An Oklahoma jury convicted Tremane Wood of first-degree felony murder for the killing of Ronnie Wipf during a botched robbery. The jury found Oklahoma had proved three aggravating circumstances associated with the murder, and the mitigating circumstances did not outweigh them. The jury accordingly sentenced Wood to death.
The conclusion of Wood's trial was only the start of his case's long legal odyssey. Wood directly appealed his conviction to the Oklahoma Court of Criminal Appeals, advancing, as relevant here, two primary arguments. First, he claimed his trial counsel performed ineffectively at the sentencing stage. Second, he argued the "heinous, atrocious, or cruel" aggravating circumstance could not be constitutionally applied to this case given the dearth of evidence that Mr. Wipf suffered before death. The OCCA ordered an evidentiary hearing on the ineffectiveness issue, but ultimately affirmed Wood's conviction and death sentence.
Wood then filed an application for post-conviction relief in state court. He claimed his appellate counsel performed ineffectively on direct appeal, including at the evidentiary hearing. The OCCA again denied relief.
Wood next filed a habeas petition under 28 U.S.C. § 2254 in the Western District of Oklahoma. The district court denied the petition. We granted Wood certificates of *874appealability on whether his trial and appellate counsel performed ineffectively.1
During the course of this appeal, we decided Pavatt v. Royal , 859 F.3d 920 (10th Cir. 2017), opinion amended and superseded on denial of rehearing on July 2, 2018 by Pavatt v. Royal , 894 F.3d 1115 (10th Cir. 2017), a challenge to Oklahoma's application of the heinous, atrocious, and cruel (HAC) aggravator in that case. Based on Pavatt , we granted an additional COA on whether the HAC aggravating circumstance could be constitutionally applied to the facts of this case.
For the reasons discussed below, we AFFIRM the district court's denial of the petition for habeas relief.
I. Background
We begin by explaining the underlying facts and the numerous previous proceedings.
A. The Crime
On December 31, 2001, Tremane Wood rang in the new year at a brewery in Oklahoma City.2 Wood's brother Zjaiton,3 Zjaiton's girlfriend Lanita, and Wood's ex-girlfriend Brandy joined him. At some point during the festivities, Lanita and Brandy began talking with two fellow brewery patrons-Ronnie Wipf and Arnold Kleinsasser. Wipf and Kleinsasser invited the women back to their motel to continue celebrating. After conferring with Wood and Zjaiton, the women agreed to leave with their new acquaintances.
But something nefarious was afoot. Before leaving, Wood, Zjaiton, and the women concocted a plan. The women would pretend to be prostitutes and, once Wipf and Kleinsasser secured the money to pay them, the Wood brothers would show up at the motel and rob the two men.
The women put the plan into action. At the motel, Wipf and Kleinsasser agreed to pay them $210 to have sex. The men had no cash on hand, however, so they all drove to an ATM. Meanwhile, Wood and Zjaiton waited outside the motel. Once the women, Wipf, and Kleinsasser arrived back at the room, Wood and Zjaiton pounded on the door. Mr. Wipf opened the door and the brothers barged in. Both were armed-Wood with a knife and Zjaiton with a gun.4 The women ran out the door and a fight ensued.
*875At first, Wood fought Mr. Wipf alone. But Zjaiton eventually joined the fray and helped Wood fight Wipf. After Zjaiton had joined the fight, Kleinsasser saw that Mr. Wipf was covered in blood. At some point during this brawl, Mr. Wipf was fatally stabbed in the chest. Autopsy diagrams and pictures of Wipf's body reveal his face was bloody and bruised, and he had a deep cut on his right hand.
B. Wood's Murder Trial
Oklahoma charged Wood, Zjaiton, Lanita, and Brandy with numerous crimes, including first-degree felony murder. The state sought the death penalty against both Wood and Zjaiton, arguing that four aggravating circumstances warranted the death sentence. One of those circumstances was that the murder was especially heinous, atrocious, or cruel.
John Albert represented Wood at the guilt and sentencing phases of trial.5 Eventually, the court severed Wood's trial from the other defendants'. At the guilt phase, the jury convicted Wood of first-degree felony murder.6
At the sentencing phase, three witnesses testified on Wood's behalf. Andre Taylor, a family friend, testified that Wood was well liked, was not a bad person, and loved his children.
Dr. Hand, a licensed psychologist, also testified. He first emphasized how chaos defined Wood's family life. To demonstrate this, he cited numerous Department of Human Services (DHS) records in which Wood's mother, Linda, claimed her husband and Wood's father, Raymond Gross, abused her. And Dr. Hand explained that when Wood did get in trouble, he was usually following Zjaiton's lead. When cross examined, however, Dr. Hand stumbled a bit. The prosecution asked him about Wood's juvenile records, specifically those detailing Wood's previous assault charge. Though Dr. Hand recalled reviewing a large stack of records, he could not recall those specific ones.
Lastly, Wood's mother Linda testified. She painted a dark portrait of her relationship with Gross. They had, she said, "a very abusive relationship. I had been beaten many, many times in front of my children. Tied up. Dragged down the highway. My bones broke." Tr., 04/05/2004, at 91. But Linda, too, floundered a bit on cross examination. The prosecution attacked her allegations of abuse, emphasizing how DHS had found most of them invalid.
The jury sentenced Wood to death.
C. Direct Appeal to the OCCA
Wood retained new lawyers for his direct appeal to the Oklahoma Court of Criminal Appeals. Wood also applied for a Rule 3.11 evidentiary hearing to develop additional evidence about his claim trial counsel performed ineffectively. In support of the application, Wood put forward seventeen affidavits and over 1,200 pages of juvenile records.
The OCCA granted the application and remanded the case to the trial court, instructing it to hold an evidentiary hearing and answer five factual questions.7 The *876court then held a three-day hearing during which Wood presented twenty-three witnesses. Ten witnesses testified about Wood's life history, including his mother, his father, and his brothers Andre and Zjaiton. Much of this testimony hit on the same themes Linda and Dr. Hand had testified to at trial-albeit, in far more detail.
But the evidentiary hearing did not merely recite the trial testimony with a few extra details; new evidence did emerge. For example, Andre Wood testified that Gross abused Wood as well as Linda. And Gross testified for the first time. He admitted to once handcuffing Linda to his car as punishment for sleeping with his nephew. He also confessed to pushing Linda in front of his sons. But Gross denied other instances of abuse, such as knocking Linda's teeth out or dragging her on the ground after he handcuffed her. Similarly, Gross admitted he had whipped his sons before, but insisted he never did so sadistically.
Trial counsel, Mr. Albert, also signed an affidavit and testified at the hearing. In his affidavit, he acknowledged time constraints caused him to "not prepare enough of a mitigation case to effectively represent and defend" Wood. EH Vol. 1 Ex. M. In his testimony, Mr. Albert similarly admitted he failed to hire an investigator. But he stressed that he had interviewed Wood, Linda, and Zjaiton. And he emphasized that he and Dr. Hand had successfully employed this same strategy in other cases.
Wood's counsel at the evidentiary hearing also wanted to introduce testimony from Dr. Kate Allen, a sociologist with a doctoral degree in family sociology who had worked as a social worker and professor for thirty-five years. The trial court granted the state's objection to Dr. Allen testifying. She was not qualified as an expert witness, the trial court concluded, and her testimony would have been cumulative.
After hearing three days of testimony, the trial court entered findings of fact answering the OCCA's questions. The OCCA then permitted the parties to submit ten-page supplemental briefs to challenge these findings. Wood's counsel did so, and the brief attacked the court's answers to the five questions. The brief did not, however, mention the exclusion of Dr. Allen's testimony.
The OCCA ultimately affirmed Wood's conviction and sentence.
D. Post-Conviction Appeal in the OCCA
Next, Wood moved for post-conviction relief in the OCCA on a number of grounds. As relevant here, Wood alleged his appellate counsel performed ineffectively on direct appeal, including during the Rule 3.11 evidentiary hearing. The OCCA denied relief on all of these claims.8
E. Wood's Habeas Petition
Wood filed a 28 U.S.C. § 2254 petition in the Western District of Oklahoma, raising ten issues. The district court denied relief on all of them. Wood then sought certificates of appealability on numerous claims; we granted COAs on two issues. First, whether Wood's trial counsel performed ineffectively at the sentencing stage by (1)
*877failing to adequately investigate, select, prepare, and present mitigation lay witnesses, (2) failing to prepare and present the mitigation expert witness, Dr. Hand, and (3) failing to investigate, obtain, and present Linda Wood's medical records, and Wood's juvenile, school, medical, mental health, and DHS records. And second, whether Wood's appellate counsel on direct appeal performed ineffectively by failing to challenge the exclusion of Dr. Allen's testimony in the supplemental brief.
Wood then requested leave to certify additional issues. We granted his request for a COA on the claim appellate counsel performed ineffectively at the evidentiary hearing for three additional reasons: (1) failing to obtain and use documents that would have undercut Gross's denials of abuse, (2) not alerting the court about Mr. Albert's subsequent disciplinary proceedings, and (3) failing to point out the trial court's factual errors in supplemental briefing to the OCCA.
After briefing concluded, Wood asked us to certify two additional issues in light of our circuit's decision in Pavatt v. Royal , 859 F.3d 920 (10th Cir. 2017), opinion amended and superseded on denial of rehearing on July 2, 2018 by Pavatt v. Royal , 894 F.3d 1115 (10th Cir. 2017). We granted a COA on one of the issues-whether constitutionally sufficient evidence supported the application of the heinous, atrocious, or cruel aggravating circumstance.
II. Standard of Review
When reviewing whether the federal district court erred in denying habeas relief, we review its legal analysis de novo and its factual findings for clear error. Smith v. Duckworth , 824 F.3d 1233, 1241-42 (10th Cir. 2016). But in proceedings under 28 U.S.C. § 2254, the Antiterrorism and Effective Death Penalty Act of 1996 (AEDPA) significantly limits our review. Under AEDPA, when a state court adjudicated a petitioner's claim on the merits, we cannot grant relief unless that 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 proceedings.
§ 2254(d)(1)-(2).
"Clearly established Federal Law" refers to the Supreme Court's holdings, not its dicta. See, e.g. , Williams v. Taylor , 529 U.S. 362, 412, 120 S.Ct. 1495, 146 L.Ed.2d 389 (2000). A state-court decision is only contrary to clearly established federal law if it "arrives at a conclusion opposite to that reached by" the Supreme Court, or "decides a case differently" than the Court on a "set of materially indistinguishable facts." Id. at 412-13, 120 S.Ct. 1495. But a state court need not cite the Court's cases or, for that matter, even be aware of them. So long as the state-court's reasoning and result are not contrary to the Court's specific holdings, § 2254(d)(1) prohibits us from granting relief. See Early v. Packer , 537 US. 3, 9, 123 S.Ct. 362, 154 L.Ed.2d 263 (2002) (per curiam).
A state court's decision unreasonably applies federal law if it "identifies the correct governing legal principle" from the relevant Supreme Court decisions but applies those principles in an objectively unreasonable manner. Wiggins v. Smith , 539 U.S. 510, 520, 123 S.Ct. 2527, 156 L.Ed.2d 471 (2003). Critically, an "unreasonable application of federal law is different from an incorrect application of federal law."
*878Williams , 529 U.S. at 410, 120 S.Ct. 1495 (2000). "[E]ven a clearly erroneous application of federal law is not objectively unreasonable." Maynard v. Boone , 468 F.3d 665, 670 (10th Cir. 2006). Rather, a state court's application of federal law is only unreasonable if "all fairminded jurists would agree the state court decision was incorrect." Frost v. Pryor , 749 F.3d 1212, 1225 (10th Cir. 2014).
Finally, a state-court decision unreasonably determines the facts if the state court "plainly misapprehend[ed] or misstate[d] the record in making [its] findings, and the misapprehension goes to a material factual issue that is central to petitioner's claim." Byrd v. Workman , 645 F.3d 1159, 1170-72 (10th Cir. 2011). But this "daunting standard" will be "satisfied in relatively few cases." Id. That is because the state court's decision must be "based on an unreasonable determination of the facts." Id.9
AEDPA thus "erects a formidable barrier to federal habeas relief." Burt v. Titlow , 571 U.S. 12, 16, 134 S.Ct. 10, 187 L.Ed.2d 348 (2013). Congress crafted such a deferential standard to ensure review under § 2254 serves only as " 'a guard against extreme malfunctions in the state criminal justice systems,' not a substitute for ordinary error correction through appeal." Harrington v. Richter , 562 U.S. 86, 102-03, 131 S.Ct. 770, 178 L.Ed.2d 624 (2011) (quoting Jackson v. Virginia , 443 U.S. 307, 322 n.5, 99 S.Ct. 2781, 61 L.Ed.2d 560 (1979) ).
With the limited nature of our review in mind, we turn to Wood's claims.
III. Analysis
Wood raises three general claims. First, that his trial counsel performed ineffectively. Second, that his appellate counsel on direct appeal performed ineffectively. And finally, that applying the HAC aggravator to the facts of this case violated the Constitution.
A. Ineffective Assistance of Trial Counsel
We first summarize the trial-counsel-ineffectiveness standard before turning to Wood's claim his trial counsel performed ineffectively.
1. Legal Standards
In Strickland v. Washington , 466 U.S. 668, 104 S.Ct. 2052, 80 L.Ed.2d 674 (1984), the Supreme Court laid out the now-familiar framework for ineffective assistance of counsel claims. Under it, Wood must demonstrate both that his counsel performed deficiently and that he suffered prejudice from this deficient performance. Strickland , 466 U.S. at 687, 104 S.Ct. 2052.
Counsel performs deficiently if his representation falls "below an objective standard of reasonableness" under prevailing professional norms. Id. at 688-89, 104 S.Ct. 2052. At the sentencing stage of a capital trial, counsel has a duty to "thoroughly investigat[e] and present[ ] mitigating evidence." Cargle v. Mullin , 317 F.3d 1196, 1221 (10th Cir. 2003). Failing to do so can constitute deficient performance. Id. And counsel's deficient performance at the *879sentencing stage prejudices the defendant if, but for counsel's unprofessional errors, "there is a reasonable probability that one juror would have chosen a sentence other than death." Matthews v. Workman , 577 F.3d 1175, 1190 (10th Cir. 2009).
Importantly, when we evaluate counsel's performance we must do so through a "most deferential" lens. Richter , 562 U.S. at 105, 131 S.Ct. 770. Though there are ordinarily "countless ways to provide effective assistance," it is still " 'all too tempting' to 'second-guess counsel's assistance after conviction or adverse sentence.' " Id. at 106, 131 S.Ct. 770 (quoting Strickland , 466 U.S. at 689, 104 S.Ct. 2052 ). We must resist this temptation and instead make "every effort ... to eliminate the distorting effects of hindsight." Strickland , 466 U.S. at 689, 104 S.Ct. 2052. We thus "indulge a strong presumption that counsel's conduct falls within the wide range of reasonable professional assistance." Id.
"Surmounting Strickland 's high bar" is therefore "never an easy task." Richter , 562 U.S. at 105, 131 S.Ct. 770. Yet raising Strickland claims in a habeas petition makes success all the more difficult. That is because, taken together, AEDPA and Strickland render our review "doubly deferential." Knowles v. Mirzayance , 556 U.S. 111, 123, 129 S.Ct. 1411, 173 L.Ed.2d 251 (2009). We "take a 'highly deferential' look at counsel's performance, through the deferential lens of § 2254." Cullen v. Pinholster , 563 U.S. 170, 190, 131 S.Ct. 1388, 179 L.Ed.2d 557 (2011) (quoting Strickland , 466 U.S. at 689, 104 S.Ct. 2052 ).
2. Discussion
We granted a COA on Wood's claim his trial counsel performed ineffectively at the sentencing stage by failing to (1) adequately investigate, select, prepare, and present mitigation lay witnesses; (2) prepare and present the mitigation expert witness, Dr. Hand; and (3) investigate, obtain, and present Linda Wood's medical records, and Wood's juvenile, school, medical, mental health, and DHS records. We address each issue in turn.
a. Investigating, Selecting, and Preparing Lay Witnesses
We first consider Wood's claim that his trial counsel performed ineffectively by failing to investigate and present additional witnesses. The OCCA denied relief on this claim, concluding Wood's trial counsel did not perform deficiently, and Wood suffered no prejudice. In concluding counsel did not perform deficiently, the OCCA emphasized that "[e]vidence of [Wood's] chaotic home life and background was presented to the jury through both an expert and lay witness." Wood v. State , 2007 OK CR 17, 158 P.3d 467, 481 (Okla. Crim. App. 2007). Though the OCCA conceded that "other witnesses not called at trial could have provided further detail to support the mitigation evidence," it nonetheless found "the trial court correctly concluded that the material testimony from those credible witnesses not called at trial was nonetheless presented to the jury." Id. Further, the OCCA held Wood suffered no prejudice because he "failed to show that the outcome of his case would have been different had the credible evidence developed at the evidentiary hearing been presented during his sentencing proceeding." Id.
Wood argues AEDPA does not bar relief for a number of reasons. None has merit.
First , Wood contends we can grant relief because the OCCA's decision was "based on an unreasonable determination of the facts." § 2254(d)(2). More specifically, he insists the OCCA's finding that the "material testimony from those credible witnesses not called at trial was nonetheless *880presented to the jury" qualifies as an unreasonable factual determination. Wood , 158 P.3d at 481. But the OCCA's determination about what evidence was "material" is not a factual finding at all.
In Williams , the Supreme Court explained that the prejudice inquiry is a "mixed question of law and fact." 529 U.S. at 371, 120 S.Ct. 1495. And the Court recognized that the "factual part of the mixed question" was whether evidence had, in fact, been "presented at trial." Id. at 398, 120 S.Ct. 1495. The "legal part" of the prejudice analysis, in contrast, related to the "strength of the ... evidence." Id.10 Accordingly, the OCCA could have made a factual error if it concluded evidence had been presented at trial when it, in fact, had not been. But Wood alleges no such error. Instead, he claims the OCCA made an erroneous factual determination when it categorized as immaterial the evidence developed at the evidentiary hearing but not presented at trial. This categorization plainly relates to the "legal part" of the prejudice analysis-the strength of the purportedly immaterial evidence, and whether it would have affected the proceeding's outcome. Simply put, since the factual determination Wood claims is unreasonable is not a factual determination, § 2254(d)(2) is inapplicable.
Second , Wood contends the OCCA unreasonably applied Strickland by concluding his trial counsel did not perform deficiently by failing to call more lay witnesses, and that he suffered no prejudice from this failure. We need not consider the OCCA's deficient-performance analysis because we can resolve this issue based solely on the OCCA's conclusion on prejudice. After a careful review of the record, we cannot conclude that "all fairminded jurists would agree" the OCCA unreasonably applied Strickland when it concluded trial counsel's failure to call additional lay witnesses did not prejudice Wood. Grant v. Royal , 886 F.3d 874, 909 (10th Cir. 2018). Indeed, the OCCA properly recognized that the themes developed at the evidentiary hearing were also developed at Wood's sentencing, albeit in less detail. And while some testimony at the hearing could be considered "new"- such as allegations Gross abused Wood and his brother, not just Linda-this evidence still related to the same themes counsel developed at trial: Wood's formative years were, as Dr. Hand testified, defined by chaos, and abuse allegations swirled around his home. The testimony developed at the evidentiary hearing was thus cumulative of the evidence trial counsel actually presented during the sentencing stage.
This is not to say Wood's trial counsel offered a textbook mitigation defense. We agree Wood's mitigation case might have been stronger had some of the witnesses from the evidentiary hearing testified during sentencing. But because the evidence and themes developed at the hearing were substantially similar to those developed at trial, the OCCA's conclusion Wood suffered no prejudice was objectively reasonable. Accordingly, AEDPA's deferential framework prevents us from disturbing the state court's decision.
Third , Wood advances two arguments that relate to the fact the court at the evidentiary hearing barred Dr. Allen from testifying as an expert witness. To start, we doubt these contentions fit within the *881COA we granted on the issue, which was limited to trial counsel's alleged failure to investigate and call additional lay , not expert, witnesses. But even if these arguments do fall within the COA, neither has any merit.
Wood first argues the OCCA unreasonably applied Strickland because it failed to consider Dr. Allen's testimony and report. Wood contends that Strickland required the OCCA to consider the "totality of the evidence" developed at the evidentiary hearing, which included this report and testimony. 466 U.S. at 695, 104 S.Ct. 2052. But Dr. Allen did not testify at the evidentiary hearing because the court concluded she was not qualified to do so. Accordingly, the OCCA could not have considered this testimony because Dr. Allen never actually testified; the evidence Wood faults the OCCA for not considering simply did not exist. And while the court admitted Dr. Allen's report into the record at the evidentiary hearing, it did so only to ensure its decision not to allow her to testify could "be fully reviewed at a later time." Tr., 2/27/06, at 220. The OCCA thus could not consider this report as evidence of trial counsel's alleged ineffectiveness. It instead could only have considered the report if Wood challenged the OCCA's exclusion of Dr. Allen's testimony in his supplemental brief. But he did not do so. The OCCA therefore did not need to (and in fact could not) consider Dr. Allen's report.11
Wood also contends the trial court's refusal to allow Dr. Allen to testify at the Rule 3.11 hearing directly conflicts with Lockett v. Ohio , 438 U.S. 586, 98 S.Ct. 2954, 57 L.Ed.2d 973 (1978) (plurality), and Eddings v. Oklahoma , 455 U.S. 104, 102 S.Ct. 869, 71 L.Ed.2d 1 (1982). But we are reviewing the OCCA's decision affirming Wood's sentence and conviction; we are not reviewing any and every issue decided by the state trial court at an evidentiary hearing. And the OCCA did not rule on whether the trial court properly barred Dr. Allen from testifying because Wood failed to raise this error in his supplemental brief. This claim therefore cannot serve as a basis for meeting § 2254(d) 's requirements.
b. Preparing and Presenting Dr. Hand
We also granted a COA on whether Wood's trial counsel performed ineffectively by failing to prepare and present Dr. Hand's expert testimony. Wood claims his counsel failed to obtain all of his juvenile records before trial and, as a result, could not adequately prepare Dr. Hand before his testimony. In Wood's view, this had disastrous consequences. On cross examination, Dr. Hand floundered when confronted with the juvenile records detailing Wood's prior assault charge: he admitted that, though he had reviewed many juvenile records, he did not recall those specific ones. Thus, Wood insists his trial counsel performed ineffectively by failing to obtain and utilize these records.
The OCCA again rejected Wood's trial-ineffectiveness claim based on both Strickland prongs: counsel did not perform deficiently and Wood endured no prejudice. In so concluding, the OCCA stated the district court's finding that "the defense psychological expert had possession of [Wood's] background records, including his relevant juvenile records" was "supported by the record." Wood , 158 P.3d at 480. Wood argues AEDPA does not apply because this qualifies as an unreasonable determination of fact under § 2254(d)(2). He highlights how the OCCA found trial counsel *882did not have Wood's Central Oklahoma Juvenile Center (COJC) records, which detailed Wood's academic and personal success during his time at the Center. Thus, when the OCCA stated trial counsel had all of Wood's "relevant juvenile records," it necessarily found the COJC records were not "relevant." This, Wood insists, qualifies as an unreasonable determination of fact.
But whether the CJOC records were "relevant" is not a factual determination. Evidence is considered relevant if it has a "tendency to make a fact more or less probable" or is "of consequence in determining" something. Fed. R. Evid. 401. Whether the CJOC records were relevant, then, relates to whether the records supported Wood's mitigation case. This is a quintessentially legal determination. See Williams , 529 U.S. at 398, 120 S.Ct. 1495.
Even if the OCCA's comment on relevance did constitute a factual determination, and even assuming this determination was unreasonable, § 2254(d)(2) would still bar relief. This is because "an 'unreasonable determination of the facts' does not, itself, necessitate relief." Byrd , 645 F.3d at 1170-72 (quoting Collier v. Norris , 485 F.3d 415, 423 (8th Cir. 2007) ). Rather, to receive relief under § 2254(d)(2) the OCCA's adjudication of this claim must have been "based on" the unreasonable determination. Id.
The OCCA did not base its prejudice conclusion on its finding about what records were relevant. Rather, that conclusion was based on the fact that even if all the "credible evidence developed at the evidentiary hearing" had "been presented during [Wood's] capital sentence proceeding," the proceeding's outcome would not have been different. Wood , 158 P.3d at 481. In other words, the OCCA based its prejudice conclusion on the strength, weakness, and cumulativeness of the mountain of evidence presented at the evidentiary hearing-namely, the twenty new witnesses who testified and the thousands of documents produced. Its prejudice conclusion did not rest on its finding about whether an exceedingly small subset of that evidence-the CJOC records-were relevant. Consequently, the OCCA's decision was not based on its finding that the CJOC records were not "relevant."
Thus, § 2254(d)(2) 's bar on relief is not satisfied on this basis.
c. Wood's Juvenile Records and Linda's Medical Records
AEDPA bars relief on the claim Wood's counsel failed to investigate and present all of Wood's juvenile records for the reasons we just described: the OCCA did not unreasonably determine the facts by concluding Wood's counsel possessed all the relevant juvenile records.
Beyond its general argument about the records we already addressed, Wood's briefing does not advance any specific arguments about Linda's medical records. We thus do not grant relief on this basis.12
*883* * *
In sum, § 2254(d) prohibits us from granting relief on Wood's claim his trial counsel performed ineffectively. We therefore must affirm the district court's denial of the petition on that issue.
B. Ineffective Assistance of Appellate Counsel on Direct Appeal
Wood also contends his appellate counsel performed ineffectively on direct appeal, including at the Rule 3.11 hearing, in five ways: (1) by not challenging, in the supplemental brief, the trial court's conclusion Dr. Allen could not testify at the evidentiary hearing; (2) by failing to obtain and use documentary evidence of Raymond Gross's abuse that would have undermined his testimony; (3) by not updating the record with evidence his trial counsel had been suspended from the practice of law and held in contempt of court in a different state proceeding; (4) by not using the supplemental brief to correct the factual errors the evidentiary hearing court made in its findings of facts; and (5) by neglecting to raise a claim trial counsel performed ineffectively by failing to object to juror separation.
Strickland 's same deferential framework (which is doubly deferential when coupled with AEDPA) applies to ineffectiveness claims based on appellate counsel's performance. Thus, to demonstrate ineffective assistance of appellate counsel, a petitioner "must establish both (1) that his counsel's performance fell below an objective standard of reasonableness, and (2) that there is a reasonable probability that, but for counsel's unreasonable errors, the outcome of his appeal would have been different." Ellis v. Hargett , 302 F.3d 1182, 1186-87 (10th Cir. 2002) (citing Williams , 529 U.S. at 390-91, 120 S.Ct. 1495 and Strickland , 466 U.S. at 688, 694, 104 S.Ct. 2052 ).
Under this doubly deferential framework, Wood is not entitled to relief on any of these five claims.
1. Challenging the Exclusion of Dr. Allen's Testimony
As we already explained, the trial court barred Dr. Allen from testifying as an expert at the evidentiary hearing because it concluded she was unqualified and her testimony would have been cumulative. After the evidentiary hearing court issued its findings of fact, it allowed the parties to submit ten-page supplemental briefs raising any issues from the hearing. Wood's supplemental brief failed to challenge the exclusion of Dr. Allen's testimony.
The OCCA again denied relief because counsel did not perform deficiently and Wood was not prejudiced. In so concluding, Wood contends the OCCA unreasonably applied Strickland .13 We disagree. We need not consider the OCCA's deficient-performance analysis. Since the OCCA applied Strickland in an objectively reasonable manner when it concluded Wood suffered no prejudice, AEDPA precludes us from granting relief. See Strickland , 466 U.S. at 695, 104 S.Ct. 2052 (explaining courts need not "address both components of" the Strickland inquiry if "the defendant makes an insufficient showing on one").
It was reasonable for the OCCA to conclude appellate counsel's failure to challenge the exclusion of Dr. Allen's testimony in the supplemental brief caused Wood no prejudice. This is because Dr. *884Allen's testimony would have been substantially cumulative of other evidence already presented at both the hearing and at trial.
Dr. Allen planned to testify about six topics: (1) Wood's early childhood development and the effect his parents' rocky relationship had on him; (2) that Wood suffered from an attachment disorder; (3) how Zjaiton took over parenting Wood; (4) how Wood did well in structured situations but relapsed to his old bad habits when he went home; (5) that Wood suffered from depression, PTSD, and anxiety; and (6) how Wood's biracial heritage affected his development. At both the evidentiary hearing and at trial, other witnesses' testimony touched on these same topics. Linda and Andre, for example, testified about Wood's childhood and how his parents' abusive relationship affected it. Many witnesses, moreover, testified about Zjaiton's influence on Wood, and Wood's success in structured environments. And while no one testified that Wood suffered from PTSD specifically, at trial Dr. Hand testified that Wood was paranoid and suffered from depression and anxiety. It was therefore not objectively unreasonable for the OCCA to find Dr. Allen's largely cumulative testimony would likely not have affected the proceeding's outcome. As the OCCA explained, Dr. Allen's report was only a "new assessment of previously presented and discovered evidence" that would not have affected the outcome of trial. App. at 654.
Wood's argument to the contrary misconstrues the prejudice inquiry. He focuses on how, had appellate counsel included this issue in the supplemental brief, "it is likely Wood would have prevailed on the issue ." Aplt. Br. at 48 (emphasis added). In other words, Wood argues the OCCA would have concluded that, under Oklahoma law, the evidentiary court erred by not permitting Dr. Allen to testify. But the prejudice analysis turns on whether the result of the entire "proceeding would have been different," which here is the OCCA's denial of relief on direct appeal because trial counsel did not perform ineffectively by failing to use the supplemental brief to challenge the exclusion of Dr. Allen's testimony. Cargle , 317 F.3d at 1202 (emphasis added). And for the reasons we discussed above, Wood cannot make this showing.
2. Failing to Obtain and Use Documents Substantiating the Abuse Allegations Against Gross
Wood next argues his appellate counsel was constitutionally ineffective at the Rule 3.11 hearing because he failed to "undercut [Gross's] minimization and denial of abuse with readily available records." Aplt. Supp. Br. at 8. More precisely, Wood insists his appellate counsel should have questioned Gross during his direct examination about the following documents: Gross's divorce decree, divorce records, protective orders issued against him, and his criminal history.14 Yet counsel only introduced one of those documents into the record at the evidentiary hearing-the divorce decree. By failing to discover and utilize these other documents, Wood contends his counsel performed ineffectively.
The OCCA denied relief because counsel did not perform deficiently and Wood suffered no prejudice. In so concluding, it emphasized that the divorce decree contained only allegations. Likewise, the protective *885orders offered "no proof that abuse occurred as they memorialize only [Linda's] allegations in support of her petition." App. at 659. The OCCA therefore "fail[ed] to see how the allegations of abuse contained in these documents would have bolstered [Linda's] testimony or changed the trial court's findings of fact or the outcome of the evidentiary hearing." Id. at 651 (emphasis added).
The OCCA never used the words "deficient performance" or "prejudice." But the substance of its concluding statement reaches both Strickland prongs. If the documents "would not have bolstered [Linda's] testimony," counsel could not have performed deficiently by failing to impeach Gross with them. Id. And if impeaching Gross with the documents would not have affected the direct appeal's outcome, then Wood suffered no prejudice from his counsel's failure to do so. The OCCA's holding therefore reached two distinct conclusions-one on deficient performance and one on prejudice.15
Wood argues AEDPA does not bar relief because the OCCA unreasonably applied Strickland in two ways.
He first claims the OCCA failed to, as Strickland requires, consider the cumulative prejudicial effect of his counsel's errors. But the OCCA had no need to consider cumulative prejudice because it only concluded appellate counsel performed deficiently in one instance. And in any event, the OCCA did consider cumulative prejudice-it was an inherent part of its prejudice analysis. We address this argument more fully below in Part III.B.4.
Nor would "all fairminded jurists agree" the OCCA unreasonably applied Strickland when it concluded that appellate counsel did not perform deficiently by failing to discover and use these documents, and that this failure did not prejudice Wood. We need not analyze the OCCA's conclusion on deficient performance because it did not unreasonably apply Strickland by concluding Wood suffered no prejudice.
Before analyzing this argument, it is helpful to consider how this case's procedural history makes proving prejudice difficult. Counsel's deficient performance prejudices a defendant if there is "a reasonable probability that, but for counsel's unprofessional error(s), the result of the proceeding would have been different." Cargle , 317 F.3d at 1202 (emphasis added). Ordinarily, proving prejudice is no easy task. Doing so is all the more difficult here, however, because the relevant "proceeding" is Wood's direct appeal of his conviction and sentence. To conclude Wood was prejudiced by his appellate counsel's failure to use these documents at the evidentiary hearing , then, the OCCA had to conclude there was a reasonable probability that, had counsel done so, the result of Wood's entire direct appeal would have been different. That is, questioning Gross about these documents at the evidentiary hearing would have likely caused the OCCA on direct appeal to conclude Wood's trial counsel performed ineffectively.
It is therefore no exaggeration to say that these documents would need to have a talismanic quality-or, put differently, be *886nearly outcome-determinative on the trial-counsel-ineffectiveness claim-in order for Wood to have suffered prejudice from his counsel's failure to obtain and use them. After all, in rejecting Wood's trial-counsel-ineffectivess claim, the OCCA considered the avalanche of evidence presented at the evidentiary hearing-testimony from twenty-three witnesses and thousands of pages documents. Thus, if these documents would have likely changed the OCCA's thorough conclusion considering all this other evidence, they would have to be extraordinarily powerful.
The documents fall below this high bar. We agree that all these documents cast doubt on Gross's minimizations of abuse. But other testimony at both the evidentiary hearing and Wood's sentencing proceeding served this same purpose. That is, Gross's minimization of the abuse allegations during his testimony at the evidentiary hearing was already undercut by other testimony at both the evidentiary hearing and at sentencing. Simply stated, these documents did not contain a newfound smoking gun-rather, they were filled with the same abuse allegations against Gross many others had already made.
For one thing, introducing these documents would have been cumulative of Andre Wood's and Linda Wood's testimony at the Rule 3.11 hearing . There, both of them directly countered Gross's minimizations of abuse. Andre recalled Gross knocking Linda's front teeth out and seeing his "dad sitting there slapping [his] mom around or punching her in the face." Tr., 2/23/06, at 159. He often thought Gross "was going to kill" his mother. Id. And he described how his mother withstood "beatings that most grown men couldn't walk away from." Id. Likewise, at the evidentiary hearing Linda recalled the litany of times Gross abused her. She described seeking help from a battered women's shelter four or five times due to Gross's abuse. And she cited the numerous scars Gross left her with. She has false teeth because Gross "hit [her] in the mouth." Id. at 142. A scar on her head came from Gross whacking her "with the butt of a gun." Id. at 143. And Gross once broke her nose by "slam[ming] it into the hood of a car." Id.
Further, asking Gross about the allegations in these documents would also have been cumulative of Linda's and Dr. Hand's testimony at the sentencing itself . During the sentencing phase, Linda told the jury how she and Gross had "a very abusive relationship." Tr., 04/05/2004, at 91. Indeed, she recounted being "beaten many, many times in front of my children. Tied up. Dragged down the highway. My bones broke." Id. And Dr. Hand testified that he "reviewed a file of the Department of Human Services' records that include[d] referrals for abuse and neglect that go back to when" Wood was five years old. Id. at 43.
Thus, introducing these documents would not have broken new ground. To be sure, the documents Wood cites support the charges of abuse against Gross. Linda's petition for a protective order, for example, alleged that her "husband hit me [four] times with his fist in my face." PCA Ex. 9A. And Gross's criminal history revealed he had been charged with pointing a gun at Linda to threaten and intimidate her. PCA Ex. 12A. The allegations in these documents thus sprinkle extra credibility on Andre and Linda's testimony. But they are not substantially more powerful than the testimony actually presented at both the Rule 3.11 hearing and the sentencing. Accordingly, the OCCA did not unreasonably apply Strickland when it held Wood was not prejudiced by appellate counsel's failure to obtain and use these documents.
*887Wood also claims the OCCA unreasonably determined the facts by "ignor[ing] the independent corroboration of abuse contained in Gross's criminal records." Aplt. Sup. Br. at 10. We fail to understand how ignoring evidence can be considered a factual determination. But in any event, the OCCA based its denial of relief on the "documents Wood provide[d]." App. at 659 (emphasis added). The OCCA therefore expressly considered all the documents Wood pointed it to, including Gross's criminal records. True, the OCCA specifically referenced some of the documents, such as the divorce petition, but did not mention the criminal records. This does not matter. "The Supreme Court has never required state courts to be verbose for AEDPA purposes." Curiel v. Miller , 830 F.3d 864, 870 (9th Cir. 2016).
Finally, Wood argues that, by describing the allegations of abuse in the divorce records and restraining order as "practically meaningless," the OCCA "failed to consider whether a reasonable probability existed that their use by appellate counsel might have changed the trial court's factual findings." Aplt. Sup. Br. at 10. We agree the OCCA did not consider the effect impeaching Gross would have had on the evidentiary hearing's factual findings. But this was because the OCCA's prejudice inquiry rightly turned on what effect impeaching Gross would have had on the OCCA's decision on direct appeal, not on factual findings at an evidentiary hearing.
To conclude, we emphasize that the OCCA on Wood's post-conviction appeal had these documents detailing Gross's abuse before it, along with all the other testimony about Gross's abuse that was developed at the evidentiary hearing and sentencing. Viewing the documents in this context, the OCCA stated they were "insufficient to convince us that appellate counsel was ineffective." App. at 659. Having closely reviewed these documents and the record from the evidentiary hearing ourselves, we cannot conclude the OCCA's conclusion was unreasonable under § 2254(d).
3. Failure to Update the Record with Mr. Albert's Suspension and Disciplinary Proceedings
Third, Wood argues his appellate counsel performed ineffectively by failing to update the record with his trial counsel's subsequent disciplinary proceedings. These proceedings flowed from Mr. Albert's abuse of alcohol and cocaine, which began in March and April of 2005, about a year after Wood's trial and sentencing, but about a year before the Rule 3.11 evidentiary hearing. State ex rel. Okla. Bar Ass'n v. Albert , 2007 OK 31, 163 P.3d 527, 539. Just days after the Rule 3.11 hearing, at which Mr. Albert testified, an Oklahoma County court held him in contempt for failing to enter a substance-abuse treatment program. During those proceedings, the court repeatedly admonished Mr. Albert for lacking candor. Eventually, the Oklahoma Supreme Court suspended Mr. Albert from the practice of law.
The OCCA concluded appellate counsel's failure to update the record did not render his performance ineffective. In its view, "evidence of trial counsel's involvement in a contempt proceeding in unrelated cases and his suspension from the practice of law two years after Wood's conviction does not prove that trial counsel was ineffective during Wood's trial." App. at 656.
Again, Wood argues AEDPA's bar on relief does not apply for a number of reasons. None has merit.
First , he contends the OCCA unreasonably applied Strickland because it failed to consider the cumulative prejudicial effect of the many alleged errors. But the OCCA did not need to consider cumulative prejudice *888because it found no deficient performance. And the OCCA did, in fact, consider cumulative prejudice, as we explain in Part III.B.4.
Second , he argues the OCCA unreasonably applied Strickland when it concluded Wood was not prejudiced by his counsel's failure to update the record. He emphasizes that Mr. Albert's contempt proceeding "bore directly upon [Albert's] truthfulness and occurred within two weeks of when he testified at Wood's Rule 3.11 hearing." Aplt. Supp. Br. at 12. As a consequence, had the OCCA known about these disciplinary actions, Wood argues it would have doubted Mr. Albert's testimony because (1) he may have been abusing drugs and alcohol when he testified at the evidentiary hearing, and (2) he likely lacked candor during the hearing, just as he had in the disciplinary proceedings. Yet the OCCA concluded this evidence "would have made no difference on the outcome of appeal." App. at 656. This, Wood insists, was an unreasonable application of Strickland .
We disagree. The OCCA reasonably applied Strickland when it concluded Wood endured no prejudice from his counsel's failure to update the record with Albert's suspension. The reason for this is simple: Albert's testimony and affidavit helped Wood's argument that Albert performed deficiently, so undercutting Albert's credibility by introducing his subsequent disciplinary troubles could have hurt Wood's case.
At the evidentiary hearing, appellate counsel introduced an affidavit Mr. Albert signed in which he largely fell on his sword. In it, Mr. Albert admitted he "did not review the records properly in order to present a meaningful mitigation case, and that due to my lack of proper review, in hindsight, I see that I was not effective." EH Vol. 1 Ex. M. And Mr. Albert similarly conceded that he "relied upon Tremane Wood to properly explain and develop the mitigation portion of the defense, and in hindsight, I see that I could have achieved a better and effective result for Tremane Wood had I been more involved." Id.
When he testified at the evidentiary hearing, Mr. Albert affirmed the admissions in his affidavit. He acknowledged he "could have achieved a better" result for Wood had he been more prepared at sentencing. Tr., 2/27/06, at 248. And Mr. Albert explained that he had "too many cases" while representing Wood, and he had consequently "decided to quit doing death penalties" after Wood's case. Id. at 247. All in all, Mr. Albert acknowledged he "could have done better" in the case. Id. at 251.
So Mr. Albert's affidavit and testimony largely strengthened Wood's argument that Albert performed ineffectively at trial. Undercutting Mr. Albert's credibility by bringing up these disciplinary issues could have thus washed away his helpful testimony and hurt Wood's case. Given this, the OCCA's conclusion Wood suffered no prejudice was reasonable.
We recognize that Mr. Albert's upcoming disciplinary proceedings could have influenced his testimony. That is, Mr. Albert could have been defensive during the evidentiary hearing and held back even more helpful testimony because he feared it would be used as evidence in the disciplinary case against him. Had counsel let the OCCA know about these disciplinary proceedings, Wood asserts the OCCA would have presumed Mr. Albert was downplaying his ineffectiveness and, in turn, would have been more likely to conclude he performed ineffectively.
We agree this chain of events is theoretically possible. But we cannot conclude the OCCA unreasonably applied Strickland by concluding this scenario was not reasonably probable. When determining if *889§ 2254(d) applies and bars relief, our review is "limited to the record that was before the state court that adjudicated the claim on the merits." Cullen , 563 U.S. at 182, 131 S.Ct. 1388.16 And the fact of the matter is, Wood presented no evidence during the post-conviction proceeding that Mr. Albert was defensive or not fully forthcoming at the evidentiary hearing because he was worried about an adverse finding affecting his disciplinary proceeding. Without such evidence, it was not objectively unreasonable for the OCCA to conclude this scenario was not reasonably probable.
Third , Wood points to allegations that Mr. Albert had begun "consuming large amounts of alcohol during the work day in early 2004, when he was representing Wood in the penalty phase." Aplt. Sup. Br. at 13 (citing PC Not. of Suppl. Auth., Att. ¶¶ 3, 6 11/19/07). Surely, says Wood, his appellate counsel performed ineffectively by failing to alert the OCCA-which was considering whether Mr. Albert performed ineffectively at trial-of these allegations that Mr. Albert was abusing alcohol around the time of Wood's trial and sentencing. The OCCA's conclusion to the contrary, he says, unreasonably applied Strickland .
We disagree. To be sure, evidence of alcohol abuse can be, and often is, strong evidence of attorney misconduct. But the evidence Wood cites is not particularly powerful. Wood cites factual findings from another disciplinary proceeding in which various individuals claimed that in January 2004, Mr. Albert began drinking beers in the afternoon on some workdays and, in turn, missed some afternoon court appearances.
Critically, though, none of this testimony is connected in any way to Wood's trial. Indeed, no one claims Mr. Albert drank alcohol before meeting with Wood, was intoxicated during Wood's trial, or that alcohol interfered in any way with Mr. Albert's representation of Wood. And strikingly, despite the fact that one of the main charges in the document is that Mr. Albert missed court appearances because of his drinking, Wood cites no evidence Mr. Albert missed any court appearances in his case. Simply put, these allegations of alcohol abuse, while troubling, lack any link to Wood's case.
And on direct appeal, moreover, Wood's theory of the case did not turn on, or indeed even relate to, Mr. Albert performing ineffectively because he abused alcohol. Rather, Wood argued that for whatever reason-alcohol or otherwise-Mr. Albert failed to present a thorough mitigation defense by calling additional witnesses. The OCCA disagreed, and concluded Mr. Albert did not perform deficiently. And even if he had, it held Wood suffered no prejudice because the extra evidence Wood could have offered would not have affected the proceeding's outcome. We do not see why allegations of alcohol abuse would have affected this conclusion.
In sum, allegations of alcohol abuse are a serious charge. But the charge is militated if it has no reasonable relationship to the defendant's case. Because these allegations lack any connection to Mr. Albert's representation of Wood, the OCCA's conclusion no prejudice flowed from counsel's failure to update the record with this information is not an unreasonable application of Strickland .
Fourth , Wood claims the OCCA contravened Strickland by applying the wrong prejudice standard. In concluding Wood suffered no prejudice, the OCCA
*890said "Wood [could] not show that the outcome of his appeal would have been different." App. at 656. But, Wood protests, Strickland defines the prejudice inquiry differently-as whether there is "a reasonably probability" the proceeding's outcome would have been different, Strickland , 466 U.S. at 694, 104 S.Ct. 2052, not whether it "would have been," App. at 656. The OCCA therefore applied the wrong prejudice standard, says Wood, rendering its decision contrary to Strickland .
Wood admits he never raised this argument below and thus forfeited it, but he asks us to use our discretion to consider it. But even if we exercised our discretion to consider the argument, it lacks merit. Elsewhere in its opinion, the OCCA recited the correct prejudice standard. App. at 655 ("Under [ Strickland ], Wood must not only overcome the presumption of competence but show that there is a reasonable probability that, but for counsel's unprofessional error, the result of the proceeding would have been different."). And under AEDPA's deferential framework, we "presum[e] that state courts know and follow the law." Woodford v. Visciotti , 537 U.S. 19, 24, 123 S.Ct. 357, 154 L.Ed.2d 279 (2002) (per curiam). We therefore do not think the OCCA's occasional use of shorthand to describe the prejudice standard renders its decision contrary to Strickland . See id.
4. Failure to Challenge Factual Findings in the Supplemental Brief
Next, Wood argues his appellate counsel performed ineffectively by failing to challenge the trial court's factual errors at the evidentiary hearing in his supplemental brief. He advances two arguments as to why § 2254(d) does not prevent us from granting relief.
He first contends the OCCA unreasonably applied Strickland because it failed to consider the cumulative prejudicial effect of his counsel's alleged errors.17 In short, he points to Strickland 's requirement that courts "consider the totality of counsel's errors in assessing whether a defendant was thereby prejudiced." Aplt. Supp. Br. at 5. Yet he claims the OCCA only considered the prejudicial effect of each alleged error in isolation. Put differently, in Wood's view the OCCA considered whether each alleged error, standing alone, prejudiced him. This unreasonably applied Strickland , he says, because the OCCA had to consider whether the combined effect of these errors prejudiced Wood.
But courts need only consider the cumulative prejudicial effect of counsel's alleged errors if they first conclude counsel performed deficiently in numerous ways. See Ellis v. Raemisch , 872 F.3d 1064, 1090 (10th Cir. 2017). So if the OCCA found "only one possible instance of deficient performance," then it had no need to consider cumulative prejudice. Id.
Here, the OCCA did not need to consider cumulative prejudice because it concluded appellate counsel did not perform *891deficiently. On the claim related to counsel's failure to challenge Dr. Allen's exclusion in the supplemental brief, the OCCA concluded counsel did not perform deficiently. See App. at 661 (noting that appellate counsel employed a "reasonable strategy" in selecting what, and what not, to raise in the supplemental brief). Similarly, on the claim about counsel's failure to obtain and use documents substantiating Gross's abuse, the OCCA concluded counsel did not perform deficiently and Wood also suffered no prejudice. And on the claim relating to Mr. Albert's disciplinary records and alcohol use, the OCCA likewise found nothing to show counsel was "ineffective during Wood's trial." App. at 652. Finally, on the claim that counsel failed to challenge the trial court's factual findings, the OCCA held counsel did not perform deficiently. The OCCA therefore did not need to consider cumulative error. Ellis , 872 F.3d at 1090. Its decision consequently cannot be contrary to, or an unreasonable application of, Strickland for failing to engage in this analysis.
Even if the OCCA did need to consider cumulative prejudice, Wood would still not be entitled to relief. At least in the AEDPA context, considering the cumulative prejudicial effect of counsel's numerous errors is an inherent part of the prejudice inquiry. By Strickland 's own terms, concluding the petitioner suffered no prejudice involves considering the totality of the circumstances, including whether "counsel's unprofessional errors" likely affected the proceeding's outcome. Strickland , 466 U.S. at 703, 104 S.Ct. 2052 (emphasis added). And when we analyze state-court decisions through AEDPA's deferential lens, we must "presum[e] that state courts know and follow the law" and give "state-court decisions ... the benefit of the doubt." Woodford , 537 U.S. at 24, 123 S.Ct. 357. As a result, where, as here, a state court analyzes numerous errors separately and concludes each one did not prejudice the defendant, we presume this analysis considers the prejudicial impact of all the alleged errors together. Put differently, giving the state-court decision the benefit of the doubt requires assuming that when the court says numerous alleged errors did not prejudice the defendant, it has considered both the individual and cumulative prejudicial effect of each alleged error.
At bottom, Wood asks us to impose an opinion-writing requirement on state courts. Under his reasoning, when a state court concludes multiple alleged errors did not prejudice the defendant, its opinion must have a separate section or sentence explicitly stating that, cumulatively, these errors were not prejudicial. If it does not, the court unreasonably applied Strickland . This rule is at war with the Supreme Court's constant refrain that AEDPA does not empower federal courts to "impose mandatory opinion-writing standards on state courts." Johnson v. Williams , 568 U.S. 289, 300, 133 S.Ct. 1088, 185 L.Ed.2d 105 (2013) ; see also Lafler v. Cooper , 566 U.S. 156, 183, 132 S.Ct. 1376, 182 L.Ed.2d 398 (2012) (Scalia, J., dissenting) (noting that while a "state court's analysis was admittedly not a model of clarity ... federal habeas corpus is a 'guard against extreme malfunctions in the state criminal justice system,' not a license to penalize a state court for its opinion-writing technique") (quoting Richter , 562 U.S. at 102, 131 S.Ct. 770 ); cf. Richter , 562 U.S. at 99, 131 S.Ct. 770 ("Opinion-writing practices in state courts are influenced by considerations other than avoiding scrutiny by collateral attack in federal court.").18
*892We accordingly give the OCCA's thorough decision the benefit of the doubt and conclude that when it decided none of these four errors caused Wood any prejudice, it considered both the individual and cumulative effect of the alleged errors.
Wood offers a second possible reason why AEDPA does not preclude us from granting relief: the OCCA unreasonably applied Strickland by concluding appellate counsel acted strategically by failing to correct the evidentiary hearing court's factual findings in the supplemental brief. In support, he cites Cargle v. Mullin , 317 F.3d 1196 (10th Cir. 2003). That case held that to determine if appellate counsel performed ineffectively by failing to raise issues in a brief, courts should "look to the merits of the omitted issue." Id. at 1202. The OCCA, Wood contends, never considered the merits of the issues his appellate counsel omitted from the supplemental brief, rendering its analysis unreasonable. We agree the OCCA never explicitly considered the merits of these issues. But this does not entitle Wood to relief. AEDPA does not empower us to "impose mandatory opinion-writing standards on state courts." Johnson , 568 U.S. at 300, 133 S.Ct. 1088. And we presume that state courts know and follow the law. See Woodford , 537 U.S. at 24, 123 S.Ct. 357. With this presumption, we conclude the OCCA did not unreasonably apply Strickland by failing to explicitly consider the merits of the omitted issues when it concluded appellate counsel's failure to include these issues in the supplemental brief did not render his performance deficient.19
5. Failure to Object to Juror Separation
Wood claims his appellate counsel performed ineffectively in a fifth and final way: by failing to raise a claim that his trial counsel performed ineffectively by failing to object to the jurors leaving the courtroom after the court charged them and before they began deliberations.
At both the guilt and sentencing phase of trial, after the court charged the jurors, it permitted them to leave the courtroom to move their cars. Wood claims this violates Oklahoma Statute Title 22 § 857, which requires a court officer to keep the jurors together after the court charges them.20 If trial counsel objects to jurors separating in violation of § 857, the error is presumed to prejudice the rights of the defendant, "and the burden falls to the [s]tate to prove otherwise."
*893Johnson v. State , 2004 OK CR 23, 93 P.3d 41, 47. If the state fails to prove "no prejudicial injury resulted" from the separation, this "vitiates the verdict." Page v. State , 332 P.2d 693, 695 (Okla. Crim. App. 1958). But if trial counsel failed to object to the separation, this "waives any potential error." Elliott v. State , 753 P.2d 920, 922-23 (Okla. Crim. App. 1988).
Wood's trial counsel failed to object to the jurors separating on both occasions. Yet on direct appeal, Wood's appellate counsel did not argue trial counsel performed ineffectively for this reason. Wood claims his appellate counsel's failure to do so rendered his performance constitutionally ineffective.
Even on de novo review, Wood is not entitled to relief. "When considering a claim of ineffective assistance of appellate counsel for failure to raise an issue, we look to the merits of the omitted issue." Neill v. Gibson , 278 F.3d 1044, 1057 (10th Cir. 2001). The omitted issue's merits determine both deficient performance and prejudice. Cargle , 317 F.3d at 1202. To rule on Wood's ineffective assistance of appellate counsel claim, then, we must assess the merits of the ineffective assistance of trial counsel claim his counsel failed to raise.
We conclude the trial-counsel-based claim lacks merit, so the appellate-counsel-based claim fails. In Warner v. State , 2006 OK CR 40, 144 P.3d 838, 875, jurors separated to move their cars after deliberation had begun. The OCCA held this "infrequent and short" separation "did not constitute a separation of the jury during deliberations within the meaning of § 857." 144 P.3d at 875. Indeed, the OCCA recognized that jurors briefly separating to move their cars "was a common occurrence for juries in Oklahoma County that deliberated into the evening." Id. Warner therefore demonstrates that under Oklahoma law, jurors separating to move their cars does not qualify as a separation under § 857. Wood's trial counsel consequently did not perform ineffectively for failing to object to the jurors separating to move their vehicles, since no violation of § 857 had occurred. In turn, Wood's appellate counsel did not perform ineffectively by failing to raise this non-meritorious claim.
* * *
In sum, Wood is not entitled to relief on the claim his appellate counsel performed ineffectively on direct appeal.21
C. The HAC Aggravator
In light of Pavatt v. Royal , 859 F.3d 920 (10th Cir. 2017),22 we granted a COA on whether constitutionally sufficient evidence was presented to the jury to prove beyond a reasonable doubt the capital aggravating circumstance that murder of Mr. Wipf was especially heinous, atrocious, or cruel. As an initial matter, Oklahoma argues we should deny relief because Wood failed to exhaust this argument, forfeited it, abandoned it, and § 2254(d) limits our ability to grant relief. We need not address these arguments, however, because even assuming this issue is properly presented to us *894and we could consider the issue under de novo review, we deny relief.
To understand Wood's claim based on Pavatt , some background is necessary. We first explain the two ways in which a petitioner can challenge the evidence supporting the imposition of an aggravator. We then turn to Pavatt and apply its approach to the facts of this case.23
1. Background Law
Under Oklahoma law, a jury may only impose the death penalty when it unanimously finds at least one statutory aggravating circumstance beyond a reasonable doubt, and also concludes the mitigating circumstances do not outweigh the aggravating circumstances. See Ross v. Ward , 165 F.3d 793, 799 (10th Cir. 1999). When a jury finds an aggravating circumstance exists, capital defendants can challenge the aggravator in two ways.
First, a defendant can bring a sufficiency of the evidence claim under Jackson v. Virginia 443 U.S. 307, 316, 99 S.Ct. 2781, 61 L.Ed.2d 560 (1979). It violates the Fourteenth Amendment's guarantee of due process if a jury sentences a defendant to death based on an aggravator, even though there was insufficient evidence for any rational juror to have concluded the aggravator was met. Pavatt , at 1124-26. Because state law defines aggravators, this question turns on state law. Id.
Second, petitioners can challenge an aggravating circumstance as unconstitutionally vague. It violates the Eighth and Fourteenth Amendments for death sentences to be arbitrarily imposed. See Lewis v. Jeffers , 497 U.S. 764, 774, 110 S.Ct. 3092, 111 L.Ed.2d 606 (1990). As a consequence, if an aggravating circumstance is so vague it could apply to any and every murder, then sentencing a defendant to death because that aggravator was met violates the Constitution. See, e.g. , Godfrey v. Georgia , 446 U.S. 420, 427-28, 100 S.Ct. 1759, 64 L.Ed.2d 398 (1980).
The Supreme Court applied this reasoning in Maynard v. Cartwright when it held the same Oklahoma HAC aggravating circumstance at issue here was unconstitutionally vague. 486 U.S. 356, 359-61, 108 S.Ct. 1853, 100 L.Ed.2d 372 (1988). Even so, after Maynard our circuit has repeatedly upheld death sentences based on the HAC aggravator. We do so when the aggravator's vagueness has been "cure[d]" by "a narrowing construction." Hatch v. State of Okla. , 58 F.3d 1447, 1468 (10th Cir. 1995), overruled on other grounds by Daniels v. United States , 254 F.3d 1180, 1188 n.1 (10th Cir. 2001) (en banc). As we previously explained, Oklahoma courts have narrowed the HAC aggravator so it could only apply if "the murder involve[d] 'torture of the victim or serious physical abuse.' " Hatch , 58 F.3d at 1468 (quoting Stouffer , 742 P.2d at 563 ). And a victim only suffers serious physical *895abuse if he or she endures "conscious physical suffering" before death. Medlock v. Ward , 200 F.3d 1314, 1321 (10th Cir. 2000) (quoting Cheney v. State , 909 P.2d 74, 80 (Okla. Crim. App. 1995) ).
In Pavatt , we nonetheless held the OCCA unconstitutionally construed the HAC aggravator in that case. Pavatt explained that our prior acceptance of that aggravator, when it was narrowly construed, did not "immunize [the OCCA's] decision[ ] from review of whether it ha[d] departed from that acceptable construction." Pavatt , at 1131. "[E]ven when a State has previously applied a constitutionally valid narrowing construction of an aggravator," Pavatt concluded, a "death penalty imposed under the aggravator must still be based on a construction that in a 'principled way' can distinguish the case from the many in which the penalty was not imposed." Id. at 1131 (quoting Godfrey , 446 U.S. at 433, 100 S.Ct. 1759 ).
Pavatt then asked whether, by applying the HAC aggravating circumstance to the facts of that case, the OCCA construed the aggravator in a manner that complied with the Constitution. Id. at 1130-33. It held the OCCA had not. Id. at 1131-33. That was because the OCCA failed to construe the aggravator "so that it distinguishes in a principled way" the differences between "crimes deserving death and the many cases in which the death penalty is not imposed." Id. In other words, Pavatt held that if the HAC aggravator could apply to that case, then it could apply to "[v]irtually any murder in which the victim did not die instantly," "the act of murder did not immediately render the victim unconscious[,] and the wounds could have caused pain." Id. Thus, Pavatt held that by applying the HAC aggravator to the facts of that case, the OCCA "did not apply a constitutionally acceptable interpretation of Oklahoma's HAC aggravator." Id.
2. Application of Pavatt to the Facts of this Case
Under Pavatt , Wood is not entitled to relief. Wood advances two arguments. First, he argues that, like the victim in Pavatt , Mr. Wipf endured only the suffering that necessarily accompanies every murder. Wood accordingly contends the OCCA contravened the Constitution by construing the HAC aggravator so that it could apply to the facts of this case. Second, even if the evidence establishes that Mr. Wipf suffered, Wood argues no evidence shows he-as opposed to Zjaiton-inflicted that suffering.
Both arguments fail. To understand why, it is helpful to first summarize the trial testimony detailing Mr. Wipf's death and the altercation that occurred before it. Three pieces of evidence are key.
First , Kleinsasser, Mr. Wipf's friend who was in the hotel room, testified that he witnessed a fight between Wood, Zjaiton, and Wipf, during which Mr. Wipf became covered in blood and screamed in pain. Kleinsasser testified that two men "burst[ ] in the door," one with a gun and one with a knife. Tr., 3/31/04 at 131. Other evidence proved Wood had the knife and Zjaiton the gun. "[A]s soon as they opened the door," Kleinsasser recalled, Mr. Wipf "started struggling with" Wood. Id. at 135. Wood and Wipf "were struggling and fighting," and Wipf was "screaming in pain or terror." Id. at 136. Meanwhile, Zjaiton pointed the gun at Kleinsasser and demanded the money. Id. at 133. Eventually, Zjaiton joined Wood in fighting Mr. Wipf, and the three men moved towards the bathroom.
Kleinsasser testified that Wood then walked away from Mr. Wipf and towards him. Wood thumped Kleinsasser on the head with the handle of the knife and *896demanded more money. When Kleinsasser said he had no more, Wood walked back over to Mr. Wipf.
As Wood, Zjaiton, and Mr. Wipf exited the bathroom, Kleinsasser saw Mr. Wipf's whole body "covered in blood." Id. at 139. Before he ran out of the room, Kleinsasser saw Wood, Zjaiton, and Wipf "fighting amongst themselves" in a "big fight" just in front of the bed. Id. at 140-41. Mr. Wipf continued to scream as Kleinsasser ran out of the room.
Second , the state's autopsy report and photographs of Mr. Wipf's body reveal he had many cuts and bruises besides the fatal stab wound. The autopsy report shows Mr. Wipf had a significant number of cuts on his right eye, chin, and on his right hand. The state's photographs confirm this in gruesome detail.
Third , the medical examiner, Dr. Jordan, reviewed the autopsy drawing of the injuries and the photographs of Mr. Wipf's body. He explained that the injuries around Wipf's eyes were "very fresh" contusions and abrasions. Tr., 4/2/2004 at 8. And the cuts on Wipf's hand, he said, were consistent with "defensive wounds"-which, the examiner agreed, are "something like if you are about to be struck with a knife and hold your hand up to defend yourself." Id. at 15. On cross examination, however, Dr. Jordan conceded the cuts on Wipf's face could also have been caused by him falling down and not bracing himself. On redirect, Dr. Jordan again iterated that Mr. Wipf's non-lethal injuries "[c]ould have been caused by being in a fight." Id. at 19-20.
Wood first argues this evidence demonstrates Mr. Wipf's death involved only the amount of suffering that necessarily accompanies every murder. We disagree. Sufficient evidence exists that prior to his death, Mr. Wipf endured far more suffering than every murder victim experiences. An eyewitness testified that Mr. Wipf was engaged in a fistfight with two men. And the photographs reveal that Wipf suffered serious injuries during the fight-the Wood brothers badly beat and bruised his face, and his hand had deep cuts on it. All this evidence demonstrates that before Mr. Wipf died, he endured serious pain and suffering. This is unsurprising since he was involved in a two-on-one fistfight in which both of his opponents were armed. Contrary to Wood's argument, then, we conclude Mr. Wipf's death did not involve merely the kind of suffering that necessarily accompanies every murder. Accordingly, applying the aggravator to these facts does not mean the aggravator could apply to every murder.
Wood focuses on the fact that no one saw Wood stab Mr. Wipf. He therefore theorizes that Mr. Wipf was stabbed after Kleinsasser left the room and died instantaneously. In his view, this makes his case indistinguishable from Pavatt . In that case, the evidence did not show it probable that the victim survived long enough after he was shot with a shotgun to suffer. Pavatt , at 1130-31. So too here, says Wood, no evidence demonstrates Mr. Wipf survived after being stabbed; he may well have died instantly like the victim in Pavatt .
Even if we assume that Mr. Wipf was, in fact, stabbed after Kleinsasser fled the room and died instantaneously, Mr. Wipf endured conscious physical suffering before being stabbed. Indeed, Kleinsasser directly testified to the brutal beating Mr. Wipf endured from the Wood brothers. And although the medical examiner testified that the cuts on Wood's face could have been by caused by Mr. Wipf falling down after being stabbed, the examiner gave no such testimony about the deep cut on his hand. Thus, that Mr. Wipf endured conscious *897physical suffering before the fatal wound was inflicted distinguishes this case from Pavatt .24
Wood cites a number of cases which, he contends, demonstrate the beating Mr. Wipf endured before his death does not constitute serious physical abuse. None of the cases, however, persuade us.
In Stouffer v. State , for example, the OCCA held sufficient evidence did not support the HAC aggravator because there was "no reason to believe from the evidence that [the victim] was conscious after" she was shot. 742 P.2d 562, 564 (Ok. Cr. App. 1987). In other words, in that case there was no evidence the victim endured serious physical abuse because she died instantaneously. Here, in contrast, even assuming Mr. Wipf immediately died from his stab wound, he nonetheless endured serious physical abuse before his death when Wood and Zjaiton brutally beat him.
Two other cases Wood relies on, Cudjo v. State , 925 P.2d 895 (Okla. Crim. App. 1996), and Hawkins v. State , 891 P.2d 586 (Okla. Crim. App.), are equally inapplicable. In both cases, the OCCA concluded insufficient evidence supported the application of the HAC aggravator because there was no evidence the victim experienced any suffering "beyond the scope of the act of killing itself." Cudjo , 925 P.2d at 901-02 (emphasis added); see also Hawkins , 891 P.2d at 596-97 ("No evidence of serious physical abuse, that is, gratuitous violence inflicted on the victim beyond the act of killing , is present in this case." (emphasis added) ). But in this case, Mr. Wipf did not only suffer when the knife was fatally thrust into his chest. To the contrary, before *898his killing he was brutally beaten by two men. Accordingly, Cudjo and Hawkins are not on point.
Wood's second argument likewise fails. He argues that even if Mr. Wipf experienced conscious physical suffering prior to his death, nothing demonstrates he, rather than Zjaiton, caused this suffering. This argument ignores Kleinsasser's testimony. Indeed, Kleinsasser testified that Wood and Wipf immediately began fighting when the Wood brothers entered the room. And Kleinsasser further detailed how Wood, sometimes accompanied by Zjaiton, was fighting Mr. Wipf while Wipf screamed in pain. This eyewitness testimony establishes that Wood, perhaps with Zjaiton at times, caused Wipf to endure conscious physical suffering before his death.
In sum, applying Pavatt 's approach does not entitle Wood to relief.25
IV. Conclusion
Wood is not entitled to relief on any of his claims. We therefore AFFIRM the district court's denial of habeas relief.