In this habeas corpus case, Carlos Cuesta-Rodriguez challenges his Oklahoma conviction for first-degree murder and his accompanying sentence of death. The district court denied relief and denied a certificate of appealability (COA). We granted a COA, agreeing to hear a number of Cuesta-Rodriguez's claims. Exercising jurisdiction under 28 U.S.C. § 2253(a), we agree *890with the district court and conclude that Cuesta-Rodriguez isn't entitled to relief.

BACKGROUND

I. The Crime of Conviction

The following facts come from the direct-appeal decision of the Oklahoma Court of Criminal Appeals (OCCA), Cuesta-Rodriguez v. State , 241 P.3d 214 (Okla. Crim. App. 2010). We presume that the OCCA's factual findings are correct. See 28 U.S.C. § 2254(e)(1) (establishing that state-court determinations of fact "shall be presumed to be correct" unless rebutted by "clear and convincing evidence").

Olimpia Fisher-the victim-and her adult daughter, Katya Chacon, lived with Cuesta-Rodriguez in a home Fisher and Cuesta-Rodriguez had purchased together. In the year following the home purchase, Cuesta-Rodriguez and Fisher's relationship was strained. Fisher was working long hours as a moving-company packer, and Cuesta-Rodriguez feared she was cheating on him. Whenever Fisher and Chacon would leave the house, Cuesta-Rodriguez would question them "about where they were going and what they would be doing." Cuesta-Rodriguez , 241 P.3d at 222. The relationship deteriorated to the point that both Cuesta-Rodriguez and Fisher wanted the other to move out.

On May 20, 2003, Fisher went to the local police station "to make a complaint of domestic abuse." Id. The interviewing officer "observed bruising on her right upper arm and stomach." Id. But when Fisher realized that the officer "was going to take photographs of the bruising and that Cuesta-Rodriguez would be arrested, she ran out of the station." Id.

On May 31, 2003, Cuesta-Rodriguez called Fisher on her cell phone. She answered and replied that she was at work. But Cuesta-Rodriguez had gone by her place of work earlier and knew she wasn't there. "Believing she was cheating on him, he went home, drank some tequila, and went to bed." Id.

Around 10 p.m., Chacon came home to a dark house. She saw an empty bottle of tequila1 with a note beside it. The note, written on the back of an envelope, read, "fuck you bitches and puntas, goodbye." Trial Tr. vol. II at 381:2. After realizing Cuesta-Rodriguez was home, Chacon attempted to contact her mother. Unable to reach her by phone, Chacon left the house and joined Fisher as she was getting off work. The two ate a late dinner at McDonald's *891and went home. Though they initially planned to pack and leave that night, they decided to stay overnight, Chacon sleeping in her own bedroom and Fisher sleeping in a third bedroom.

Around 4:30 a.m., Chacon awoke to the sounds of Fisher and Cuesta-Rodriguez arguing. She went to the bedroom where the two were fighting and persuaded Fisher to come back to her (Chacon's) bedroom "in the hope that Cuesta-Rodriguez would leave them alone." Cuesta-Rodriguez , 241 P.3d at 222. But "Cuesta-Rodriguez followed the women into [Chacon's] bedroom while continuing to argue loudly with Fisher." Id.

Fisher picked up a phone, but Cuesta-Rodriguez grabbed it and tossed it from her reach. At the same time, he pulled out a pistol "and blasted Fisher in the right eye."2 Id. Chacon "retrieved a baseball bat from under the bed and tried to hit Cuesta-Rodriguez in the hand." Id. He "grabbed the bat as [she] swung it and threw it to the floor." Id. Chacon ran from the building and called 911 from a neighbor's house.

After being shot, Fisher was still conscious. Cuesta-Rodriguez "took her to his bedroom where, despite having an eye blown out, Fisher continued to fight and struggle." Id. at 223. Around 4:41 a.m., the first police officers arrived on the scene (within two minutes of being dispatched by 911). Officers approached the house and heard Fisher "screaming and banging on a bedroom window as if she was trying to escape." Id. The house's windows and doors "were covered with burglar bars that not only prevented her escape, but also prevented entry by police." Id. The officers attempted to enter by "kicking in the front door," but that failed. Id. While attempting to enter the building, the officers heard a gunshot-and then Fisher's screams stopped. An autopsy later revealed a second, fatal gunshot wound to Fisher's left eye.

Certain that Fisher was dead and "that Cuesta-Rodriguez was armed, police summoned their tactical team." Id. Meanwhile, a police hostage negotiator attempted to convince Cuesta-Rodriguez to come outside.

Using a specialized tool called a "jam-ram," the tactical team forced their way through the front-door burglar bars. Id. Officers arrested Cuesta-Rodriguez and took him to the police station. He gave statements to detectives that day and the following day-and in both interviews admitted to shooting Fisher (though he claimed the first shot was accidental). Photographs of Fisher's face showed gunshot wounds to both eyes.3

II. The Trial4

The state of Oklahoma put Cuesta-Rodriguez on trial for first-degree murder, and prosecutors sought the death penalty.

*892A. The Guilt Phase

During the trial, the court admitted testimony from Dr. Jeffrey Gofton based on the report of an autopsy performed by another doctor (Dr. Fred Jordan) who wasn't present and wouldn't be subject to cross-examination.5 "Dr. Gofton testified regarding the examination of the body conducted by Dr. Jordan and gave his own opinions on Fisher's injuries and cause of death based on Dr. Jordan's observations as recorded in his autopsy report." Cuesta-Rodriguez , 241 P.3d at 226-27. "Dr. Gofton explained to the jury the nature of [Fisher's] injuries ... and recited other observations mentioned in Dr. Jordan's report." Id. at 229. "He concluded that a firearm injury to the head was the cause of death and opined that among several possibilities, the method of death was most likely choking on blood that had entered the airway from bone fracturing in the nasal area." Id. He explained that "Fisher would have lost consciousness in a matter of seconds to minutes and could have taken as long as eight minutes to aspirate on the blood." Id. He also pronounced that the second gunshot "was the likely cause of death." Id.

At the end of the trial, the jury found Cuesta-Rodriguez guilty of murder in the first degree.

B. The Penalty Phase

The defense presented evidence of several mitigating circumstances, detailing, among other things, Cuesta-Rodriguez's troubled childhood, his history of alcohol and substance abuse, as well as his experiences emigrating from Cuba.6 His counsel introduced testimony about Cuesta-Rodriguez's good behavior in jail. And his employer and co-workers testified regarding his work ethic and abilities. Family members (both in taped interviews and in person) discussed Cuesta-Rodriguez's background and good qualities. And they expressed their love for him and asked the jury to impose a non-capital sentence.

The jury heard from a psychologist (Dr. James Choca) who testified "ostensibly" in mitigation.7 Appellant's Opening Br. at 7. Dr. Choca told the jury about a childhood injury from when Cuesta-Rodriguez "hit his head against [a] windshield and fractured his skull." Trial Tr. vol. V at 982:19-20. After hospitalization "a metal plate had to be put in" his skull.8 Id. at 982:21. The *893doctor also told the jury about an injury that took place years later in the United States: while working at a lumber yard and driving a tractor, Cuesta-Rodriguez "fell off the tractor and was dragged by the tractor for a few yards until someone was able to stop it."9 Id. at 983:18-20. As a result of that incident, Dr. Choca testified, Cuesta-Rodriguez suffered from back pain and took pain medication. The doctor discussed Cuesta-Rodriguez's history of depression and substance abuse. And he discussed Cuesta-Rodriguez's "social history" "to get some sense for what he had been through." Id. at 985:3, 6-7, 985:9-991:24 (discussing Cuesta-Rodriguez's "difficult life"). Dr. Choca determined that Cuesta-Rodriguez had borderline-personality disorder and discussed the effect of that condition.

Allegedly due to the failure of trial counsel, the jurors didn't hear any additional mitigation evidence regarding Cuesta-Rodriguez's organic brain damage from the childhood incident. Nor did they hear about his post-traumatic stress disorder.10

At the penalty phase of trial, the state argued that Cuesta-Rodriguez deserved the death penalty based on two aggravating circumstances: (1) the heinousness, atrociousness, or cruelty of the murder and (2) the continuing risk Cuesta-Rodriguez posed to society. We now outline the prosecution's comments that are at issue on appeal. These fall into two categories: (1) comments regarding the jury instruction on mitigating circumstances and (2) comments regarding the mitigation evidence that the defense presented.

1. Comments Regarding Jury Instruction

During the penalty phase, the court gave the jury an instruction (instruction nine) that defined mitigating circumstances and explained the jury's role in considering them. Instruction nine states:

Mitigating circumstances are those which, in fairness, sympathy, and mercy, may extenuate or reduce the degree of moral culpability or blame. The determination of what circumstances are mitigating is for you to resolve under the facts and circumstances of this case.

While all twelve jurors must unanimously agree that the State has established beyond a reasonable doubt the existence of at least one aggravating circumstance prior to consideration of the death penalty, unanimous agreement of jurors concerning mitigating circumstances is not required. In addition, mitigating circumstances do not have to be proved beyond a reasonable doubt in order for you to consider them.

Original R. vol. VII at 1284.

Next, the court gave instruction ten, which states, "Evidence has been introduced as to the following mitigating circumstances," and then lists sixteen facts about Cuesta-Rodriguez. Id. at 1285. The *894court told the jury that "[e]vidence ha[d] been introduced as to the following mitigating circumstances": (1) Cuesta-Rodriguez's emigration "from the poverty-stricken Communist country of Cuba"; (2) his journey to the United States during the Mariel Boat Lift of 1980; (3) his time in federal detention after his heroin-possession conviction; (4) the revolt of "some Cubans in the prison who feared repatriation" during Cuesta-Rodriguez's time in federal custody, compared to Cuesta-Rodriguez's "volunteer[ing] for and welcome[ing]" of repatriation "so that he would see his family again"; (5) Cuesta-Rodriguez's "productive[ ]" use of his time in federal detention "to learn to speak and read English"; (6) his "long, stable work history" and status as a "valued employee" who remained a "cherished and trusted friend" to his boss; (7) his volunteer work for seven years helping make "the homes of elderly and needy persons ... safe and habitable"; (8) his status as a likely "asset to a prison community where productive inmate workers are needed" because of his "past employment experiences and willingness to work"; (9) his family in Cuba, with whom "he has maintained regular contact with throughout the years," and who "asked [the jury] to spare" his life; (10) Cuesta-Rodriguez's love for his son, Carlos (Kery) Cuesta Gonzalez, who was inspired by his father to become a writer; (11) Cuesta-Rodriguez's "serious, debilitating depression," which was "made worse by self medication with alcohol and other substances"; (12) his rapidly deteriorating mental condition that, "combined with alcohol and other substances[,] culminated in [his] actions on May 31, 2003 which caused the death of Olimpia Fisher"; (13) Cuesta-Rodriguez's since-improved mental condition, which was "effectively stabilized by medications" that "ease the symptoms of depression and delusions"; (14) Cuesta-Rodriguez's participation in and successful completion of the Oklahoma Department of Mental Health's Wellness Recovery Action Program; (15) his good behavior in the county jail for four years awaiting trial; and (16) his remorse for causing Fisher's death. Id. at 1285-88.

And, in a separate instruction-instruction sixteen-the jury was told: "All the previous instructions given you in the first part of this trial apply where appropriate, except that in this part of the trial, you may consider sympathy or sentiment for the defendant in deciding whether to impose the death penalty." Id. at 1295 (emphasis added).

The prosecution, in its closing argument, referenced instruction nine discussing mitigating circumstances, arguing that the mitigation evidence presented to the jury didn't reduce Cuesta-Rodriguez's moral culpability for the crime. The prosecutor asked, "[H]ow does [the defense's evidence (referring to "the evidence [the jury] heard the last two or three days") ] mitigate what this defendant did on the date in question?" Trial Tr. vol. VII at 1281:21-22, 1282:1-2.

And then the prosecutor referred the jury to "the instructions from His Honor up there," id. at 1282:3-4, stating that mitigating circumstances are circumstances "which, in fairness, sympathy, and mercy, may extenuate or reduce the degree of moral culpability or blame," id. at 1282:6-8. The prosecutor went on to ask what evidence had been presented "that might reduce the moral culpability or blame of" Cuesta-Rodriguez for shooting Fisher. Id. at 1282:17-18. The prosecutor concluded that Cuesta-Rodriguez's emigration from Cuba didn't "reduce the moral culpability of this murder." Id. at 1283:14-15.

And the prosecutor continued on with this theme. See id. at 1283:16-17 ("How *895does it mitigate it? I pose the question to you ...."); id. at 1284:12-14 ("[T]he State of Oklahoma submits that [the family testimony in mitigation] doesn't do anything to reduce the moral culpability of what he did to Olimpia Fisher."). Discussing the family testimony, the prosecutor had this to say: "Do they love him? Sure, they do, even though they haven't seen him in a long time. It's not surprising. It's not helpful to you either." Id. at 1284:14-17.

But the prosecutor interwove with those statements suggestions that the jury could consider the mitigation evidence. The prosecutor told the jury: "And again, I'm not telling you don't listen to them; by all means, you consider what they have to say." Id. at 1284:9-11; see also id. at 1281:17-19 ("[Y]ou still say, all right, does that outweigh the mitigating evidence that we've heard."); id. at 1283:20 ("I'm not going to disparage [the mitigation witnesses].").

After rejecting the import of the mitigation evidence, the prosecutor reminded the jury of the victim-impact testimony. See id. at 1285:18-21 ("You are to go up there and inquire into the moral culpability of what he did and, in doing so, remember the impact testimony that came from these young ladies [Fisher's daughters]."). The prosecutor concluded his argument by stating, "There is one punishment that doesn't undermine the seriousness of [the murder], and that is the punishment of death." Id. at 1286:20-22.

The defense's closing argument also touched on instruction nine. The defense emphasized to the jury that it had flexibility to consider mitigating circumstances, stating that mitigating circumstances "don't have to be proven beyond a reasonable doubt" and needn't be agreed on unanimously. Id. at 1301:24-25. Defense counsel told the jury, "Any level of proof that is enough for you is good enough." Id. at 1301:25-1302:1. And defense counsel stressed to the jurors that they "may consider sympathy or sentiment for the defendant ... because the law says it's right for you to consider them; otherwise, the Court would not have allowed them to come before you." Id. at 1301:13-20. The defense further emphasized that the jury could rely on different mitigating circumstances, including mitigating circumstances not on the list and not talked about during the trial, because "if it's mitigating to you, it's enough." Id. at 1302:5.

The prosecution in rebuttal returned to the theme that the mitigation evidence didn't reduce Cuesta-Rodriguez's culpability. After referencing instruction nine, the prosecutor said: "Counsel told you many times mitigating circumstances are those which, in fairness, sympathy, and mercy-and that's true but there's more-may extenuate or reduce the degree of moral culpability or blame. May extend or reduce the degree of culpability or blame." Id. at 1313:9-14. "So," the prosecutor said, "now let's look at the mitigating evidence they offer." Id. at 1313:15-16. Referencing Cuesta-Rodriguez's Cuban heritage, the prosecutor stated: "And you ask yourselves, looking at the law, does that reduce his degree of culpability or blame? State submits no." Id. at 1313:17-20. And, going through various pieces of the defense's mitigating evidence, the prosecutor again and again reached the same conclusion. See id. at 1314:12-14 ("Ask yourselves how does [the fact that he came to the United States in the Mariel boat lift] reduce his degree of culpability or blame?"); id. at 1315:1-3 ("[W]hat you have to ask yourselves under the law is do you find [the fact that he welcomed repatriation] reduces his degree of moral culpability or blame for this case?"); id. at 1315:6-11 ("[A]nd I won't go through all these [mitigating circumstances] .... And you ask *896yourselves how in the world does that reduce his degree of moral culpability or blame for this case?"). But the prosecution did encourage the jury to consider all the evidence, stating: "[W]e're not asking you to ignore the evidence, but embrace it." Id. at 1315:11-12.

2. Comments Regarding Mitigation Evidence

During its closing argument, the prosecution stated that "the State of Oklahoma does not want to denegrate [sic] any of the evidence you've heard the last two or three days. I will not denegrate [sic] it." Id. at 1281:20-22; see also id. at 1281:24-25 (referring to mitigation witnesses as "fine, upstanding people"). Later, discussing Cuesta-Rodriguez's proffered mitigation evidence, the prosecutor stated:

And as far as them tearfully pleading for his life there, I say to you on behalf of the State, ladies and gentlemen, shame on him for putting them in that position. Shame on him for making them act as a human shield between justice and himself.

Id. at 1284:18-22.

After the prosecution closing, the defense gave its closing argument. In it, defense counsel stated: "In fairness, sympathy, and mercy, refuse the death penalty because there's a family 90 miles from our shores who are a world away who will be hurt. His mother Evi, his sister Arelie, and his brother Juaquin." Id. at 1303:17-20. A few lines later, closing out the argument, counsel stated asked the jury to "refuse the death penalty because there is a son," id. at 1303:21-22, who told Cuesta-Rodriguez that "I want to sit one day across from you. Refuse the death penalty because there is a son who tells his father, I am your son, I have the right to know you. Don't deny Kery Rodriguez [his son] that opportunity. In fairness, sympathy, and mercy, refuse the death penalty," id. 1304:1-6.

The prosecution started its rebuttal closing argument (presented by a different prosecutor) by noting that it planned to "rebut a couple of things [defense] Counsel said." Id. at 1304:13-14. Soon after, the prosecutor, referring to defense counsel's closing argument, told the jury that "what you've heard for 20 minutes is the guilt trip." Id. at 1304:19-20. Defense counsel objected, and the judge asked the prosecutor to rephrase. The prosecutor then told the jury: "You know, when I say guilt trip, you don't need to feel guilty about doing your job. He's the one that brought us together. It is his actions. And I want to talk about that because you can consider sympathy absolutely." Trial Tr. vol. VII at 1306: 5-9. Soon after, the prosecutor continued: "So, yeah, when they want to talk to you about mercy, which you can consider, and I submit to you[,] you decide if you should feel guilty about doing your job. You've got [intervening objection] So when they ask you about mercy, and I say, you don't have to feel guilty if you're sitting on this jury; you're doing your civic duty." Id. at 1309:22-25, 1310:18-20. Later, the prosecutor stated: "As [my colleague] said, you know, shame on him. He puts those people in a terrible position." Id. at 1316:7-8. The prosecution rested after asking the jury to sentence Cuesta-Rodriguez to death: "His actions brought us here. Sentence him accordingly." Id. at 1317:18-19.11

*897That spelled the end of closing arguments, so the jury left to deliberate. During deliberations, the jury asked for the legal definition of culpability. The court answered, with both parties' consent, that the definition is "blame or blameable." Id. at 1318:23-24.

In the end, the jury found the existence of two aggravating circumstances: (1) that the murder was especially heinous, atrocious, or cruel and (2) that Cuesta-Rodriguez posed a continuing threat to society. And the jury recommended a death sentence. Later, the court formally sentenced Cuesta-Rodriguez to death.

III. The Appeals

The OCCA affirmed Cuesta-Rodriguez's conviction and sentence on direct appeal. Cuesta-Rodriguez , 241 P.3d at 247. In doing so, the OCCA found two errors-a Confrontation Clause error and a prosecutorial-misconduct error-but found both individually harmless. Id. at 230-31, 243-44. The OCCA concluded that the two errors were also cumulatively harmless. Id. at 246.

Admitting Dr. Gofton's testimony, the OCCA determined, was in fact error under the Confrontation Clause because "Cuesta-Rodriguez was denied the opportunity to confront Dr. Jordan in order to test his competence and the accuracy of his findings." Id. at 229. But the OCCA determined this error was harmless.12 Id. at 231. The court concluded that yes, Dr. Gofton's testimony was "potentially relevant to proving the heinous, atrocious, or cruel aggravator in the sentencing phase by showing that Fisher consciously suffered before she died." Id. at 230. Yet the OCCA decided that "even if Dr. Gofton's testimony is discounted in its entirety, there was still more than sufficient evidence for the jury to conclude that Fisher consciously suffered before her death." Id. at 231. Specifically, the OCCA pointed to the testimony of police officers and Chacon, as well as Cuesta-Rodriguez's statements to police that "showed that when Cuesta-Rodriguez fired the first blast from his pistol into Fisher's right eye, she was not rendered unconscious." Id. Therefore the OCCA concluded that even excluding Dr. Gofton's testimony, "the jury could have reasonably concluded that Fisher consciously experienced great physical and mental suffering." Id. ; see also id. ("Consequently, even if Dr. Gofton's testimony about how long Fisher may have remained conscious after the second gunshot is eliminated from consideration, there was enough remaining evidence to show conscious suffering in the interval between the first and second shots.").

Regarding his claims of prosecutorial misconduct, Cuesta-Rodriguez argued that "the prosecutors made many statements designed to diminish, denigrate, or completely invalidate the mitigating evidence that was presented." Id. at 243. The OCCA identified just one-"the prosecutor's first 'guilt trip comment' "-which, it concluded, "pushe[d] beyond the limits of permissible argument because it was not a comment on the evidence, but instead was an obvious attempt to denigrate Cuesta-Rodriguez's mitigation defense." Id. at 244. The OCCA noted that the "prosecutor's other two comments referring to 'guilt trip' or feeling *898guilty both c[a]me very close to crossing this line." Id. But it recognized only the first comment as error. See id.

Nonetheless, the OCCA determined that the comments weren't "verdict determinative" and that "given the strength of the evidence supporting imposition of the death penalty, they were harmless." Id. The OCCA concluded that "Cuesta-Rodriguez was not denied a fair or reliable sentencing proceeding." Id.

The OCCA later denied relief on Cuesta-Rodriguez's two post-conviction applications. Cuesta-Rodriguez v. Oklahoma , No. PCD-2012-994 (Okla. Crim. App. Feb. 8, 2013); Cuesta-Rodriguez v. Oklahoma , No. PCD-2007-1191 (Okla. Crim. App. Jan. 31, 2011).

The federal district court then denied Cuesta-Rodriguez's petition for habeas relief. Cuesta-Rodriguez v. Royal , No. CIV-11-1142-M, 2016 WL 5485117, at *1 (W.D. Okla. Sept. 29, 2016). The district court also denied him a COA. But we granted a COA to consider (1) Cuesta-Rodriguez's prosecutorial-misconduct claims; (2) his ineffective-assistance-of-counsel claims, including his procedural-default arguments and the district court's denial of his request for an evidentiary hearing; and (3) his cumulative-error claim. Those claims are now before us on appeal.

DISCUSSION

Cuesta-Rodriguez makes three main arguments on appeal: (1) that he isn't procedurally barred from asserting his ineffective-assistance-of-counsel claims regarding failure to introduce evidence of his organic brain damage and post-traumatic-stress disorder, and that those ineffective-assistance claims warrant relief; (2) that prosecutorial misconduct infringed his right to a fundamentally fair and reliable sentencing proceeding in violation of the Sixth, Eighth, and Fourteenth Amendments; and (3) that even if each individual error was harmless, the cumulative effect of the errors warrants relief. After laying out the standard of review, we address each in turn.

I. Standard of Review

The Antiterrorism and Effective Death Penalty Act of 1996 (AEDPA), 28 U.S.C. § 2254, governs our review of habeas petitions and focuses on how the state court resolved the claim. Byrd v. Workman , 645 F.3d 1159, 1165 (10th Cir. 2011). "In general, if a convicted state criminal defendant can show a federal habeas court that his conviction rests upon a violation of the Federal Constitution, he may well obtain a writ of habeas corpus that requires a new trial, a new sentence, or release." Trevino v. Thaler , 569 U.S. 413, 421, 133 S.Ct. 1911, 185 L.Ed.2d 1044 (2013).

A habeas petitioner must first exhaust his claims in state court before a federal court may review them. 28 U.S.C. § 2254(b)(1)(A). For claims that the state court adjudicated on the merits, we will grant habeas relief only if the petitioner establishes that the state-court decision was "contrary to, or involved an unreasonable application of, clearly established Federal law, as determined by the Supreme Court of the United States," id. § 2254(d)(1), or that the state-court decision "was based on an unreasonable determination of the facts in light of the evidence presented in the State court proceeding," id. § 2254(d)(2). (This is the standard we apply to every issue herein unless otherwise specified.) Claims that the state court didn't adjudicate on the merits, we review de novo. Hooks v. Workman (Hooks II ), 689 F.3d 1148, 1163-64 (10th Cir. 2012).

The focus of § 2254(d) is the reasonableness of the state court's decision.

*899"The question under AEDPA is not whether a federal court believes the state court's determination was incorrect but whether that determination was unreasonable-a substantially higher threshold." Schriro v. Landrigan , 550 U.S. 465, 473, 127 S.Ct. 1933, 167 L.Ed.2d 836 (2007). Relief is warranted only "where there is no possibility fairminded jurists could disagree that the state court's decision conflicts with [the Supreme Court's] precedents." Harrington v. Richter , 562 U.S. 86, 102, 131 S.Ct. 770, 178 L.Ed.2d 624 (2011).

II. Procedural Bar Regarding Mitigation Evidence

Cuesta-Rodriguez argues that, "[d]ue to failures of his trial counsel," the jury that sentenced him to death didn't hear "gold-standard mitigation" evidence about his organic brain damage and post-traumatic-stress disorder that "could readily have resulted in" the selection of a lesser punishment ("life or life without parole").13 Appellant's Opening Br. at 13. But before reaching the merits, we need to decide whether his ineffective-assistance-of-trial-counsel claim is procedurally barred.

Cuesta-Rodriguez didn't bring his ineffective-assistance-of-trial-counsel claim on direct appeal, triggering a state procedural bar. See Okla. Stat. Ann. tit. 22, § 1089 ("The only issues that may be raised in an application for post-conviction relief are those that [w]ere not and could not have been raised in a direct appeal ...."). And Cuesta-Rodriguez didn't claim his appellate counsel was ineffective in his first post-conviction appeal. See Hatch v. State , 924 P.2d 284, 294 (Okla. Crim. App. 1996) ("The issue of ineffective assistance of appellate counsel, like any other claim, must be raised at the first available opportunity."). He first raised his ineffective-assistance claims in his second state post-conviction application in the OCCA-claiming the ineffectiveness of trial, direct appellate, and first post-conviction counsel. Cuesta-Rodriguez , No. PCD-2012-994, slip op. at 3, 5, 6. The OCCA deemed his ineffective-assistance claims waived. See id. at 3-4, 5, 7.

Reviewing Cuesta-Rodriguez's habeas petition, the district court concluded that Cuesta-Rodriguez's ineffective-assistance-of-counsel claims were also procedurally *900barred. Cuesta-Rodriguez , 2016 WL 5485117, at *19.

On appeal, Cuesta-Rodriguez urges us to review his ineffective-assistance-of-trial-counsel claim (and so to review whether the assertedly deficient mitigation presentation violated the Sixth, Eighth, and Fourteenth Amendments), claiming that ineffective appellate and first post-conviction counsel justify our excusing the procedural bar. See Appellant's Opening Br. at 9 ("The district court erred in dealing with this huge and harmful deficit by holding the [ineffective-assistance-of-trial-counsel] claim was procedurally barred from the reach of the federal courts.").

"[T]o bar federal review, a state procedural rule must be adequate to support the judgment and independent from federal law." Banks v. Workman , 692 F.3d 1133, 1145 (10th Cir. 2012). When the adequacy and independence requirements are met, we don't review defaulted issues "unless the petitioner can demonstrate cause and prejudice or a fundamental miscarriage of justice." Anderson v. Sirmons , 476 F.3d 1131, 1140 (10th Cir. 2007) (quoting English v. Cody , 146 F.3d 1257, 1259 (10th Cir. 1998) ).

On appeal, Cuesta-Rodriguez claims (1) that the Oklahoma procedural bar isn't adequate, (2) that he demonstrated cause and prejudice for failing to bring his ineffective-assistance-of-trial-counsel claim on direct appeal, and (3) that a miscarriage of justice occurred that allows us to review his claim. We address each claim in turn.

A. Adequacy of the Procedural Bar14

Cuesta-Rodriguez claims that Oklahoma's procedural bar requiring him to raise ineffective-assistance claims on direct appeal is inadequate.15 He asserts that because his trial and direct-appeal counsel both hailed from the Oklahoma County Public Defender's Office (OCPD), a structural conflict of interest prevented appellate counsel from properly bringing an ineffective-assistance-of-trial-counsel claim.

To be adequate, "a state procedural rule must be 'strictly or regularly followed' and applied 'evenhandedly to all similar claims.' " Banks , 692 F.3d at 1145 (quoting Duvall v. Reynolds , 139 F.3d 768, 797 (10th Cir. 1998) ). And, as Oklahoma points out, we have previously found the Oklahoma procedural bar applied here to be both adequate and independent. See id. at 1144-47.

Oklahoma's system for raising ineffective-assistance-of-counsel claims on direct appeal is inadequate when trial and appellate counsel are too closely intertwined. Cannon v. Mullin , 383 F.3d 1152, 1173 (10th Cir. 2004) ; English , 146 F.3d at 1263-64. Such conflict exists when trial and appellate counsel are one and the same. English , 146 F.3d at 1263-64. And we have held that, sometimes, counsel from the same office are conflicted in choosing to raise ineffective-assistance claims implicating their colleagues. Cannon , 383 F.3d at 1173 ("If a criminal defendant is represented by trial and appellate counsel from the same office, appellate counsel's assessment of trial counsel's performance may be less than completely objective. An understandable, although inappropriate, regard *901for collegiality may restrain appellate counsel from identifying and arguing trial-attorney error.").

"[W]hether trial and appellate attorneys from the same 'office' should be deemed 'separate' counsel will turn on the specific circumstances." Id. "Presenting an ineffective-assistance-of-counsel claim may well damage the reputation of the trial attorney and the office for which both trial and appellate counsel work." Id. ; see also id. ("Arguing ineffective assistance with respect to a colleague's performance is saying that the performance was not only inferior, but unreasonable."). Thus, we must be wary about assuming that counsel is "separate" merely because the individual lawyers are distinct. See id. ("[T]wo lawyers from the same private law firm are often treated as one for conflict-of-interest purposes.").

"[T]he state bears the burden of proving the adequacy of a state procedural bar in order to preclude federal habeas review." Hooks v. Ward , 184 F.3d 1206, 1217 (10th Cir. 1999) ; see also id. at 1216-17 ("[T]he state is undoubtedly in a better position to establish the regularity, consistency and efficiency with which it has applied Rule 3.1116 in the past to allow direct appellants to develop a factual record challenging the adequacy of trial counsel than are habeas petitioners, who often appear pro se, to prove the converse."). But a defendant complaining of such a conflict needs, "at a minimum," to provide "specific allegations ... as to the inadequacy of the state procedure." Id. at 1217 ; see also Cannon , 383 F.3d at 1173-74 (concluding counsel wasn't separate based "[o]n the record before" the court).

Oklahoma highlights a number of cases in which appellate counsel at OCPD, including Cuesta-Rodriguez's appellate counsel, have pursued ineffective-assistance-of-counsel claims. See, e.g. , Coddington v. State , 254 P.3d 684, 692, 713-14 (Okla. Crim. App. 2011) (claim of ineffective assistance of trial counsel based on failure to introduce mitigation evidence during capital penalty phase raised by Cuesta-Rodriguez's appellate counsel); Jiminez v. State , 144 P.3d 903, 904-07 (Okla. Crim. App. 2006) (claim of ineffective assistance of trial counsel raised by Cuesta-Rodriguez's appellate counsel)17 ; see also, e.g. , Frederick v. State , 400 P.3d 786, 825-32 (Okla. Crim. App. 2017) (claim of ineffective assistance of appellate counsel for not raising claim of ineffective assistance of trial counsel), overruled by Williamson v. State , 422 P.3d 752 (Okla. Crim. App. 2018) ; Davis v. State , 268 P.3d 86, 97, 129-38 (Okla. Crim. App. 2011) (same)18 ; Warner v. State , 144 P.3d 838, 861, 868, 872-77, 891-96 (Okla. Crim. App. 2006) (same), overruled by Taylor v. State , 419 P.3d 265 (Okla. Crim. App. 2018).

In light of those cases, Cuesta-Rodriguez hasn't explained how and why his trial and direct-appeal counsel were *902problematically interconnected. He asserts only that trial and appellate counsel both worked for the OCPD-and that they work "just down the hall" from each other. Appellant's Opening Br. at 44 n.21; cf. Cannon , 383 F.3d at 1173 ("A statewide public defender's office with independent local offices, and perhaps even a distinct appellate office, would not raise the same concerns as when trial and appellate counsel work in adjacent rooms."). From that proximity, he infers potential bases for conflicts, like budgetary concerns and loyalty-potential conflicts that exist whenever counsel share an employer. See Appellant's Opening Br. at 43 ("Thus, the office budget must be tapped for an expert to investigate a colleague or such investigative funding must be humiliatingly and improbably requested from the court fund if that is even possible."). Cuesta-Rodriguez also invites us to "[i]magine" the dilemma appellate counsel might be placed in. Id. at 45 n.23. Hence Cuesta-Rodriguez claims that an evidentiary hearing is needed "to explore the specific circumstances and ascertain whether counsel could be deemed separate."19 Id. at 44.

But Cuesta-Rodriguez hasn't shown that a relationship to trial counsel hindered his appellate counsel. So his case bears little resemblance to our prior cases. See 383 F.3d at 1173-74 (concluding trial counsel wasn't separate when the record was "strongly suggestive" of the fact that "appellate counsel had a policy of not claiming ineffective assistance by public defenders at trial"); Carter v. Gibson , 27 F. App'x 934, 943 (10th Cir. 2001) (finding procedural bar inadequate when appellate counsel failed to raise ineffective assistance of trial counsel on direct appeal after trial counsel assisted in writing the appellate brief).

Oklahoma's cases showing regularly-made ineffective-assistance claims suffice to defeat Cuesta-Rodriguez's argument when weighed against the nonexistent conflict evidence proffered. See Cannon , 383 F.3d at 1173-74 ("The culture of an office can also make a substantial difference. A history of raising ineffective-assistance claims could allay concerns."); Smallwood v. Gibson , 191 F.3d 1257, 1270 (10th Cir. 1999) (rejecting the "contention that office policy prevented ... appellate counsel from bringing ineffective assistance ... claims" when "[t]he record contain[ed] no *903evidence that such a policy existed" and instead, "the record indicate[d] that petitioner's appellate counsel aggressively raised" multiple issues, including ineffective assistance, on direct appeal). Thus, we reject Cuesta-Rodriguez's argument and conclude that Oklahoma's procedural bar here was adequate (and that Cuesta-Rodriguez isn't entitled to an evidentiary hearing on the adequacy of that bar).20

B. Cause to Overcome Default

To avoid the application of the procedural bar, Cuesta-Rodriguez argues that he can demonstrate cause for his failure to raise his ineffective-assistance-of-trial-counsel claim on direct appeal.

First, he claims that appellate counsel was ineffective because his appellate counsel wasn't "truly separate" from his trial counsel. Appellant's Opening Br. at 46. But he immediately runs into a problem-his ineffective-assistance-of-appellate-counsel claim is procedurally defaulted because he failed to bring it in his first post-conviction application. See Hatch , 924 P.2d at 294. Thus, Cuesta-Rodriguez argues that his first post-conviction counsel was also ineffective, thereby establishing cause for the failure.

Generally, "ineffective assistance of counsel in postconviction proceedings does not establish cause for the procedural default of a claim." Fairchild v. Trammell , 784 F.3d 702, 720 (10th Cir. 2015) (citing Coleman v. Thompson , 501 U.S. 722, 756-57, 111 S.Ct. 2546, 115 L.Ed.2d 640 (1991) ); see also Coleman , 501 U.S. at 752, 111 S.Ct. 2546 ("There is no constitutional right to an attorney in state post-conviction proceedings. Consequently, a petitioner cannot claim constitutionally ineffective assistance of counsel in such proceedings." (citing Pennsylvania v. Finley , 481 U.S. 551, 107 S.Ct. 1990, 95 L.Ed.2d 539 (1987) ; and Murray v. Giarratano , 492 U.S. 1, 109 S.Ct. 2765, 106 L.Ed.2d 1 (1989) ) ); 28 U.S.C. § 2254(i) ("The ineffectiveness or incompetence of counsel during Federal or State collateral post-conviction proceedings shall not be a ground for relief in a proceeding arising under section 2254.").

We make an exception when "the initial-review collateral proceeding is the first designated proceeding for a prisoner to raise a claim of ineffective assistance at trial," because then "the collateral proceeding is in many ways the equivalent of a prisoner's direct appeal as to the ineffective-assistance claim." Martinez v. Ryan , 566 U.S. 1, 11, 132 S.Ct. 1309, 182 L.Ed.2d 272 (2012). This exception also applies when the "state procedural framework, by reason of its design and operation, makes it highly unlikely in a typical case that a defendant will have a meaningful opportunity to raise a claim of ineffective assistance of trial counsel on direct appeal." Trevino , 569 U.S. at 429, 133 S.Ct. 1911. So when a state's scheme makes a post-conviction proceeding the defendant's first opportunity to raise his trial counsel's ineffective assistance, the ineffective assistance of post-conviction counsel can serve as cause to excuse a failure to raise it then.

Cuesta-Rodriguez claims that the "[l]ack of truly separate counsel on direct appeal means ... that post-conviction was the first opportunity ... to raise trial counsel's ineffectiveness." Appellant's Opening Br.

*904at 46. Thus, Cuesta-Rodriguez asserts that post-conviction counsel's failure to raise an ineffective-assistance claim constitutes cause under the exception established in Martinez , 566 U.S. at 11, 132 S.Ct. 1309, and Trevino , 569 U.S. at 429, 133 S.Ct. 1911.

But Cuesta-Rodriguez's argument ignores the reality of Oklahoma's ineffective-assistance-claim system.21 "Oklahoma provides a reasonable time to investigate a claim of ineffective assistance before raising it on direct appeal." Fairchild , 784 F.3d at 721. An ineffective-assistance claim can be raised in the opening brief on appeal, and that brief can be accompanied by a request to supplement the record. Id. In Fairchild , we noted that Oklahoma's system "allowed appellate counsel to file the [appellate] brief, along with a Rule 3.11 motion to supplement the trial record, 16 months after Defendant was sentenced, with access to the transcript and record for nearly six months." Id. at 722.

Oklahoma points us to our decisions in Fairchild , 784 F.3d at 723, and Banks , 692 F.3d at 1148, in which we determined that Oklahoma's ineffective-assistance-claim structure voids the need for the Martinez and Trevino safety-valve exception. In Fairchild and Banks , we determined that Oklahoma's procedural safeguards allow for ineffective-assistance claims to be brought on direct appeal. That determination controls this case. Indeed, the district court concluded that Oklahoma's Rule 3.11"allows defendants a meaningful opportunity to raise ineffective-assistance-of-trial-counsel claims" on direct appeal. Cuesta-Rodriguez , 2016 WL 5485117, at *19.

But Cuesta-Rodriguez makes two attempts to distinguish his case, arguing: (1) that because he didn't have separate counsel at trial and on direct appeal, his first opportunity to challenge his trial counsel's performance was his first post-conviction application and (2) that "the structure and operation of the Oklahoma system," which regularly results in defendants in Oklahoma and Tulsa Counties receiving representation by the OCPD and the Tulsa County Public Defender's Office, respectively, both at trial and on direct appeal, restricts such defendants from "hav[ing] full access to Rule 3.11."22 Appellant's Opening Br. at 48. Both arguments center on Cuesta-Rodriguez's not having had full access to Rule 3.11 due to conflicted advocates. But we have already concluded that Cuesta-Rodriguez had separate counsel for his trial and direct appeal, so these arguments are foreclosed.

And our conclusions in Fairchild pose an uphill battle for Cuesta-Rodriguez. There, we concluded that Oklahoma's regime was unlike the legal and structural barriers that had worried the Supreme Court in Martinez and Trevino . Fairchild , 784 F.3d at 723 ("Oklahoma law did not preclude raising on direct appeal a claim of ineffective assistance of trial counsel-either as prohibited by state law, as in Martinez , or as a practical consequence of that law, as in Trevino ...."). After Oklahoma presented evidence of public defenders having asserted ineffective-assistance-of-counsel claims, we concluded that Mr. Fairchild hadn't "shown that the 'design *905and operation' of Oklahoma's procedural framework 'ma[d]e[ ] it highly unlikely in a typical case that a defendant w[ould] have a meaningful opportunity to raise a claim of ineffective assistance of trial counsel on direct appeal.' " Id. (alteration in original) (quoting Trevino , 569 U.S. at 429, 133 S.Ct. 1911 ). Likewise, here, Cuesta-Rodriguez agrees that Oklahoma allows for ineffective-assistance claims on direct appeal but claims that the public-defender system's structure prevents defendants from accessing that right. But he has failed to show that "the practical consequence" of Oklahoma's set-up denies the average defendant a meaningful opportunity to raise an ineffective-assistance claim. Id. So this isn't a Martinez or Trevino case. Cf. Pavatt v. Royal , 894 F.3d 1115, 1137 (10th Cir. 2017) (Briscoe, J., concurring and dissenting) ("[T]o bypass the OCCA's procedural bar ruling and review [the petitioner's] ineffective assistance claims on the merits" based on his separate-counsel argument "would be to adopt an entirely new, and potentially much broader, rule than was announced in Martinez and Trevino ").

Martinez and Trevino don't apply to Cuesta's case, so we can't review his first post-conviction counsel's alleged ineffectiveness.23 Hence Cuesta-Rodriguez didn't show cause for his failure to timely raise his ineffective-assistance claims, and the procedural bar holds. And we reject his request for an evidentiary hearing on the same basis-namely, that he hasn't provided specific allegations suggesting that Oklahoma's system was working unfairly.24

C. Fundamental Miscarriage of Justice

On appeal, Cuesta-Rodriguez argues for the first time that he has shown actual innocence of the death penalty-i.e. , that but for constitutional error, no reasonable jury could have found that the aggravating circumstances of his crime outweighed the mitigating circumstances-and that this Court should therefore review his procedurally defaulted claims under the miscarriage-of-justice exception. But we agree with Oklahoma that Cuesta-Rodriguez failed to preserve this argument for appellate review, and so we decline to consider it. See, e.g. , Stouffer v. Trammell , 738 F.3d 1205, 1221 n.13 (10th Cir. 2013) ("We do not generally consider issues that were not raised before the district court as part of the habeas petition."); Heard v.Addison , 728 F.3d 1170, 1175 (10th Cir. 2013) ("We do not reach [the petitioner's argument] in this case, however, because ... we conclude *906that [the petitioner] never raised such a claim, in his petition or otherwise, before the federal district court.").

Cuesta-Rodriguez maintains that we should choose to address his actual-innocence argument because the Supreme Court's opinion in Jenkins v. Hutton , --- U.S. ----, 137 S.Ct. 1769, 198 L.Ed.2d 415 (2017) (per curiam), changed the legal landscape. See United States v. Mora , 293 F.3d 1213, 1218 (10th Cir. 2002) (noting that though "[w]e generally do not consider issues raised for the first time on appeal," we will "occasionally" do so). But we aren't persuaded.

As the Supreme Court explained in Sawyer v. Whitley , 505 U.S. 333, 112 S.Ct. 2514, 120 L.Ed.2d 269 (1992), actual-innocence claims are limited to arguments that "no reasonable juror would have found the petitioner eligible for the death penalty under the applicable state law," i.e. , the elements of the crime itself and the existence of aggravating circumstances. Id. at 336, 344-45, 112 S.Ct. 2514 ; see also id. at 347, 112 S.Ct. 2514 ("The 'actual innocence' requirement must focus on those elements that render a defendant eligible for the death penalty ...."). But, the Court explained, the existence of "additional mitigating evidence that was prevented from being introduced as a result of a claimed constitutional error" is not a proper basis for an actual-innocence claim. Id. at 347, 112 S.Ct. 2514. Later, discussing Sawyer , we explained that "even if state law considers the outweighing of mitigating circumstances by aggravating circumstances an 'element' of a capital sentence, it is not an element for purposes of the actual-innocence inquiry." Black v. Workman , 682 F.3d 880, 916 (10th Cir. 2012). These precedents foreclose Cuesta-Rodriguez's actual-innocence claim.

Hutton hasn't changed that. Indeed, in reversing a Sixth Circuit decision reviewing the merits of a case under the miscarriage-of-justice exception to procedural default, Hutton reaffirmed the core holding of Sawyer . See Hutton , 137 S.Ct. at 1773. The Hutton Court explained that a reviewing court must analyze "whether a properly instructed jury could have recommended death," not "whether the alleged error might have affected the jury's verdict." Id. at 1772.

Cuesta-Rodriguez, though, seizes on the Hutton Court's "[a]ssuming" that a court could excuse default based on a "trial court's failure to specify that, when weighing aggravating and mitigating factors, the jury could consider only the aggravating circumstances it found at the guilt phase." Id. ; see also Appellant's Opening Br. at 40 ("This explication overruled [the Tenth Circuit's] prior jurisprudence that held the concept of innocence of the death penalty did not extend to the weighing process." (citing Black , 682 F.3d at 916 ) ). But the Hutton Court assumed potential error only to reverse the Sixth Circuit's faulty application of Sawyer -nothing in the Court's discussion contradicted its earlier decisions. See Hutton , 137 S.Ct. at 1772-73. In short, nothing in Hutton supports our reviewing Cuesta-Rodriguez's actual-innocence claim. And addressing this fact-laden inquiry when no lower reviewing court did-even tangentially-isn't justified here.

* * *

Having rejected all of Cuesta-Rodriguez's arguments, we don't reach the merits of his ineffective-assistance claims. We turn next to his second proposition on appeal.

III. Prosecutorial Misconduct

Cuesta-Rodriguez claims that "[i]n the penalty phase closing arguments, the prosecutors engaged in a flagrant campaign to *907denigrate or completely invalidate the mitigating evidence." Appellant's Opening Br. at 55. He claims that "[t]hese prosecutorial efforts" "precluded [the jury] from considering as a mitigating factor , an[ ] aspect of [Cuesta-Rodriguez's character] ... and [some] circumstances of the offense that [Cuesta-Rodriguez] proffer[ed] as a basis for a sentence less than death." Id. at 56 (quoting Lockett v. Ohio , 438 U.S. 586, 604, 98 S.Ct. 2954, 57 L.Ed.2d 973 (1978) (plurality opinion) ). That prosecutorial misconduct, he claims, denied him a fundamentally fair trial in violation of the Sixth, Eighth, and Fourteenth Amendments.

"[O]ur interest is in whether [Cuesta-Rodriguez] got a fair trial; 'inappropriate prosecutorial comments, standing alone, [do] not justify a reviewing court to reverse a criminal conviction obtained in an otherwise fair proceeding.' " Matthews v. Workman , 577 F.3d 1175, 1186 (10th Cir. 2009) (second alteration in original) (quoting United States v. Young , 470 U.S. 1, 11, 105 S.Ct. 1038, 84 L.Ed.2d 1 (1985) ). Instead, we reverse for prosecutorial misconduct when errant remarks "so infected the trial with unfairness as to make the resulting conviction a denial of due process." Donnelly v. DeChristoforo , 416 U.S. 637, 643, 94 S.Ct. 1868, 40 L.Ed.2d 431 (1974) ; see also Hanson v. Sherrod , 797 F.3d 810, 843 (10th Cir. 2015) ("Prosecutors are prohibited from violating fundamental principles of fairness, which are basic requirements of Due Process."). That "objectionable content was invited by ... the defense" doesn't "excuse improper comments," but it may be considered in "determin[ing] [the misconduct's] effect on the trial as a whole." Darden , 477 U.S. at 182, 106 S.Ct. 2464 ; see also Tillman v. Cook , 215 F.3d 1116, 1129 (10th Cir. 2000) ("When a prosecutor responds to an attack made by defense counsel, [this Court] evaluate[s] that response in light of the defense argument." (quoting Moore v. Reynolds , 153 F.3d 1086, 1113 (10th Cir. 1998) ) ). "[T]he appropriate standard of review for such a claim on [habeas] is the narrow one of due process, and not the broad exercise of supervisory power." Hanson , 797 F.3d at 843 (second alteration in original) (quoting Darden v. Wainwright , 477 U.S. 168, 181, 106 S.Ct. 2464, 91 L.Ed.2d 144 (1986) ).

Cuesta-Rodriguez contests the OCCA's conclusions that only one prosecutorial-misconduct error occurred and that the one error-the first guilt-trip comment-was harmless. He asserts other comments were also errors, and not harmless ones.

A. Which Statements, if Any, Were Errors?

Cuesta-Rodriguez highlights multiple prosecution statements-(1) those suggesting that "the [defense's] mitigation case was an effort to send jurors on a guilt trip" and that Cuesta-Rodriguez "shamefully" tried to use his family as a human shield and (2) those discussing instruction nine, regarding mitigation-claiming all were error. Appellant's Opening Br. at 55. The OCCA determined that only the prosecution's first guilt-trip statement was an error. We first address the prosecution's comments that allegedly denigrated the defense's mitigation evidence before turning to the comments regarding jury instruction nine. After doing so, we address whether the errors, if any, were harmless.

1. Guilt and Shame Comments

The OCCA found that the first guilt-trip comment was error. The OCCA didn't find the other statements regarding guilt to be error, although the court did find that the comments came "very close to crossing this line." Cuesta-Rodriguez , 241 P.3d at 244. But Cuesta-Rodriguez argues that other guilt-based comments were also errors, *908as well as the shame-on-him (said twice) and human-shield comments. Cuesta-Rodriguez thus claims that "[t]he OCCA's holding is contrary to or an unreasonable application of Supreme Court precedent." Appellant's Opening Br. at 59.

Two facts counter Cuesta-Rodriguez's argument. First, defense counsel didn't object to the shame-on-him and human-shield comments contemporaneously. See Trice v. Ward , 196 F.3d 1151, 1167 (10th Cir. 1999) (noting that the lack of an objection, "while not dispositive, is relevant"). Second, in evaluating prosecution comments' impact, we consider whether the defense invited the comments. See Darden , 477 U.S. at 182, 106 S.Ct. 2464 ("[T]he idea of 'invited response' is used not to excuse improper comments, but to determine their effect on the trial as a whole." (quoting Young , 470 U.S. at 13, 105 S.Ct. 1038 ) ); see also Tillman , 215 F.3d at 1129. Statements of family members that they "love" a defendant aren't " 'relevant mitigating evidence' on which a jury legitimately might ... ground[ ] feelings of sympathy." Coleman v. Saffle , 869 F.2d 1377, 1393 (10th Cir. 1989). And defense counsel attempted to elicit sympathy for Cuesta-Rodriguez's family-his son in particular-based on the pain they would feel if he received the death penalty. Thus, the second shame-on-him comment, as well as the guilt-trip comments, when viewed in light of the defense's approach, were less harmful than they otherwise might have been.25

Beyond the context in which the prosecutor's comments arose, Cuesta-Rodriguez points to little federal law to support his proposition that the OCCA's conclusion (that the comments weren't error) was contrary to established federal law. He cites Dodd v. Trammell , 753 F.3d 971 (10th Cir. 2013), as a case in which "an experienced prosecutor knowingly crossed the line in the penalty phase for an expected effect on the sentencing determination." Appellant's Opening Br. at 60. But Dodd concerned inappropriate victim-impact testimony, and its holding has little bearing on this case. In Dodd , prosecutors had introduced victim-impact evidence in clear violation of Supreme Court precedent, but the OCCA had concluded that the violation was harmless. 753 F.3d at 996-97.26 Though we reversed the OCCA's decision, we took care to note that the case was an outlier, as evidenced by "the sheer volume" of problematic testimony and a "weak[ ] case for the death penalty." Id. at 998.

Ignoring the differences between the two cases, Cuesta-Rodriguez contends that Dodd announces a rule ("the Dodd inference") that reversal is particularly appropriate when prosecutorial misconduct was purposeful. Appellant's Reply Br. at 35. We see no such rule in Dodd -but even if we did, we aren't persuaded that Cuesta-Rodriguez has shown that the prosecutors' statements amounted to purposeful (and *909erroneous) manipulation. As Oklahoma highlights, the prosecutors told the jury that the mitigation evidence could factor into its decision. For example, one prosecutor told the jury to "consider" what Cuesta-Rodriguez's family members "ha[d] to say." Trial Tr. vol. VII at 1284:10-11.

Our decision in Hanson is more on point. See 797 F.3d at 840 (reiterating the standard that "we cannot say that '[a] state court's determination that a claim lacks merit' is wrong on habeas 'so long as "fairminded jurists could disagree" on the correctness of the state court's decision' " (alteration in original) (quoting Harrington , 562 U.S. at 101, 131 S.Ct. 770 ) ). In that case, we upheld the OCCA's conclusion that it wasn't error for a prosecutor to state that it "is also clear that life without parole is not enough accountability for this defendant." Id. at 846. We rejected the argument that the OCCA's conclusion "effectively precluded the jury from considering mitigating evidence." Id. at 847. So too here: We can't say that the prosecution's comments prevented the jury from examining the defense's mitigation evidence. Cuesta-Rodriguez fails to point to a Supreme Court case suggesting otherwise, thus failing to meet his burden under AEDPA.

And as Oklahoma points out, we have denied habeas relief in cases involving similar prosecutorial comments. See, e.g. , Simpson v. Carpenter , 912 F.3d 542, 587 (10th Cir. 2018) (noting that the prosecutor had "improperly denigrated [the petitioner's] mitigating evidence" by "suggesting the defense should be ashamed for relying on [the petitioner's] family support and mental health," but concluding that none of the prosecutor's comments, "separately or cumulatively ... deprived [the petitioner] of a fundamentally fair sentencing proceeding"); Bland v. Sirmons , 459 F.3d 999, 1026 (10th Cir. 2006) (rejecting the claim that the prosecutor had "improperly demeaned [the petitioner's] mitigating evidence" by calling pieces of that evidence "excuses" and asking whether those pieces should "act [as a] shield from accepting the full responsibility for his actions" (internal quotation marks and citation omitted) ); Pickens v. Gibson , 206 F.3d 988, 999-1000 (10th Cir. 2000) (denying habeas relief in a case where the prosecutor referred to a defense argument as a "guilt trip"). So we don't think the OCCA's conclusion-that just one prosecution comment was error-was contrary to established federal law.

But Cuesta-Rodriguez makes another argument.27 He claims that the OCCA (and the district court) incorrectly analyzed the errors separately (rather than together), correctly pointing out that "all the conduct must be considered 'in toto because individual harmless prosecutorial errors can add up to make a trial fundamentally unfair in the aggregate.' " Appellant's Opening Br. at 60 (quoting Le v. Mullin , 311 F.3d 1002, 1022 (10th Cir. 2002) ). Building this argument, Cuesta-Rodriguez argues that the OCCA failed to give the shame-on-him comments any weight in assessing the impact of the erroneous guilt-trip comment.

We aren't persuaded. Under the heading "Prosecutorial Misconduct," the OCCA addressed Cuesta-Rodriguez's "claims that numerous instances of improper argument and questioning of witnesses during the sentencing phase of his trial produced a sentence that failed to meet the heightened standard of reliability in death penalty cases."

*910Cuesta-Rodriguez , 241 P.3d at 243. After "review[ing] the prosecutors' questions and comments cited by Cuesta-Rodriguez as improper," the OCCA concluded that "[w]ith one exception, ... nothing in any of those questions or comments, individually or cumulatively , [went] beyond an attempt to minimize the effect of the evidence presented by the defense, or [went] beyond discussing the evidence in arguing for an appropriate sentence." Id. at 243 (emphasis added). The OCCA wasn't required to address the harmlessness of non-errors. Only actual errors need be included. Le , 311 F.3d at 1023. The OCCA did address the harmlessness of the one error it found-the merits of which we discuss a little later.

2. Jury Instruction Comments

"During the sentencing phase of a capital case, the defendant has a well-established right to introduce 'relevant' mitigating evidence that he proffers as a basis for a sentence less than death." Coleman , 869 F.2d at 1392 (citing Lockett , 438 U.S. at 604, 98 S.Ct. 2954 ).

Cuesta-Rodriguez-pointing to various prosecution statements concerning the jury instruction-alleges that the prosecution improperly exploited instruction nine's language to preclude consideration of mitigating evidence.28

Reviewing the prosecution's approach to instruction nine, the OCCA concluded that "the prosecutor in this case did not urge the jury to categorically disregard the proffered mitigation evidence, but instead argued that the evidence offered in mitigation did not support an inference of reduced culpability." Cuesta-Rodriguez , 241 P.3d at 243. To prevail here, Cuesta-Rodriguez needs to show us that the OCCA's determination was unreasonable.

The OCCA's conclusion that the prosecution didn't try to make the jury ignore mitigation evidence wasn't unreasonable. The prosecution didn't tell the jury not to consider Cuesta-Rodriguez's mitigation evidence. Instead, the prosecution argued that the mitigating testimony shouldn't weigh against a sentence of death-and that's permissible. The prosecution can advocate what evidence the jury should value. It just can't tell the jury that it can't consider the mitigation evidence unless it speaks to culpability.

Cuesta-Rodriguez relies on our decision in Le , where we noted that the prosecution's arguments "may have implied that the jury had the ability to ignore the legal requirement that it must consider mitigating evidence."29 311 F.3d at 1018. But he fails to mention that in Le , we concluded that "the jury was appropriately informed by the jury instructions and by closing arguments that it had to consider mitigating evidence before deciding to impose a death sentence." Id. The same is true here.

We reached a similar conclusion in Hanson , 797 F.3d at 851-52. In that case, "the prosecutor told the jury to consider whether any of the mitigating circumstances 'really extenuate or reduce [the defendant's *911] degree of culpability or blame in this case.' " Id. at 851. But we upheld the OCCA's decision affirming the defendant's death sentence because the prosecutor also encouraged the jury to consider the mitigation evidence and the judge instructed the jury to consider mitigation evidence. Id. at 851-52 ; see also Grant v. Royal , 886 F.3d 874, 939 (10th Cir. 2018) (describing our decision in Hanson , and noting that "because the moral-culpability text itself was not unconstitutional-at least in the context of other, broadening instructions-the prosecutor's isolated references to that text, without more, did not effect a constitutional violation").

We again upheld a death sentence in Grant -a case in which the prosecution made statements like, "[W]hat the law says is that before something can be mitigating it must reduce the moral culpability or blame of the defendant. " 886 F.3d at 937. But the trial court in Grant gave the same instruction given here and in Hanson , listing non-culpability-related circumstances as mitigation. Grant , 886 F.3d at 940 ; Hanson , 797 F.3d at 851. And in affirming the sentence, we relied on the ameliorating jury instructions as a whole and prosecution comments interpreting mitigation circumstances more broadly. Grant , 886 F.3d at 939-40.

We have comparable circumstances here. The prosecution similarly gave a broader view of the mitigating evidence than some isolated comments might suggest, saying things like:

• "[Y]ou still say, all right, does that outweigh the mitigating evidence that we've heard?" Trial Tr. vol. VII at 1281:17-19.

• "And, again, I'm not telling you don't listen to [Cuesta-Rodriguez's family members]; by all means, you consider what they have to say." Id. at 1284:9-11.

• "[W]e're not asking you to ignore the evidence, but embrace it." Id. at 1315:11-12.

See also id. at 1273:3 ("Pay close attention to the Judge's instruction."). So here, like in Hanson , "the prosecutor made a number of other comments to the jury that encouraged them [sic] to consider any and all mitigating evidence they thought relevant." 797 F.3d at 851.

And the jury here heard the same instructions we found curative in Hanson . See Original R. vol. VII at 1285-88; Hanson , 797 F.3d at 851.

First, the judge instructed the jury on sixteen specific mitigating circumstances, "some of which had nothing to do with [the petitioner's] moral culpability." Hanson , 797 F.3d at 851 (listing mitigating circumstances including the petitioner's emotional history, family history, history while incarcerated, and his having a son). "In other words, in this instruction ... the trial judge specifically characterized as 'mitigating' factors that ordinarily would not be deemed to have extenuated or reduced [the petitioner's] moral culpability or blame." Grant , 886 F.3d at 940 (describing mitigating circumstances involving the petitioner's family and emotional history as not extenuating moral culpability or blame). The court instructed the jury on "mitigating" evidence, such as Cuesta-Rodriguez's volunteer work, his learning to read and write English while in federal detention, his family ties, and his relationship with his son. Original R. vol. VII at 1285. And just like in Hanson and Grant , that evidence was described as mitigating even though it doesn't speak to culpability. Hanson , 797 F.3d at 851 ("Viewing the challenged instruction in the context of all the instructions, we do not think the jury would have felt precluded from considering any mitigating evidence ....").

*912Next, "in the same instruction that included the moral-culpability text, there was language that vested the jury with the responsibility for determining what evidence was mitigating." Grant , 886 F.3d at 940 (discussing Hanson ). Here, in instruction nine, the district court told the jury that "[t]he determination of what circumstances are mitigating is for you to resolve under the facts and circumstances of this case." Original R. vol. VII at 1284; see also Grant , 886 F.3d at 940 (relying on this exact instruction and declaring that "a jury is presumed to follow the trial court's instructions"); Hanson , 797 F.3d at 851 (finding that this sentence "broadened any potential limitations imposed by the first sentence of the instruction"). So, again, the court correctly informed the jury of the law.

Last, in the court's final instruction in this case, it told the jury that "in this part of the trial, you may consider sympathy or sentiment for the defendant in deciding whether to impose the death penalty." Original R. vol. VII at 1295. So the jury instructions accurately described the law (including emphasizing the inclusion of mitigating circumstances). See Grant , 886 F.3d at 941 (explaining that accurate, clear jury instructions are relevant "in concluding that the OCCA would not have been unreasonable in determining that the prosecution's closing argument did not have the unconstitutional effect of precluding the jury from considering the petitioner's proffered mitigating evidence that did not extenuate or reduce moral culpability or blame"). We accord that substantial weight. See Boyde v. California , 494 U.S. 370, 384, 110 S.Ct. 1190, 108 L.Ed.2d 316 (1990) (noting that instructions from the court, "we have often recognized, are viewed as definitive and binding statements of the law" (citations omitted) ). And it's worth noting that defense counsel spent substantial time informing the jury of its ability to consider mitigating evidence as well.

Given all that, we can't say that the OCCA's decision was contrary to federal law.

Attempting to escape that conclusion, Cuesta-Rodriguez argues that the jury's question regarding the legal definition of culpability shows that the prosecution's comments misled the jurors. See Appellant's Opening Br. at 69-70 ("The note they sent out during sentencing deliberations asking for guidance on the 'legal definition of culpability' tells us the prosecution's false boundary was working great for them, though unconstitutionally." (quoting Trial Tr. vol. VII at 1318:22) ). From this, Cuesta-Rodriguez argues, "We know to near certainty the prosecution's improper argument worked." Id. at 70.

But, Oklahoma counters, "[t]hat is far too speculative a basis to find that the prosecutor misled the jury into believing it could not consider Petitioner's mitigating circumstances." Appellee's Response Br. at 93. We agree. Determining culpability is a big part of the jury's job at sentencing. All the jury note shows is that the jury read the instruction and had a question: What does legal culpability mean?

Nor do we find persuasive Cuesta-Rodriguez's reliance on Hooks v. Workman (Hooks I ), 606 F.3d 715, 743 (10th Cir. 2010). In that case, we found a jury note to be "a singularly clear indication" that prosecutorial misconduct "did, in fact, mislead the jury." Id. at 745. Cuesta-Rodriguez is right that Hooks I stands for the proposition that questions can be relevant indicators of juror misperception. But that doesn't suggest that the OCCA impermissibly erred here. Desiring to know the definition of culpability-a definition central to the jury's penalty-phase job-doesn't lead us to believe that the prosecution *913led the jury astray. So the note doesn't show that the OCCA's conclusions regarding prosecutorial misconduct were unreasonable.

But Cuesta-Rodriguez makes a couple more arguments that need addressing. He claims that because "the type of misconduct at issue ... invades specific constitutional rights," strict scrutiny applies (which the OCCA didn't use). Appellant's Opening Br. at 67. Cuesta-Rodriguez's argument relies on the claim that "[p]rosecutorial misconduct impinging a specific right demands strict scrutiny." Id. at 68 (citing Caldwell v. Mississippi , 472 U.S. 320, 340, 105 S.Ct. 2633, 86 L.Ed.2d 231 (1985) ). The specific right here is the right to present mitigation evidence.

But Cuesta-Rodriguez hasn't shown that he was denied his right to present mitigation evidence. Indeed, Cuesta-Rodriguez presented substantial mitigation evidence during his trial. See Original R. vol. VII at 1285-88 (listing sixteen mitigating circumstances ranging from Cuesta-Rodriguez's Cuban emigration to his successful work history and strong familial relationships). So he has failed to make the threshold showing.

Moreover, Oklahoma counters Cuesta-Rodriguez's argument by pointing out that no clearly established law supports Cuesta-Rodriguez's proposition that the prosecutorial misconduct here required strict-scrutiny review. We agree.30

Cuesta-Rodriguez relies on Paxton v. Ward , 199 F.3d 1197, 1217-18 (10th Cir. 1999), to prove his point. In Paxton , we noted:

[T]his court has drawn an important distinction between an ordinary claim of prosecutorial misconduct, which warrants habeas relief only when the entire proceeding is rendered fundamentally unfair, and a claim that the misconduct effectively deprived the defendant of a specific constitutional right, which may be the basis for habeas relief without proof that the entire proceeding was unfair.

199 F.3d at 1217. But as we made clear when addressing exactly this issue in Littlejohn v. Trammell , our past decisions can't create clearly established law for AEDPA purposes. See 704 F.3d 817, 838 n.9 (10th Cir. 2013) ("It goes without saying, however, that Paxton cannot supply clearly established federal law to support [the petitioner's] claim.").

Cuesta-Rodriguez also cites two Supreme Court cases in support of his claim that because the prosecutorial statements infringed a constitutional right, the OCCA ought to have applied strict scrutiny. See Appellant's Opening Br. at 67-68 (citing Caldwell , 472 U.S. at 340, 105 S.Ct. 2633 ; and Donnelly , 416 U.S. at 643, 94 S.Ct. 1868 ). Neither proves his point. In Donnelly , addressing prosecutorial-misconduct claims, the Supreme Court stuck with a fundamental-fairness analysis. See 416 U.S. at 643, 94 S.Ct. 1868 ("When specific guarantees of the Bill of Rights are involved, [the Supreme Court] has taken special care to assure that prosecutorial conduct in no way impermissibly infringes them."). And Caldwell distinguished Donnelly , concluding that in Caldwell , "the prosecutor's argument sought to give the jury a view of its role in the capital sentencing procedure that was fundamentally incompatible with the Eighth Amendment[ ]."

*914Caldwell , 472 U.S. at 340, 105 S.Ct. 2633. So there, the Supreme Court reversed (without clarity as to the proper standard of review). But the prosecutors' stray comments here are a far cry from those facts. Cuesta-Rodriguez hasn't provided any sufficiently similar Supreme Court case to prove his point.31

Last, Cuesta urges us to look at the OCCA's decision in Harris v. Oklahoma , 164 P.3d 1103 (Okla. Crim. App. 2007), suggesting that the OCCA acted unreasonably in finding no prosecutorial error here after it concluded in Harris that the instruction was exploitable and that "the kind of prosecutorial argument made here exploited the statutory language improperly." Appellant's Opening Br. at 66 (citing Harris , 164 P.3d at 1113 ). But the relevant inquiry is whether the OCCA acted contrary to Supreme Court precedent, not its own, so we swiftly reject the argument. No potential inconsistency between the two cases allows this court to grant Cuesta-Rodriguez relief. See Lockyer v. Andrade , 538 U.S. 63, 71, 123 S.Ct. 1166, 155 L.Ed.2d 144 (2003) ("[T]he only question that matters under § 2254(d)(1) [is] whether a state court decision is contrary to, or involved an unreasonable application of, clearly established federal law." (citing Weeks v. Angelone , 528 U.S. 225, 120 S.Ct. 727, 145 L.Ed.2d 727 (2000) ) ).

And Cuesta-Rodriguez ignores the distinguishing facts between Harris and this case. In Harris , the OCCA found a prosecutor's comments that "told jurors not to consider [the defendant's] mitigating evidence" improper. 164 P.3d at 1113. But "[u]nlike Harris , ... the prosecutor in this case did not urge the jury to categorically disregard the proffered mitigation evidence, but instead argued that the evidence offered in mitigation did not support an inference of reduced culpability." Cuesta-Rodriguez , 241 P.3d at 243. And despite "the consistent misuse of the language in th[e] instruction" in Harris , the OCCA ultimately concluded that "[t]he prosecutor's improper argument on this issue was cured by further argument and instruction." Harris , 164 P.3d at 1114. Thus, Cuesta-Rodriguez's "reliance on Harris [wa]s misplaced." Hanson , 797 F.3d at 850.

B. Was the Error Harmless?

We next review whether the OCCA acted contrary to established federal law in finding the first guilt-trip comment harmless.

The OCCA concluded that none of the guilt-trip comments "were verdict determinative" and concluded that "given the strength of the evidence supporting imposition of the death penalty, they were harmless." Cuesta-Rodriguez , 241 P.3d at 244. In doing so, the OCCA referenced the statements' impact on the trial as a whole. Id. And, earlier in the opinion, the OCCA laid out in detail the evidence supporting the jury's determination that two aggravating circumstances (heinousness and continuing risk to society) existed. Id. at 237-39. The first guilt-trip comment didn't deny Cuesta-Rodriguez his right to a fundamentally fair trial. And Cuesta-Rodriguez provides us no federal-law basis to reject the OCCA's conclusion that the comment, though disfavored, was harmless.

*915Cuesta-Rodriguez also argues that the OCCA's harmlessness conclusion was contrary to or an unreasonable application of Chapman v. California , 386 U.S. 18, 87 S.Ct. 824, 17 L.Ed.2d 705 (1967), because "the OCCA failed to find the error was harmless beyond a reasonable doubt."32 Appellant's Opening Br. at 59. But the OCCA stated plainly that it analyzed "the context of the entire trial." Cuesta-Rodriguez , 241 P.3d at 243. Here, as in Hanson , "we find it hard to imagine that the jurors thought they were prohibited from considering any of the mitigating evidence they heard at the resentencing hearing." 797 F.3d at 852 (citing Boyde , 494 U.S. at 378-86, 110 S.Ct. 1190 ). Thus, we can't conclude that the OCCA's determination that the guilt-trip comment was harmless was contrary to established federal law.

IV. Cumulative Error

This leaves us with Cuesta-Rodriguez's claim of cumulative error. He argues that even if each individual error was harmless, the cumulative effect of the errors impacted the penalty-phase verdict. Cuesta-Rodriguez highlights three errors to include in the cumulative analysis: (1) the ineffective-assistance-of-counsel claim, (2) the prosecutorial-misconduct errors described above, and (3) the Confrontation Clause error the OCCA determined was harmless.

The OCCA denied Cuesta-Rodriguez's cumulative-error claim on direct appeal, concluding that while "Cuesta-Rodriguez's trial was not error free, the errors do not require relief because when considered in the aggregate, they did not render his trial fundamentally unfair, taint the jury's verdict, or render the sentencing unreliable." Cuesta-Rodriguez , 241 P.3d at 246. Thus, the OCCA concluded that "[a]ny errors were harmless beyond a reasonable doubt, individually and cumulatively." Id.

"In the federal habeas context, a cumulative-error analysis aggregates all constitutional errors found to be harmless and analyzes whether their cumulative effect on the outcome of the trial is such that collectively they can no longer be determined to be harmless."33 Cole v. Trammell , 755 F.3d 1142, 1177 (10th Cir. 2014) (quoting Alverson v. Workman , 595 F.3d 1142, 1162 (10th Cir. 2010) ). "The cumulative-error analysis applies where there are two or more actual errors. It does not apply, however, to the cumulative effect of non-errors." Smith , 824 F.3d at 1255 (quoting United States v. Franklin-El , 555 F.3d 1115, 1128 (10th Cir. 2009) ). To receive habeas relief, Cuesta-Rodriguez must show that "the cumulative effect of the errors determined to be harmless had a 'substantial and injurious effect or influence in determining the jury's verdict.' " Hanson , 797 F.3d at 852 (quoting Brecht v. Abrahamson , 507 U.S. 619, 637, 113 S.Ct. 1710, 123 L.Ed.2d 353 (1993) ).

We first analyze each of Cuesta-Rodriguez's proposed errors to determine whether the error should be included in our cumulative-error analysis.

*916A. Ineffective Assistance

Oklahoma argues that this court cannot rely on procedurally defaulted claims in considering a cumulative-error claim. We agree.34 "[I]n a cumulative error analysis, a court ... may not consider claims that are procedurally defaulted." Ray v. Simmons , 125 F. App'x 943, 946-47 (10th Cir. 2005) ; see also Hughes v. Dretke , 412 F.3d 582, 597 (5th Cir. 2005) ("Meritless claims or claims that are not prejudicial [or claims that are procedurally barred] cannot be cumulated." (alteration in original) (quoting Westley v. Johnson, 83 F.3d 714, 726 (5th Cir. 1996) ) ). So Cuesta-Rodriguez's ineffective-assistance claims, having been ruled procedurally barred, have no place in our cumulative-error analysis.

B. Confrontation Clause

Oklahoma argues that because Cuesta-Rodriguez didn't receive a COA on the underlying Confrontation Clause claim, we can't consider it in the cumulative error analysis.

We disagree. The COA explicitly included the Confrontation Clause claim as one to be included in the cumulative-error analysis: We stated that "harmless constitutional errors found by the Oklahoma Court of Criminal Appeals in appellant's direct appeal concerning the admission of autopsy diagrams and the testimony of Dr. Gofton" were to be included. Order at 2, Cuesta-Rodriguez v. Carpenter , No. 16-6315, (10th Cir. Apr. 10, 2017). True, we can't simply adopt the OCCA's constitutional conclusions. But a lack of explicit mention of the merits in our COA doesn't mean we can't reach them. The COA did what it intended-flagged the potential Confrontation Clause error as one this court could look at in its cumulative error analysis. See Appellant's Opening Br. at 77 n.40 ("Cuesta was granted a certificate of appealability on the cumulative penalty prejudice flowing from [the Confrontation Clause] violation.").

Oklahoma's reliance on Young v. Sirmons , 551 F.3d 942, 972-73 (10th Cir. 2008), for the contrary position is unpersuasive. See Appellee's Response Br. at 100 (describing Young as standing for the proposition that "a cumulative error claim cannot encompass a substantive claim for which a petitioner does not have a COA"). In Young , we rejected a petitioner's argument that we should cumulatively assess all of the trial errors found by the OCCA where: (1) those errors weren't mentioned in the COA35 and (2) some of the errors the petitioner sought to cumulate involved state-law issues. 551 at 973. So this case is clearly distinguishable: (1) the COA explicitly mentions the Confrontation Clause error and (2) the Confrontation Clause issue is squarely one of federal, not state, law.

*917So we can look at the Confrontation Clause issue in the cumulative-error analysis.

It's worth noting, though, that no Supreme Court case has squarely resolved the issue of whether the Confrontation Clause applies at penalty-stage proceedings. See Carter v. Bigelow , 787 F.3d 1269, 1294 (10th Cir. 2015) ("The Supreme Court has never held that the Confrontation Clause applies at capital sentencing."); Wilson v. Sirmons , 536 F.3d 1064, 1111-12 (10th Cir. 2008) ("[W]e have recently stated that it is 'far from clear' whether the Confrontation Clause even applies at capital sentencing proceedings." (quoting United States v. Barrett , 496 F.3d 1079, 1099 (10th Cir. 2007) ) ). But that doesn't preclude our inclusion of the error in our determination whether, in the aggregate, the various errors denied Cuesta-Rodriguez a fair trial.36 See Littlejohn , 704 F.3d at 843 ("Allowing Mr. Littlejohn maximum latitude in addressing his claim, we assume without deciding that the Confrontation Clause applies in capital sentencing proceedings.").

Having concluded we can look at it, we assume without deciding that the Confrontation Clause error found by the OCCA was indeed error, and plug it into our cumulative-error analysis.37

C. Prosecutorial Misconduct

Having concluded earlier that only one applicable error survives-the initial guilt-trip comment, that's the only prosecutorial misconduct we include in our analysis.

D. Cumulative-Error Analysis

To start, Oklahoma argues that because the guilt-trip comment is the only error we can consider, there aren't multiple errors to cumulate. True, a cumulative-error analysis requires more than one error to aggregate. But because we assume, without deciding, that the Confrontation Clause error was error, we have more than one error to address, and so we proceed to the cumulative-error analysis.

AEDPA deference controls our analysis. Cuesta-Rodriguez asks that we include other errors in our cumulative-error analysis-and he then claims that adding more errors means we are evaluating a claim that the OCCA didn't address on the merits, so we should apply de novo review. See Appellant's Opening Br. at 77 ("This Court's review for cumulative error is not under so-called AEDPA deference but rather is de novo ."); Hooks II , 689 F.3d at 1163-64. But we are evaluating the same two errors that the OCCA analyzed and so will uproot the OCCA's decision only if it was contrary to or an unreasonable application of established federal law.38

*918We have already determined that the OCCA's conclusion that the prosecutorial-misconduct error was harmless wasn't unreasonable. All that's left to determine is whether the Confrontation Clause error combined with the prosecutorial misconduct error denied Cuesta-Rodriguez a right to a fundamentally fair trial. In doing so, we are mindful that cumulative error does not require any synergistic effect. Grant v. Trammell , 727 F.3d 1006, 1026 (10th Cir. 2013).

Even so, recognizing that Cuesta-Rodriguez claims such a synergy, Oklahoma points out that he fails to explain how the Confrontation Clause error would have any synergistic effect with the prosecutorial error such that it denied him a fundamentally fair trial. Indeed, Cuesta-Rodriguez acknowledges that the Confrontation Clause violation was relatively minor. See Appellant's Reply Br. at 37 ("[T]he Confrontation Clause violation is not the strongest cumulative error element nor the one on which Mr. Cuesta most relies.").

All Cuesta-Rodriguez tells us on this point is that the Confrontation Clause error could have affected the jury's determination of the heinous, atrocious, or cruel aggravator. On that, the OCCA concluded that ample other evidence in the record showed that Fisher consciously experienced physical and mental suffering before her death. Cuesta-Rodriguez , 241 P.3d at 231. The OCCA pointed to other testimony (that of police officers and Chacon) that showed evidence of a struggle "for at least seven minutes until Cuesta-Rodriguez delivered the fatal shot to her left eye." Id. From there, the OCCA concluded that even without the medical examiner's testimony, "the jury could have reasonably concluded that Fisher consciously experienced great physical and mental suffering." Id.

The OCCA's conclusions weren't unreasonable. As the prosecutor said to the jury in discussing Fisher's suffering, "In this case we don't even have to take just [the medical examiner's] word for it. We know [the first shot] wasn't fatal."39 Trial Tr. vol. VII at 1279:6-8. The prosecutor went on to reference Cuesta-Rodriguez's statement to police and testimony from officers on the scene who heard Fisher's "blood-curdling scream." Id. at 1279:12; see also id. (discussing Fisher's behavior during her last minutes of life). So the OCCA's conclusion that the jury needn't have relied on the medical examiner's testimony stands.

The first guilt-trip comment concerned the defense's mitigation evidence. See Appellant's Opening Br. at 75 ("The prosecution minimized and sought to side step mitigation by falsely describing it as a guilt trip."). This error, though adjudged harmless, was weightier than the Confrontation Clause error. Indeed, the OCCA cautioned "prosecutors in future cases to keep their argument focused on the evidence and to avoid making comments that do nothing but denigrate the defense." Cuesta-Rodriguez , 241 P.3d at 244.

Combining the two errors though, we can't see how the admittedly minor error perhaps influencing the jury's conclusion that the crime was particularly heinous, atrocious, or cruel could have combined with the guilt-trip comment on mitigation to violate Cuesta-Rodriguez's constitutional rights. The two errors relate to different jury findings.

Cuesta-Rodriguez hasn't persuaded us that the combined errors led to a trial that wasn't "fundamentally fair." Cole , 755 F.3d at 1177. So Cuesta-Rodriguez's cumulative-error claim fails.

*919CONCLUSION

For the above reasons, we affirm the district court's judgment.