concurring:
I fully concur in all of the Court’s opinion. I write separately to respond to the dissent’s insistence that the Georgia Superior Court, in rejecting the petitioner’s Estelle claims on state collateral review, clearly erred in applying Brecht’s harm*1273less-error standard instead of Chapman’s more petitioner-friendly standard, and that the asserted error is somehow relevant to whether the petitioner is entitled to federal habeas relief. See Diss. Opn. at 1216-19. It did not, and it is not.1 •
The central premise of the dissent’s position is that circuit precedent, as well as what it sees as the “obvious implications” of the United States Supreme Court’s decisions in Brecht and Fry, requires state courts to apply Chapman’s harmless-beyond-a-reasonable-doubt standard on collateral review, not just on direct review. Id. That premise and the conclusions that the dissent draws from it are wrong for a number of reasons.
First, the dissent’s focus on the Georgia Superior Court’s application of Brecht is inconsistent with our task under AEDPA, which is to evaluate whether the highest state court decision addressing the petitioner’s claims is incompatible with clearly established Supreme Court precedent. See 28 U.S.C. § 2254(d)(1); Newland v. Hall, 527 F.3d 1162, 1199 (11th Cir.2008) (“[T]he highest state court decision reaching the merits of a habeas petitioner’s claim is the relevant state court decision.”). As Judge Tjoflat’s opinion for the Court in today’s case correctly explains, the relevant state court decisions for AEDPA purposes are the Georgia Supreme Court’s summary denials of Hittson’s applications for a certificate of probable cause (CPC), not the Georgia Superior Court’s denials of his two state habeas petitions. See Maj. Opn. at 1230-33, 1232-33 n. 25, 1236 n. 27. And “[bjecause we are not reviewing the reasoning announced by the Superior Court,” whether the Superior Court’s application of Brecht was contrary to, or involved an unreasonable application of, clearly established Supreme Court precedent is irrelevant. See Maj. Opn. at 1236 n. 27. Because of AEDPA the only question before us is whether the Georgia Supreme Court had any “reasonable basis” for denying Hittson relief on his Estelle claims. See Harrington v. Richter, 562 U.S. 86, 131 S.Ct. 770, 784, 178 L.Ed.2d 624 (2011) (“Where a state court’s decision is unaccompanied by an explanation, the habeas petitioner’s burden [under AED-PA] still must be met by .showing there was no reasonable basis for the state court to deny relief.”). We cannot presume from the Georgia Supreme Court’s silence that it followed the Superior Court in applying Brecht, especially given that Hitt-son’s application for a CPC argued that the Chapman standard applied under state law. See Maj. Opn. at 1232-33 n. 25, 1236 n. 27; Gill v. Mecusker, 633 F.3d 1272, 1288-89 (11th Cir.2011) (examining whether a summary state appellate decision was entitled to AEDPA deference even though the trial court’s ruling was “based on potentially flawed reasoning”).
Second, our own precedent does not, as the dissent contends, “clearly require[] státe courts hearing habeas corpus appeals to use the more petitioner-friendly standard provided in Chapman.” Diss. Opn. *1274at 1280. Citing our decisions in Trepal v. Secretary, Florida Department of Corrections, 684 F.3d 1088 (11th Cir.2012), and Duest v. Singletary, 997 F.2d 1336 (11th Cir.1993), the dissent asserts that “we have unambiguously required that state collateral courts apply Chapman in reviewing ... federal constitutional claims.” Diss. Opn. at 1280. That is wrong, unless you count ambiguous dicta as an unambiguous holding.
In Trepal we explicitly recognized that “Chapman was a direct-appeal case” and that “Brecht determined that [Chapman’s] harmless error standard did not apply on collateral review.” 684 F.3d at 1111. Although we said in a footnote that Brecht “does not apply to state courts’ review of their own convictions,” that statement was founded upon the observation that, as a matter of practice, “Florida courts apply the more petitioner-friendly Chapman standard of whether the constitutional error is ‘harmless beyond a reasonable doubt.’ ” Id. at 1112 n. 27 (quoting Pittman v. State, 90 So.3d 794, 811 (Fla.2011)). In Trepal we had no occasion to decide, and did not purport to decide, whether state courts are constitutionally required to apply Chapman on collateral review because the state collateral courts in that case had rejected the petitioner’s claims on the merits, not on grounds of harmlessness; they did not apply either the Chapman or the Brecht standard. See id. at 1104-06. The most that can be said about the dicta in our Trepal opinion is that it recognized that state courts are required to apply Chapman on direct review and that they usually do apply that same standard on collateral review. We did not say, even in dicta, that they must do so.
In any event, our dissenting colleague fails to recognize that whatever one may attempt to read into one of our opinions, or indeed whatever they may say, a decision can never hold anything beyond the facts of the case before it; all else is dicta, which is not binding on anyone for any purpose. See Edwards v. Prime, Inc., 602 F.3d 1276, 1298 (11th Cir.2010) (“We have pointed out many times that regardless of what a court says in its opinion, the decision can hold nothing beyond the facts of that case. All statements that go beyond the facts of the case ... are dicta. And dicta is not binding on anyone for any purpose.”) (citations omitted); Pretka v. Kolter City Plaza II, Inc., 608 F.3d 744, 762 (11th Cir.2010) (“We are not required to follow dicta in our own prior decisions. Nor for that matter is anyone else.”) (citation omitted); Watts v. BellSouth Telecomms., Inc., 316 F.3d 1203, 1207 (11th Cir.2003) (“Whatever their opinions say, judicial decisions cannot make law beyond the facts of the cases in which those decisions are announced.”); United States v. Aguillard, 217 F.3d 1319, 1321. (11th Cir.2000) (“The holdings of a prior decision can reach only as far as the facts and circumstances presented to the Court in the case which produced that decision.”) (quotation marks omitted); Browning v. AT & T Paradyne, 120 F.3d 222, 225 n. 7 (11th Cir.1997) (“Since this statement was not part of any holding in the case, it is dicta and we are not bound by it.”).
Our decision in Duest is no different. In that case we merely acknowledged that Brecht’s harmless-error standard for federal habeas review is identical to “the harmless-error standard federal appellate courts use on direct review as to noncon-stitutional error,” and then vaguely noted that Chapman applies “in all other situations.” Duest, 997 F.2d at 1338 & n. 2. In context, the most that can be said about our statements in Duest is that they recognize that Chapman applies to constitutional errors reviewed on direct appeal, whether in state or federal court. Like Trepal, our Duest decision did not address, did not purport to address, and could not have *1275held anything about whether Chapman must be applied in state collateral proceedings, because the state collateral courts in that case had actually applied Chapman. Id. at 1339 n. 4. Neither Trepal nor Duest nor any other decision issued by this circuit has held, or had occasion to hold, that state courts must apply Chapman not only on direct review, but also on collateral review.
And even if there were an actual holding to that effect in one of our decisions, it would still be irrelevant under AEDPA, which forecloses federal habeas relief unless the relevant 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.” 28 U.S.C. § 2254(d)(1) (emphasis added). By relying on our precedent, our dissenting colleague forgets what the Supreme Court has repeatedly told us, which is that the only thing that can clearly establish federal law for AEDPA purposes is a holding of the United States Supreme Court, not dicta, and certainly not the holdings of lower federal courts. See Parker v. Matthews, — U.S. -, 132 S.Ct. 2148, 2155, 183 L.Ed.2d 32 (2012) (explaining that the decisions of federal appeals courts “cannot form the basis for habeas relief under AEDPA” because they do “not constitute clearly established Federal law, as determined by the Supreme Court”) (quotation marks omitted); Marshall v. Rodgers, U.S. -, 133 S.Ct. 1446, 1450, 185 L.Ed.2d 540 (2013) (“The Court of Appeals’ contrary conclusion rested in part on the mistaken belief that circuit precedent may be used to refíne or sharpen a general principle of Supreme Court jurisprudence into a specific legal rule that this Court has not announced.”); see also White v. Woodall, — U.S. -, 134 S.Ct. 1697, 1702 n. 2, 188 L.Ed.2d 698 (2014); Renico v. Lett, 559 U.S. 766, 778-79, 130 S.Ct. 1855, 1865-66, 176 L.Ed.2d 678 (2010); Dombrowski v. Mingo, 543 F.3d 1270, 1274 (11th Cir.2008) (“We have held that the ‘clearly established law’ requirement of § 2254(d)(1) does not include the law of the lower federal courts.”); Putman v. Head, 268 F.3d 1223, 1241 (11th Cir.2001) (“Clearly established federal law is not the case law of the lower federal courts, including this Court.”).
The third reason that the premise and conclusion of the dissenting opinion is wrong is that the Supreme Court’s decisions in Brecht and Fry do not imply, let alone clearly hold (and thereby establish for AEDPA purposes), that state courts are required to apply Chapman’s harmless-error standard on collateral review. To support its belief that they do, the dissenting opinion clips language from those two decisions, disregards the context of that language, and omits most of the limiting references to direct review. See Diss. Opn. at 1280-81. Brecht and Fry both underscored that Chapman was decided on direct review and that it requires states to apply its harmless-beyond-a-reasonable-doubt standard on direct review. See Brecht, 507 U.S. at 630, 636, 113 S.Ct. at 1718, 1721 (explaining that “Chapman reached this Court on direct review” and set the standard that state courts must “engage in on direct review”) (emphasis added); Fry, 551 U.S. at 116, 127 S.Ct. at 2325 (“In Chapman, ... a case that reached this Court on direct review of a state-court criminal judgment, we held that a federal constitutional error can be considered harmless only if a court is able to declare a belief that it was harmless beyond a reasonable doubt.”) (quotation marks omitted). In Brecht, the Supreme Court declined to extend Chapman’s standard to federal habeas review, broadly concluding that a “substantial and injurious effect” standard was “better tailored to the nature and purpose of collateral review than the Chapman standard.” 507 *1276U.S. at 623, 113 S.Ct. at 1714 (emphasis added).
One of the primary reasons the Brecht Court gave for refusing to extend Chapman to federal collateral review — a reason equally applicable to state collateral review — is the fundamental difference between direct review, which “is the principal avenue for challenging a conviction,” and collateral review, where the state has an “interest in the finality of convictions that have survived direct review within the state court system.” Id. at 633-35, 113 S.Ct. at 1719-20; see also Mansfield v. Sec’y, Dep’t of Corr., 679 F.3d 1301, 1307 (11th Cir.2012) (“The Supreme Court emphasized in Brecht that ‘collateral review is different from direct review,’ and, therefore, that ‘an error that may justify reversal on direct appeal will not necessarily support a collateral attack on a final judgment.’ ”) (quoting Brecht, 507 U.S. at 633-34, 113 S.Ct. at 1719-20). Nothing in Brecht implies, let alone clearly establishes, that state courts must apply Chapman on collateral review. If anything, Brecht’s principal rationale — that “collateral review is different from direct review” and that the “substantial and injurious effect” standard is “better tailored to the nature and purpose of collateral review than the Chapman standard” — implies that state courts, like federal courts, are not bound to apply the Chapman standard when conducting collateral review.
The Supreme Court’s Fry decision took Brecht one step further away from the position of the dissenting opinion in the present case. Fry held that federal habeas courts must apply the “substantial and injurious effect” standard to evaluate constitutional errors regardless of whether the state courts recognized the error and reviewed it for harmlessness under the Chapman standard. Fry, 551 U.S. at 121— 22, 127 S.Ct. at 2328. Fry did not imply, much less definitively decide, that state courts must apply Chapman on collateral review. Indeed, it had no occasion to address that question because the relevant state court decision in that case — which notably failed to apply Chapman — was issued on direct appeal, not on collateral review. See id. at 115, 166 n. 1, 127 S.Ct. at 2324, 2325 n. 1. Thus, when the Supreme Court in Fry noted that “state courts are required to evaluate constitutional error under Chapman,” it was simply reiterating the well-worn principle that Chapman sets forth the appropriate harmless-error standard on direct appeal, which is what that case was. Id. at 118, 127 S.Ct. at 2326. That Brecht, Fry, and circuit precedent establish that regardless of what the state courts do, federal habeas courts must apply Brecht’s harmless-error standard does not imply, as the dissent suggests, that Chapman is the only appropriate standard on state collateral review. See Diss. Op. at 1279-80; cf. Brooks v. Kyler, 204 F.3d 102, 108 (3d Cir.2000) (“[Djrawing instruction from Supreme Court pássages through the use of the negative pregnant is risky and unsatisfactory.”). It implies nothing at all about the proper harmlessness standard for state collateral review.
From its unfounded premise that state collateral courts are categorically required to apply Chapman instead of Brecht, the dissent suggests that the harmless-error standard employed by a state collateral court may affect whether a petitioner is entitled to federal habeas review. See Diss. Opn. at 1281-82. Setting aside the flaws in its underlying premise, which have already been discussed, the dissent’s suggestion cannot be squared with the Supreme Court’s decision in Fry. In that case a state appellate court had rejected a petitioner’s constitutional claim on direct appeal, concluding that the asserted error resulted in “no possible prejudice.” Fry, 551 U.S. at 115, 127 S.Ct. at 2324 (quota*1277tion marks omitted). The state court, however, “did not specify which harmless-error standard it was applying,” and the Supreme Court assumed for the sake of argument that the state court .“did not determine the harmlessness of the error under the Chapman standard,” as it was required to do on direct appeal. Id. at 115, 116 n. 1, 127 S.Ct. at 2324, 2325 n. 1. Even with that assumption, the Supreme Court concluded that the petitioner in Fry was not entitled to federal habeas relief because the error was harmless under the Brecht standard. Id. at 116, 122, 127 S.Ct. at 2324, 2328. The Fry Court held that a federal habeas court must apply Brecht “whether or not the state appellate court recognized the [constitutional] error and reviewed it for harmlessness under the harmless beyond a reasonable doubt standard set forth in Chapman.” Id. at 121-22, 127 S.Ct. at 2328 (quotation marks omitted). And it explained that a petitioner, even one who can demonstrate that the state court’s “harmlessness determination ... was unreasonable,” is not entitled to federal habeas relief unless he “also satisfies] Brecht’s standard.” Id. at 119, 127 S.Ct. at 2326-27 (emphasis omitted); see also Mansfield, 679 F.3d at 1308 (“[A] federal court may deny habeas relief based solely on a determination that the constitutional error is harmless under the Brecht standard.”).
In other words, even if a state court failed to recognize a constitutional error, failed to review it for harmlessness, or failed to apply the correct harmless-error standard, the Supreme Court’s Fry decision requires that federal habeas relief must still be denied if the error was harmless under Brecht’s “substantial and injurious effect” standard. We have recognized exactly that. See Hodges v. Att’y Gen., State of Fla., 506 F.3d 1337, 1342-43 (11th Cir.2007) (“[I]f the state court did not apply the correct harmless error standard, or even if it did not recognize that there was error, federal habeas relief is still due to be denied if the constitutional error was harmless [under Brecht].”).
The dissenting opinion’s insistence that Fry, despite its clear language, is unclear on this point rests on the spurious notion that Fry “assumes that state collateral courts will properly apply the Chapman standard.” Diss. Opn. at 1281 n. 4. Fry assumed no such thing. To the contrary, it explicitly assumed just the opposite. The Court plainly stated that it was assuming that the state appellate court in that case had not applied the correct harmlessness standard. Fry, 551 U.S. at 116 n. 1, 127 S.Ct. at 2325 n. 1 (“We also assume that the state appellate court did not determine the harmlessness of the error under the Chapman standard_”). Judge Tjoflat’s opinion for this Court is absolutely right when it says that Hittson would not be entitled to federal habeas relief under Brecht’s harmless-error standard even if we assumed both that the Georgia Supreme Court had applied Brecht and that that application was contrary to clearly established Supreme Court precedent. That is what Fry says and holds. The dissenting opinion errs in suggesting otherwise.
I have put off until the end the most noteworthy of the dissenting opinion’s errors, one that would have the most profound and far reaching impact if it were the law, which thankfully it is not. The error is embodied in that opinion’s attitude about the relationship between this Court and the state courts. The dissenting opinion takes the position that we have the authority and the duty to lecture state courts about federal law and to admonish them to follow the law, that we are their teachers, and that the state courts in this circuit are bound to follow our views. In footnote 4, for example, the dissenting opinion says that regardless of whether *1278the state collateral trial court’s application of the Brecht standard is a basis for granting federal habeas relief, “we should be clear that it was error, that it should not be repeated, and that it matters.” Diss. Opn. at 1281 n. 4. Instead, what we should be clear about is that outside of what is necessary to decide if federal habeas relief is due to be granted, it is not the role of inferior federal courts, of which we are one, to sit in judgment of state courts on issues of federal law, and we should not arrogate to ourselves the role of lecturing them that one of their rulings is an error of federal law that “should not be repeated.” The Supreme Court alone, and not any inferior federal court, may do that. We have no more right to lecture state courts about federal law than they have to lecture us about it.
The Supreme Court has rejected and disparaged as “remarkable” a passage from a Ninth Circuit opinion saying that state courts are bound to follow rulings of the federal court of appeals in the circuit in which they are located. Arizonans for Official English v. Arizona, 520 U.S. 43, 58 n. 11, 117 S.Ct. 1055, 1064 n. 11, 137 L.Ed.2d 170 (1997). In making it clear that the Ninth Circuit’s view was wrong, the Supreme Court cited with favor its own decision in ASARCO Inc. v. Kadish, 490 U.S. 605, 617, 109 S.Ct. 2037, 2045, 104 L.Ed.2d 696 (1989), which instructed federal courts that “state courts ... possess the authority, absent a provision for exclusive federal jurisdiction, to render binding judicial decisions that rest on their own interpretations of federal law.” Id. The Court also cited with favor Justice Thomas’ concurring opinion in Lockhart v. Fretwell, 506 U.S. 364, 375-376, 113 S.Ct. 838, 845-846, 122 L.Ed.2d 180 (1993), for the proposition that the “Supremacy Clause does not require state courts to follow rulings by federal courts of appeals on questions of federal law.” Id. As Justice Thomas explained in that opinion: “The Supremacy Clause demands that state law yield to federal law, but neither federal supremacy nor any other principle of federal law requires that a state court’s interpretation of federal law give way to a (lower) federal court’s interpretation. In our federal system, a state trial court’s interpretation of federal law is no less authoritative than that of the federal court of appeals in whose circuit the trial court is located.” Lockhart, 506 U.S. at 376, 113 S.Ct. at 846.
We have reiterated that same core principle of federalism. See Casale v. Tillman, 558 F.3d 1258, 1260 (11th Cir.2009) (describing as a “well settled” principle of federalism that “a state court’s interpretation of federal law is no less authoritative than that of the corresponding federal court of appeals”) (brackets and quotation marks omitted); Glassroth v. Moore, 335 F.3d 1282, 1302 n. 6 (11th Cir.2003) (“[Sjtate courts when acting judicially, which they do when deciding cases brought before them by litigants, are not bound to agree with or apply the decisions of federal district courts and courts of appeal.”); see also Powell v. Powell, 80 F.3d 464, 467 (11th Cir.1996) (referring to “the dual dignity of state and federal court decisions interpreting federal law” as going “to the heart of our system of federalism”).
. The dissenting opinion seems to be of two minds about whether the harmlessness standard that the state collateral trial court applied is relevant to the issue of whether federal habeas relief should be granted. The first paragraph of that opinion states: "To be clear, I agree with the Concurrence that the state court’s application of the incorrect standard would not be an independent basis for granting habeas relief.” Diss. Opn. at 1279. But the dissenting opinion goes on to state, "I disagree with the Majority's statement that it does not matter whether Georgia’s application of Brecht instead of Chapman was unreasonable.” Id. at 1282. And it says, "Fry makes it unclear whether we can grant habe-as relief based upon the [state collateral trial] court's error....” Id. at 1281 n. 4. As a precaution, I will respond to the dissenting opinion’s second mind, which says that it does matter, as well as to that opinion’s assertion that "it was error,” id.