Appeal from the United States District Court for the Western District of North Carolina, at Statesville. Martin K. Reidinger, Chief District Judge. ( 5:23-cv-00046-MR )

Argued: May 6, 2026 Decided: July 8 , 2026

Before DIAZ, Chief Judge, and RICHARDSON and HEYTENS, Circuit Judges.

Affirmed by published opinion. Judge Richardson wrote the opinion, in which Chief Judge Diaz and Judge Heytens joined.

ARGUED: Mary Elizabeth McNeill, NORTH CAROLINA PRISONER LEGAL SERVICES, INC., Raleigh, North Carolina, for Appellant. Sherri Horner Lawrence, NORTH CAROLINA DEPARTMENT OF JUSTICE, Raleigh, North Carolina, for Appellees. ON BRIEF: Jeff Jackson, Attorney General, NORTH CAROLINA DEPARTMENT OF JUSTICE, Raleigh, North Carolina, for Appellees.

Page 2RICHARDSON, Circuit Judge:

A North Carolina jury convicted George Steen of committing sexual offenses against a foster child in his care. Steen seeks federal habeas relief on the theory that his trial counsel was ineffective because he failed to uncover additional records before trial. The state trial court held that Steen failed to establish either prong of his ineffectiveassistance claim: that his trial counsel performed deficiently or that any deficiency prejudiced Steen. The North Carolina Court of Appeals assumed deficient performance and affirmed solely because Steen failed to show prejudice.

We first address which state court receives the deference required by 28 U.S.C. § 2254(d). The district court deferred only to the state trial court's deficient-performance decision. That was an error. The appellate decision finding no prejudice is the controlling reasoned decision on the merits and is the only decision entitled to deference. But Steen cannot overcome the substantial deference owed to the state appellate court’s finding that he failed to show prejudice. So we affirm the district court’s denial of the writ.

Petitioner George Steen and his wife were asked to care for M.S., a four-year-old foster child, after M.S. was removed from his biological family. M.S. lived with the Steens for most of the next four years. The local department of social services removed M.S. from the Steens’ home after M.S. began displaying troubling sexual behaviors at school. When M.S. was eventually placed in a new foster home, he told his foster mother that Steen had sexually abused him in the shower on numerous occasions. That accusation led to the investigation that resulted in Steen’s convictions.

Page 3At trial, the prosecution’s case relied heavily on the details in M.S.’s descriptions of abuse. It argued that a child of M.S.’s age would not know about the sexual behavior he described unless he had been exposed to it. So Steen’s trial strategy focused on two themes: (1) highlighting M.S.’s reputation for untruthfulness and (2) identifying other sources of M.S.’s sexual knowledge. The jury heard substantial evidence supporting each theme. Various witnesses testified about M.S.’s history of lying. And several witnesses testified about other sources of M.S.’s sexual knowledge, including M.S.’s previous experiences with his biological family; an inappropriate overnight stay with Steen’s brother, who had previously been accused of child molestation; and a sexual experience M.S. had with two older boys in a different foster home—after he left Steen’s home but before he made any allegations against Steen. Still, the jury convicted Steen of three counts of sexual offense with a child under North Carolina law, and the North Carolina Court of Appeals affirmed.

Steen later sought state post-conviction relief, claiming ineffective assistance of counsel under the Sixth Amendment. See generally Strickland v. Washington , 466 U.S. 668 (1984) . His claim focused on his counsel’s decisions during pre-trial discovery. After the state court allowed for post-conviction discovery, Steen identified social services documents, counselor’s notes, and medical records concerning M.S. and his biological sisters that trial counsel had not requested. These records indicated that M.S. witnessed his biological father sexually abuse his sister; that M.S.’s sister sexually abused M.S.; and that one of M.S.’s social workers noted his frequent lies. Because this evidence predated M.S.’s accusation against Steen, Steen argued that it supported his fabrication theory and identified another potential source of M.S.’s sexual knowledge.

Page 4The state trial court denied relief. It recited Strickland ’s two-prong test, under which a defendant must show both deficient performance and prejudice, and concluded that Steen had established neither prong.

The North Carolina Court of Appeals affirmed the denial on prejudice alone. It assumed deficient performance and held that Steen could not show prejudice because the undiscovered records would have been cumulative. The jury had already heard evidence that M.S. lied and that he had prior exposure to sexual conduct. The court therefore concluded that Steen could not “show that there is a reasonable probability that the jury would have reached a different verdict had his trial counsel obtained' the additional records. J.A. 384 .

Steen then sought federal habeas relief under 28 U.S.C. § 2254. The district court granted summary judgment against him, but it did so by deferring to the state trial court’s deficient-performance ruling rather than by reviewing the state appellate court’s prejudice analysis. Steen appealed.

We begin with a threshold issue. When a state prisoner’s federal claim has been rejected at multiple levels of state court, a federal habeas court must identify the operative state-court decision and the deference owed to it. The district court erred in deferring to the state trial court.¹ We first address why we must reject the district court's approach.

Page 5We then address whether other grounds support the district court’s judgment. We conclude that Steen’s petition fails under the correct approach, so we affirm on alternative grounds.

The Antiterrorism and Effective Death Penalty Act of 1996 sharply limited federal habeas relief from state convictions. Chief among AEDPA’s reforms is the requirement that federal courts defer to reasonable state-court merits decisions. When a state petitioner's claim has been 'adjudicated on the merits in State court proceedings,' no federal court may grant the writ unless that adjudication “resulted in a decision” that was (1) contrary to, or involved an unreasonable application of, clearly established federal law, or (2) based on an unreasonable determination of the facts. 28 U.S.C. § 2254(d). Steen proceeds only on the theory that the state courts unreasonably applied federal law when they rejected his ineffective-assistance claim.

AEDPA’s rule instructs federal habeas courts to defer to a state court’s reasonable “decision” that “adjudicated on the merits” a petitioner’s “claim.” Id. But in many cases, multiple levels of state court have adjudicated the petitioner’s claim. So a federal habeas court must identify which state-court decision to review. “This is a straightforward inquiry when the last state court to decide a prisoner’s federal claim explains its decision on the merits in a reasoned opinion.” Wilson v. Sellers , 584 U .S. 122, 125 (2018) . In such a case, that reasoned decision is the relevant one, and “a federal habeas court simply reviews the Page 6specific reasons given by the state court and defers to those reasons if they are reasonable.” 2 Id.

That standard approach resolves many § 2254 petitions. But not all. Some state courts reject a claim without giving any reasons. Others—like the North Carolina Court of Appeals here-reject a claim by analyzing one element without addressing the rest. When a state court stays silent on part (or all) of a claim, identifying whether the relevant state-court “decision . . . involved an unreasonable application” of federal law takes more care. 28 U.S.C. § 2254(d)(1) . 3

Consider first a habeas claim rejected by the state courts without explanation.⁴ In that case, a federal habeas court must still answer the touchstone question of whether that Page 7state-court “decision . . . involved an unreasonable application” of federal law. Id. And the only way to be sure that the state court's unreasoned decision 'involved an unreasonable application” of federal law is to ask whether “there was no reasonable basis for the state court to deny relief.” Richter , 562 U.S. at 98. In other words, when left with an unreasoned decision, a federal court asks whether any reasonable analysis could have yielded the state court’s ultimate decision.

Second, consider a state-appellate-court decision that affirms a reasoned lowercourt decision without explanation. In such a case, we “look through” the unexplained appellate-court decision to the lower court’s reasoning and presume that the higher court “adopted the same reasoning.” See Wilson , 584 U.S. at 125. But that presumption can be rebutted with evidence that the higher court “relied or most likely did rely on different grounds than the lower” court. Id. at 12526 (providing the example of an alternative ground for affirmance that was central to briefing or “obvious in the record”). If the State rebuts the presumption, the lower court's explanation drops from our analysis, and we review the higher court's alternative grounds for decision. See id . 5 Either way, the ultimate question is still whether the highest state-court decision 'involved an unreasonable application” of federal law. 28 U.S.C. § 2254(d)(1).

Our case poses a third scenario. The North Carolina Court of Appeals issued a reasoned decision—but reasoned only on one prong. The court fully explained why Steen Page 8failed to show prejudice, which was enough to reject his claim. It declined to analyze deficient performance, even though the trial court had. So the relevant decision is neither silent nor fully explained: It speaks to one prong and says nothing about the other. For the prejudice prong, the path is settled—a federal court reviews the stated reasons and defers to them if reasonable. Wilson , 584 U.S. at 125. The harder question is what to do with the deficient-performance prong. May a federal court “look through” the appellate court’s reasoned decision to the trial court’s analysis of that prong, as the district court did here? Or must it review that prong de novo ? 6

One might think that-because the appellate court declined to address deficient performance—we ought to look through its decision to the trial court’s analysis of that prong. But that's wrong. Start with AEDPA's touchstone: whether the state court's “decision . . . involved an unreasonable application' of federal law. 28 U.S.C. § 2254(d)(1). Look-through helps answer that question when a court gives no reasons to review—it serves the “specific and narrow purpose” of identifying “the grounds for the higher court’s decision” when that court doesn’t explain itself. Wilson , 584 U.S. at 13 3. Even then, we only look through when the higher court’s unexplained affirmance likely adopts the lower court’s reasoning. But when the higher court explains its reasoning, the federal court no longer needs to make an AEDPA guess.

Page 9That’s the situation we have here. The Court of Appeals told us why it rejected Steen’s Strickland claim: Steen failed to show prejudice. So we review only the higher court’s reason and ask whether it was reasonable. Id. at 125. Because the court told us that deficient performance formed no part of its decision, deference cannot attach to that prong.⁷ A federal court that reaches the deficient-performance prong must therefore decide it de novo . See Porter , 558 U.S. at 39 (reviewing the deficient-performance prong de novo because “the state court did not decide whether [the petitioner’s] counsel was deficient”). The district court instead deferred to an analysis the relevant decision left behind. That was error.⁸

The contrary approach treats separate Strickland prongs as if each could have its own controlling decision. See Laird v. Sec’y, Pa. Dep’t of Corr. , 129 F.4th 227, 243 (3d Cir. 2025) . 9 But that analysis elevates adjudication of an 'element' over adjudication of aPage 10 “claim.” AEDPA demands the opposite. Rather than addressing the adjudication of an element, issue, or prong, AEDPA focuses our attention on “a decision” resulting from the adjudication of a 'claim.' 28 U.S.C. § 2254(d). So even when a state trial court has addressed several elements of a claim, a higher state court’s reasoned decision disposing of the claim with reference to only one element is still the relevant “decision” resolving 'the claim.' And AEDPA's deferential rule only extends to the higher court's stated reasons for rejecting the claim—in this case, the North Carolina Court of Appeals’ analysis of the prejudice prong.

AEDPA’s singular phrasing—“a decision”—supports our reading. The text trains federal review on one operative decision. True, in the most literal sense a trial court’s ruling is also “a decision” on the claim. But no one thinks the trial court’s decision controls once a higher court has resolved the same claim. Consider an example. A state trial court rejects a Strickland claim on both prongs. It clearly errs on the deficient-performance Page 11prong but reasonably applies the prejudice prong. On appeal, a state appellate court corrects the deficient-performance issue but affirms on the prejudice prong. Should a federal court still defer to the appellate court's reasonable analysis of the deficientperformance prong despite the trial court's earlier error? Of course. See Wilson , 584 U.S. at 132-33 (explaining that a federal court should deny the writ when a state appellate decision was decided on reasonable grounds despite the existence of an earlier unreasonable decision).

The lesson runs in both directions. A trial court’s unreasonable analysis cannot defeat deference to the appellate court's reasonable analysis; a trial court's reasonable analysis cannot command deference on a ground the appellate court never reached. Either way, one decision counts in this context: the appellate court’s.

Having rejected the district court's approach, we have two options before us: (1) remand for the district court to conduct a new analysis under the proper framework; or (2) address Steen’s claim under the correct standard. Because the state appellate court thoroughly analyzed the prejudice prong—and the parties briefed the prejudice prong under the correct standard-we take the second approach. We conclude that the state court reasonably analyzed the prejudice prong of Steen’s claim. So we affirm the district court’s ruling on alternative grounds.

A court addressing a Strickland claim can choose which prong to analyze first. See Strickland , 466 U.S. at 697. Rather than review the deficient-performance prong de novo , Page 12we follow the state appellate court’s lead in disposing of this case on the prejudice prong.¹⁰ And, because the state appellate court analyzed that prong on the merits, we review its reasoning under AEDPA’s deferential standard. Under that standard, we “may overturn a state court’s application of federal law only if it is so erroneous that ‘there is no possibility fairminded jurists could disagree that the state court’s decision conflicts with [the Supreme] Court's precedents.'' Nevada v. Jackson , 569 U.S. 505, 508 – 09 (2013) (per curiam) (quoting Richter , 562 U.S. at 102).

To show Strickland prejudice, Steen had to show “a reasonable probability that, but for counsel's unprofessional errors, the result of the proceeding would have been different.” Strickland , 466 U.S. at 694. So the state court's job was to decide whether Steen showed a reasonable probability that, had his trial counsel investigated and presented the additional records, the outcome of the trial would have been different. And, adding in AEDPA’s deferential standard of review, our job is to ask whether all fairminded jurists would agree that the state court’s decision contravened or unreasonably applied Supreme Court holdings. See McCarthy v. Hernandez , 608 U.S. ___ , 2026 WL 1780410, at *4 (June 22, 2026) (per curiam). Unless that high bar is met, we may not grant the writ.

That bar is not met here, so we affirm. The North Carolina Court of Appeals thoroughly explained its decision that Steen failed to show prejudice from his trial Page 13counsel’s alleged deficiency. The court acknowledged Steen’s argument that the additional records would have provided evidence that M.S. tended to lie and that M.S. had other exposure to sexual conduct. But the court ultimately concluded that such evidence would have been cumulative. While some fairminded jurists could conclude that Steen did show prejudice, others could disagree. So we cannot say that the state appellate court’s decision “involved an unreasonable application” of “clearly established Federal law.” 28 U.S.C. § 2254(d)(1). When reasonable minds can disagree, AEDPA calls for deference to the state-court decision. Accordingly, the judgment below is AFFIRMED.