marks suggesting physical abuse were found on the girls; expert testimony confirmed that the marks were consistent with the items that the girls testified had been used on them (belts, belt buckles, cigarettes, and cigarette lighters) and were not caused by accident; pictures of the girls’ injuries were displayed to the jury; Gabaree’s mother testified that she thought her son’s acts of purported discipline were too severe; and Gabaree admitted whipping the girls, though he insisted that it was to discipline them and was not inappropriate. Nor did the girls’ testimony suffer from the same inconsistencies as their testimony regarding the alleged sexual abuse.
. We agree with the district court. The State’s case in support of these charges was significantly stronger than its evidence of sexual abuse. Even without Dr. Sisk’s improper testimony that Gabaree likely was acting in accordance with his questionnáire answers, the wealth of other admissible evidence — including Dr. Kelly’s testimony that the marks he observed on the girls were consistent with intentional abuse — makes it unlikely that counsel’s performance affected the verdict on these counts.
III. Conclusion
For the above reasons, we affirm the district court’s judgment granting Gabar-ee’s petition with respect to his convictions [1001]*1001for sodomy and child molestation and denying his petition with respect to his convictions for child abuse. The order to retry Gabaree on the sodomy and child-molestation charges also is affirmed.