delivered the opinion of the Court.
Not infrequently, an attorney appointed to represent an indigent defendant on appeal concludes that an appeal would be frivolous and requests that the appellate court allow him to withdraw or that the court dispose of the case without the filing of merits briefs. In Anders v. California, 386 U. S. 738 (1967), we held that, in order to protect indigent defendants’ constitutional right to appellate counsel, courts must safeguard against the risk of granting such requests in cases where the appeal is not actually frivolous. We found inadequate California’s procedure — which permitted appellate counsel to withdraw upon filing a conelusory letter stating that the appeal had “no merit” and permitted the appellate court to affirm the conviction upon reaching the same conclusion following a review of the record. We went on to set *265forth an acceptable procedure. California has since adopted a new procedure, which departs in some respects from the one that we delineated in Anders. The question is whether that departure is fatal. We hold that it is not. The procedure we sketched in Anders is a prophylactic one; the States are free to adopt different procedures, so long as those procedures adequately safeguard a defendant’s right to appellate counsel.
I
A
Under California’s new procedure, established in People v. Wende, 25 Cal. 3d 436, 441-442, 600 P. 2d 1071, 1074-1075 (1979), and followed in numerous cases since then, see, e. g., People v. Rowland, 75 Cal. App. 4th 61, 63, 88 Cal. Rptr. 2d 900, 901 (1999), counsel, upon concluding that an appeal would be frivolous, files a brief with the appellate court that summarizes the procedural and factual history of the case, with citations of the record. He also attests that he has reviewed the record, explained his evaluation of the case to his client, provided the client with a copy of the brief, and informed the client of his right to file a pro se supplemental brief. He further requests that the court independently examine the record for arguable issues. Unlike under the An-ders procedure, counsel following Wende neither explicitly states that his review has led him to conclude that an appeal would be frivolous (although that is considered implicit, see Wende, 25 Cal. 3d, at 441-442, 600 P. 2d, at 1075) nor requests leave to withdraw. Instead, he is silent on the merits of the case and expresses his availability to brief any issues on which the court might desire briefing. See generally id., at 438, 441-442, 600 P. 2d, at 1072, 1074-1075.
The appellate court, upon receiving a “Wende brief,” must “conduct a review of the entire record,” regardless of whether the defendant has filed a pro se brief. Id., at 441-442, 600 P. 2d, at 1074-1075. The California Supreme Court *266in Wende required such a thorough review notwithstanding a dissenting Justice’s argument that it was unnecessary and exceeded the review that a court performs under Anders. See 25 Cal. 3d, at 444-445, 600 P. 2d, at 1077 (Clark, J., concurring in judgment and dissenting in part); see also id., at 444, 600 P 2d, at 1076 (“The precise holding in Anders was that a ‘no merit’ letter . . . ‘was not enough.’. . . Just what is ‘enough’ is not clear, but the majority of the court in that case did not require an appellate court to function as co-counsel”). If the appellate court, after its review of the record pursuant to Wende, also finds the appeal to be frivolous, it may affirm. See id., at 443, 600 P 2d, at 1076 (majority opinion). If, however, it finds an arguable (i. e., nonfrivolous) issue, it orders briefing on that issue. Id., at 442, n. 3, 600 P. 2d, at 1075, n. 3.1
B
In 1990, a California state-court jury convicted respondent Lee Robbins of second-degree murder (for fatally shooting his former roommate) and of grand theft of an automobile (for stealing a truck that he used to flee the State after committing the murder). Robbins was sentenced to 17 years to life. He elected to represent himself at trial, but on appeal *267he received appointed counsel. His appointed counsel, concluding that an appeal would be frivolous, filed with the California Court of Appeal a brief that complied with the Wende procedure.2 Robbins also availed himself of his right under Wende to file a pro se supplemental brief, filing a brief in which he contended that there was insufficient evidence to support his conviction and that the prosecutor violated Brady v. Maryland, 373 U. S. 83 (1963), by failing to disclose exculpatory evidence.
The California Court of Appeal, agreeing with counsel's assessment of the case, affirmed. The court explained that it had “examined the entire record” and had, as a result, concluded both that counsel had fully complied with his responsibilities under Wende and that “no arguable issues exist.” App. 39. The court added that the two issues that Robbins raised in his supplemental brief had no support in the record. Ibid. The California Supreme Court denied Robbins’ petition for review.
After exhausting state postconviction remedies, Robbins filed in the United States District Court for the Central District of California the instant petition for a writ of habeas corpus pursuant to 28 U. S. C. §2254.3 Robbins renewed his Brady claim, argued that the state trial court had erred by not allowing him to withdraw his waiver of his right to trial counsel, and added nine other claims of trial error. In addition, and most importantly for present purposes, he claimed that he had been denied effective assistance of appellate counsel because his appellate counsel’s Wende brief failed to comply with Anders v. California, 386 U. S., at 744. Anders *268set forth a procedure for an appellate counsel to follow in seeking permission to withdraw from the representation when he concludes that an appeal would be frivolous; that procedure includes the requirement that counsel file a brief “referring to anything in the record that might arguably support the appeal,” ibid.
The District Court agreed with Robbins’ last claim, concluding that there were at least two issues that, pursuant to Anders, counsel should have raised in his brief (in a Wende brief, as noted above, counsel is not required to raise issues): first, whether the prison law library was adequate for Robbins’ needs in preparing his defense after he elected to dismiss his appointed counsel and proceed pro se at trial, and, second, whether the trial court erred in refusing to allow him to withdraw his waiver of counsel. The District Court did not attempt to determine the likelihood that either of these two issues would have prevailed in an appeal. Rather, it simply concluded that, in the language of the Anders procedure, these issues “might arguably” have “support[ed] the appeal,” App. 51, n. 6 (citing Anders), and thus that Robbins’ appellate counsel, by not including them in his brief, deviated from the procedure set forth in Anders. The court concluded that such a deviation amounted to deficient performance by counsel. In addition, rather than requiring Robbins to show that he suffered prejudice from this deficient performance, the District Court applied a presumption of prejudice. App. 49. Thus, based simply on a finding that appellate counsel’s brief was inadequate under Anders, the District Court ordered California to grant respondent a new appeal within 30 days or else release him from custody.
The United States Court of Appeals for the Ninth Circuit agreed with the District Court on the Anders issue. In the Ninth Circuit’s view, Anders, together with Douglas v. California, 372 U. S. 353 (1963), which held that States must provide appointed counsel to indigent criminal defendants on appeal, “set forth the exclusive procedure through which ap*269pointed counsel’s performance can pass constitutional muster.” 152 F. 3d 1062, 1066 (1998). Rejecting petitioner’s argument that counsel’s brief was sufficient because it complied with Wende, the Ninth Circuit concluded that the brief was deficient because it did not, as the Anders procedure requires, identify any legal issues that arguably could have supported the appeal. 152 F. 3d, at 1066-1067.4 The court did not decide whether a counsel’s deviation from Anders, standing alone, would warrant a new appeal, see 152 F. 3d, at 1066-1067, but rather concluded that the District Court’s award of relief was proper because counsel had failed to brief the two arguable issues that the District Court identified. The Ninth Circuit remanded, however, for the District Court to consider respondent’s 11 claims of trial error. Id., at 1069. The court reasoned that if Robbins prevailed on any of these claims, it would be unnecessary to order the California Court of Appeal to grant a new direct appeal. We granted certiorari. 526 U. S. 1003 (1999).
II
A
In Anders, we reviewed an earlier California procedure for handling appeals by convicted indigents. Pursuant to that procedure, Anders’ appointed appellate counsel had filed a letter stating that he had concluded that there was “no merit to the appeal,” 386 U. S., at 739-740. Anders, in response, sought new counsel; the State Court of Appeal denied the request, and Anders filed a pro se appellate brief. That court then issued an opinion that reviewed the four claims in his pro se brief and affirmed, finding no error (or no prejudicial error). People v. Anders, 167 Cal. App. 2d 65, 333 P. 2d *270854 (1959). Anders thereafter sought a writ of habeas corpus from the State Court of Appeal, which denied relief, explaining that it had again reviewed the record and had found the appeal to be “ 'without merit/ ” Anders, 386 U. S., at 740 (quoting unreported memorandum opinion).
We held that “California’s action does not comport with fair procedure and lacks that equality that is required by the Fourteenth Amendment.” Id., at 741. We placed the case within a line of precedent beginning with Griffin v. Illinois, 351 U. S. 12 (1956), and continuing with Douglas, supra, that imposed constitutional constraints on States when they choose to create appellate review.5 In finding the California procedure to have breached these constraints, we compared it to other procedures we had found invalid and to statutory requirements in the federal courts governing appeals by indigents with appointed counsel. Anders, supra, at 741-743. We relied in particular on Ellis v. United States, 356 U. S. 674 (1958) (per curiam), a case involving federal statutory requirements, and quoted the following passage from it:
“'If counsel is convinced, after conscientious investigation, that the appeal is frivolous, of course, he may ask to withdraw on that account. If the court is satisfied that counsel has diligently investigated the possible grounds of appeal, and agrees with counsel’s evaluation of the case, then leave to withdraw may be allowed and leave to appeal may be denied.’” Anders, supra, at 741-742 (quoting Ellis, supra, at 675).
In Anders, neither counsel, the state appellate court on direct appeal, nor the state habeas courts had made any finding of frivolity.6 We concluded that a finding that the appeal *271had “no merit” was not adequate, because it did not mean that the appeal was so lacking in prospects as to be “frivolous”: “We cannot say that there was a finding of frivolity by either of the California courts or that counsel acted in any greater capacity than merely as amicus curiae which was condemned in Ellis” 386 U.S., at 743.
Having rejected the California procedure, we proceeded, in a final, separate section, to set out what would be an acceptable procedure for treating frivolous appeals:
“[I]f counsel finds his case to be wholly frivolous, after a conscientious examination of it, he should so advise the court and request permission to withdraw. That request must, however, be accompanied by a brief referring to anything in the record that might arguably support the appeal. A copy of counsel’s brief should be furnished the indigent and time allowed him to raise any points that he chooses; the court — not counsel — then proceeds, after a full examination of all the proceedings, to decide whether the case is wholly frivolous. If it so finds it may grant counsel’s request to withdraw and dismiss the appeal insofar as federal requirements are concerned, or proceed to a decision on the merits, if state law so requires. On the other hand, if it finds any of the legal points arguable on their merits (and therefore not frivolous) it must, prior to decision, afford the indigent the assistance of counsel to argue the appeal.” Id., at 744.
We then concluded by explaining how this procedure would be better than the California one that we had found deficient. Among other things, we thought that it would “induce the court to pursue all the more vigorously its own review because of the ready references not only to the record but also *272to the legal authorities as furnished it by counsel.” Id., at 745.
B
The Ninth Circuit ruled that this final section of Anders, even though unnecessary to our holding in that case, was obligatory upon the States. We disagree. We have never so held; we read our precedents to suggest otherwise; and the Ninth Circuit’s view runs contrary to our established practice of permitting the States, within the broad bounds of the Constitution, to experiment with solutions to difficult questions of policy.
In McCoy v. Court of Appeals of Wis., Dist. 1, 486 U. S. 429 (1988), we rejected a challenge to Wisconsin’s variation on the Anders procedure. Wisconsin had departed from Anders by requiring Anders briefs to discuss why each issue raised lacked merit. The defendant argued that this rule was contrary to Anders and forced counsel to violate his ethical obligations to his client. We, however, emphasized that the right to appellate representation does not include a right to present frivolous arguments to the court, 486 U. S., at 436, and, similarly, that an attorney is “under an ethical obligation to refuse to prosecute a frivolous appeal,” ibid, (footnote omitted). Anders, we explained, merely aims to “assure the court that the indigent defendant’s constitutional rights have not been violated.” 486 U. S., at 442. Because the Wisconsin procedure adequately provided such assurance, we found no constitutional violation, notwithstanding its variance from Anders. See 486 U. S., at 442-444. We did, in McCoy, describe the procedure at issue as going “one step further” than Anders, McCoy, supra, at 442, thus suggesting that An-ders might set a mandatory minimum, but we think this description of the Wisconsin procedure questionable, since it provided less effective advocacy for an indigent — in at least one respect — than does the Anders procedure. The Wisconsin procedure, by providing for one-sided briefing by counsel against his own client’s best claims, probably made a court *273more likely to rule against the indigent than if the court had simply received an Anders brief.
In Pennsylvania v. Finley, 481 U. S. 551 (1987), we explained that the Anders procedure is not “an independent constitutional command,” but rather is just “a prophylactic framework” that we established to vindicate the constitutional right to appellate counsel announced in Douglas. 481 U. S., at 555. We did not say that our Anders procedure was the only prophylactic framework that could adequately vindicate this right; instead, by making clear that the Constitution itself does not compel the Anders procedure, we suggested otherwise. Similarly, in Penson v. Ohio, 488 U. S. 75 (1988), we described Anders as simply erecting “safeguards.” 488 U. S., at 80.
It is true that in Penson we used some language suggesting that Anders is mandatory upon the States, see 488 U. S., at 80-82, but that language was not necessary to the decision we reached. We had no reason in Penson to determine whether the Anders procedure was mandatory, because the procedure at issue clearly failed under Douglas, see infra, at 280. Further, counsel’s action in Penson was closely analogous to the action of counsel that we found invalid in Anders, see Penson, supra, at 77-78, so there was no need to rely on the Anders procedure, as opposed to just the Anders holding, to find counsel’s action improper. See 488 U. S., at 77 (“The question presented by this case is remarkably similar [to the one presented in Anders] and therefore requires a similar answer”).
Finally, any view of the procedure we described in the last section of Anders that converted it from a suggestion into a straitjaeket would contravene our established practice, rooted in federalism, of allowing the States wide discretion, subject to the minimum requirements of the Fourteenth Amendment, to experiment with solutions to difficult problems of policy. In Griffin v. Illinois, 351 U. S. 12 (1956), which we invoked as the foundational ease for our holding *274in Anders, see Anders, 386 U. S., at 741, we expressly disclaimed any pretensions to rulemaking authority for the States in the area of indigent criminal appeals. We imposed no broad rule or procedure but merely held unconstitutional Illinois’ requirement that indigents pay a fee to receive a trial transcript that was essential for bringing an appeal. Justice Frankfurter, who provided the necessary fifth vote for the holding in Griffin, emphasized that it was not for this Court “to tell Illinois what means are open to the indigent and must be chosen. Illinois may prescribe any means that are within the wide area of its constitutional discretion” and “may protect itself so that frivolous appeals are not subsidized and public moneys not needlessly spent.” 351 U. S., at 24 (opinion concurring in judgment). He added that while a State could not “bolt the door to equal justice,” it also was not obliged to “support a wasteful abuse of the appellate process.” Ibid. The Griffin plurality shared this view, explaining that the Court was not holding “that Illinois must purchase a stenographer’s transcript in every case where a defendant cannot buy it. The Supreme Court [of Illinois] may find other means of affording adequate and effective appellate review to indigent defendants.” Id., at 20.
In a related context, we stated this basic principle of federalism in the very Term in which we decided Anders. We emphatically reaffirmed that the Constitution “has never been thought [to] establish this Court as a rule-making organ for the promulgation of state rules of criminal procedure.” Spencer v. Texas, 385 U. S. 554, 564 (1967) (citing, inter alia, Griffin, supra). Accord, Medina v. California, 505 U. S. 437, 443-444, 447-448 (1992). Justice Stewart, concurring in Spencer, explained further:
“If the Constitution gave me a roving commission to impose upon the criminal courts of Texas my own notions of enlightened policy, I would not join the Court’s opinion. . . . [But] [t]he question is whether those procedures fall below the minimum level the Fourteenth *275Amendment will tolerate. Upon that question, I am constrained to join the opinion and judgment of the Court.” 385 U. S., at 569.
We have continued to reiterate this principle in recent years. See Finley, 481 U. S., at 559 (refusing to accept the premise that “when a State chooses to offer help to those seeking relief from convictions, the Federal Constitution dictates the exact form such assistance must assume”); ibid, (explaining that States have “substantial discretion to develop and implement programs to aid prisoners seeking to secure post-conviction review”); Murray v. Giarratano, 492 U. S. 1, 13 (1989) (O’Connor, J., concurring) (“[N]or does it seem to me that the Constitution requires the States to follow any particular federal model in [posteonviction] proceedings. . . . States [have] considerable discretion”); id., at 14 (Kennedy, J., concurring in judgment) (“[J]udicial imposition of a categorical remedy . . . might pretermit other responsible solutions being considered in Congress and state legislatures”). Although Finley and Murray involved postconviction proceedings (in which there is no constitutional right to counsel) rather than direct appeal, we think, as the language of Griffin suggests, that the principle is the same in both contexts. For in Griffin, as here, there was an underlying constitutional right at issue.
In short, it is more in keeping with our status as a court, and particularly with our status as a court in a federal system, to avoid imposing a single solution on the States from the top down. We should, and do, evaluate state procedures one at a time, as they come before us, see Murray, supra, at 14, while leaving “the more challenging task of crafting appropriate procedures ... to the laboratory of the States in the first instance,” Cruzan v. Director, Mo. Dept. of Health, 497 U. S. 261, 292 (1990) (O’Connor, J., concurring) (citation and internal quotation marks omitted). We will not cavalierly “imped[e] the States’ ability to serve as laboratories for testing solutions to novel legal problems.” Arizona v. *276Evans, 514 U. S. 1, 24 (1995) (Ginsburg, J., dissenting). Accordingly, we hold that the Anders procedure is merely one method of satisfying the requirements of the Constitution for indigent criminal appeals. States may — and, we are confident, will — craft procedures that, in terms of policy, are superior to, or at least as good as, that in Anders. The Constitution erects no barrier to their doing so.7
III
Having determined that California’s Wende procedure is not unconstitutional merely because it diverges from the An-ders procedure, we turn to consider the Wende procedure on its own merits. We think it clear that California’s system does not violate the Fourteenth Amendment, for it provides “a criminal appellant pursuing a first appeal as of right [the] minimum safeguards necessary to make that appeal ‘adequate and effective,’” Evitts v. Lucey, 469 U.S. 387, 392 (1985) (quoting Griffin, 351 U. S., at 20 (plurality opinion)).
A
As we have admitted on numerous occasions, “ ‘[t]he precise rationale for the Griffin and Douglas lines of cases has never been explicitly stated, some support being derived from the Equal Protection Clause of the Fourteenth Amendment and some from the Due Process Clause of that Amendment.’ ” Evitts, supra, at 403 (quoting Ross v. Moffitt, 417 U. S. 600, 608-609 (1974) (footnote omitted)). But our case law reveals that, as a practical matter, the two Clauses largely converge to require that a State’s procedure “affor[d] adequate and effective appellate review to indigent defendants,” Griffin, 351 U. S., at 20 (plurality opinion). A State’s procedure provides such review so long as it reasonably en*277sures that an indigent’s appeal will be resolved in a way that is related to the merit of that appeal.8 See id., at 17-18 (plurality opinion) (state law regulating indigents’ appeals bore “no rational relationship to a defendant’s guilt or innocence”); id., at 22 (Frankfurter, J., concurring in judgment) (law imposed “differentiations ... that have no relation to a rational policy of criminal appeal”); Douglas, 372 U. S., at 357 (decision of first appeal “without benefit of counsel,... no matter how meritorious [an indigent’s] case may turn out to be,” discriminates between rich and poor rather than between “possibly good and obviously bad cases” (internal quotation marks omitted)); Rinaldi v. Yeager, 384 U. S. 305, 310 (1966) (state appellate system must be “free of unreasoned distinctions”); Evitts, supra, at 404 (law in Griffin “decided the appeal in a way that was arbitrary with respect to the issues involved”). Compare Finley, supra, at 556 (“The equal protection guarantee ... only... assure[s] the indigent defendant an adequate opportunity to present his claims fairly in the context of the State’s appellate process” (quoting Ross, supra, at 616)), with Evitts, supra, at 405 (“[D]ue process... [requires] States ... to offer eaeh defendant a fair opportunity to obtain an adjudication on the merits of his appeal” (discussing Griffin and Douglas))9
In determining whether a particular state procedure satisfies this standard, it is important to focus on the underlying goals that the procedure should serve — to ensure that those indigents whose appeals are not frivolous receive the counsel and merits brief required by Douglas, and also to enable the *278State to “protect itself so that frivolous appeals are not subsidized and public moneys not needlessly spent,” Griffin, supra, at 24 (Frankfurter, J., concurring in judgment). For although, under Douglas, indigents generally have a right to counsel on a first appeal as of right, it is equally true that this right does not include the right to bring a frivolous appeal and, concomitantly, does not include the right to counsel for bringing a frivolous appeal.10 See McCoy, 486 U. S., at 436-438; Douglas, supra, at 357; see also United States v. Cronic, 466 U. S. 648, 656, n. 19 (1984) (“Of course, the Sixth Amendment does not require that [trial] counsel do what is impossible or unethical”); cf. Nix v. Whiteside, 475 U. S. 157, 175 (1986) (no violation of Sixth Amendment right to the effective assistance of counsel when trial counsel refuses to violate ethical duty not to assist his client in presenting perjured testimony). To put the point differently, an indigent defendant who has his appeal dismissed because it is frivolous has not been deprived of “a fair opportunity” to bring his appeal, Evitts, supra, at 405; see Finley, 481 U. S., at 556, for fairness does not require either counsel or a full appeal once it is properly determined that an appeal is frivolous. The obvious goal of Anders was to prevent this limitation on the right to appellate counsel from swallowing the right itself, see Penson, 488 U. S., at 83-84; McCoy, supra, at 444, and we do not retreat from that goal today.
B
We think the Wende procedure reasonably ensures that an indigent’s appeal will be resolved in a way that is related to *279the merit of that appeal. Whatever its strengths or weaknesses as a matter of policy, we cannot say that it fails to afford indigents the adequate and effective appellate review that the Fourteenth Amendment requires. A comparison of the Wende procedure to the procedures evaluated in our chief cases in this area makes this evident.
The Wende procedure is undoubtedly far better than those procedures we have found inadequate. Anders itself, in disapproving the former California procedure, chiefly relied on three precedents: Ellis v. United States, 356 U. S. 674 (1958) (per curiam), Eskridge v. Washington Bd. of Prison Terms and Paroles, 357 U. S. 214 (1958) (per curiam), and Lane v. Brown, 372 U. S. 477 (1963). See Anders, 386 U. S., at 741-743. Although we did. not, in Anders, explain in detail why the California procedure was inadequate under each of these precedents, our particularly heavy reliance on Ellis makes clear that a significant factor was that the old California procedure did not require either counsel or the court to determine that the appeal was frivolous; instead, the procedure required only that they determine that the defendant was unlikely to prevail on appeal. Compare Anders, supra, at 741-742 (“ ‘If counsel is convinced, after conscientious investigation, that the appeal is frivolous, of course, he may ask to withdraw .... If the court . . . agrees with counsel’s evaluation of the case, then leave to withdraw may be allowed and leave to appeal may be denied’ ” (quoting Ellis, supra, at 675)), with Anders, supra, at 743 (“We cannot say that there was a finding of frivolity”). See also McCoy, supra, at 437 (quoting same passage from Ellis that we quoted in Anders). This problem also appears to have been one of the flaws in the procedures at issue in Eskridge and Lane. The former involved a finding only that there had been “‘no grave or prejudicial errors’” at trial, Anders, supra, at 742 (quoting Eskridge, supra, at 215), and the latter, a finding only that the appeal “‘would be unsuccessful,’ ” Anders, supra, at 743 (quoting Lane, supra, at 482). Wende, *280by contrast, requires both counsel and the court to find the appeal to be lacking in arguable issues, which is to say, frivolous. See 25 Cal. 3d, at 439, 441-442, 600 P. 2d, at 1073, 1075; see id., at 441, 600 P. 2d, at 1074 (reading Anders as finding old California procedure deficient largely “because the court itself did not make an express finding that the appeal was frivolous”).
An additional problem with the old California procedure was that it apparently permitted an appellate court to allow counsel to withdraw and thereafter to decide the appeal without appointing new counsel. See Anders, supra, at 740, n. 2. We resolved any doubt on this point in Penson, where we struck down a procedure that allowed counsel to withdraw before the court had determined whether counsel’s evaluation of the ease was accurate, 488 U. S., at 82-83, and, in addition, allowed a court to decide the appeal without counsel even if the court found arguable issues, id., at 83 (stating that this latter flaw was the “[m]ost significan[t]” one). Thus, the Penson procedure permitted a basic violation of the Douglas right to have counsel until a case is determined to be frivolous and to receive a merits brief for a nonfrivolous appeal. See 488 U. S., at 88 (“[I]t is important to emphasize that the denial of counsel in this case left petitioner completely without representation during the appellate court’s actual decisional process”); ibid, (defendant was “entirely without the assistance of counsel on appeal”). Cf. McCoy, supra, at 430-431, n. 1 (approving procedure under which appellate court first finds appeal to be frivolous and affirms, then relieves counsel). Under Wende, by contrast, Douglas violations do not occur, both because counsel does not move to withdraw and because the court orders briefing if it finds arguable issues. See Wende, supra, at 442, n. 3, 600 P. 2d, at 1075, n. 3; see also, e. g., Rowland, 75 Cal. App. 3d, at 61-62, 88 Cal. Rptr. 2d, at 900-901.
In Anders, we also disapproved the old California procedure because we thought that a one-paragraph letter from *281counsel stating only his “bare conclusion” that the appeal had no merit was insufficient. 386 U. S., at 742. It is unclear from our opinion in Anders how much our objection on this point was severable from our objection to the lack of a finding of frivolity, because we immediately followed our description of counsel's “no merit” letter with a discussion of Ellis, Eskridge, and Lane, and the lack of such a finding. See 386 U. S., at 742-743. In any event, the Wende brief provides more than a one-paragraph “bare conclusion.” Counsel’s summary of the case’s procedural and factual history, with citations of the record, both ensures that a trained legal eye has searched the record for arguable issues and assists the reviewing court in its own evaluation of the ease.
Finally, an additional flaw with the procedures in Eskridge and Lane was that there was only one tier of review — by the trial judge in Eskridge (who understandably had little incentive to find any error warranting an appeal) and by the public defender in Lane. See Anders, supra, at 742-743. The procedure in Douglas itself was, in part, flawed for the same reason. See 372 U. S., at 354-355. The Wende procedure, of course, does not suffer from this flaw, for it provides at least two tiers of review.
Not only does the Wende procedure far exceed those procedures that we have found invalid, but it is also at least comparable to those procedures that we have approved. Turning first to the procedure we set out in the final section of Anders, we note that it has, from the beginning, faced “ 'consistent and severe criticism.’” In re Sade C., 13 Cal. 4th 952, 979, n. 7, 920 P. 2d 716, 731, n. 7 (1996) (quoting Note, 67 Texas L. Rev. 181, 212 (1988)). One of the most consistent criticisms, one with which we wrestled in McCoy, is that An-ders is in some tension both with counsel’s ethical duty as an officer of the court (which requires him not to present frivolous arguments) and also with his duty to further his client’s interests (which might not permit counsel to characterize his *282client’s claims as frivolous).11 California, through the Wende procedure, has made a good-faith effort to mitigate this problem by not requiring the Wende brief to raise legal issues and by not requiring counsel to explicitly describe the case as frivolous. See Wende, 25 Cal. 3d, at 441-442, 600 P. 2d, at 1074-1075.
Another criticism of the Anders procedure has been that it is incoherent and thus impossible to follow. Those making this criticism point to our language in Anders suggesting that an appeal could be both “wholly frivolous” and at the same time contain arguable issues, even though we also said that an issue that was arguable was “therefore not frivolous.” Anders, supra, at 744.12 In other words, the Anders procedure appears to adopt gradations of frivolity and to use two different meanings for the phrase “arguable issue.” The Wende procedure attempts to resolve this problem as well, by drawing the line at frivolity and by defining arguable issues as those that are not frivolous.13
*283Finally, the Wende procedure appears to be, in some ways, better than the one we approved in McCoy and, in other ways, worse. On balance, we cannot say that the latter, assuming, arguendo, that they outweigh the former, do so sufficiently to make the Wende procedure unconstitutional. The Wisconsin procedure we evaluated in McCoy, which required counsel filing an Anders brief to explain why the issues he raised in his brief lacked merit, arguably exacerbated the ethical problem already present in the Anders procedure. The Wende procedure, as we have explained, attempts to mitigate that problem. Further, it appears that in the McCoy scheme counsel discussed — and the appellate court reviewed — only the parts of the record cited by counsel in support of the “arguable” issues he raised. See 486 U. S., at 440, 442. The Wende procedure, by contrast, requires a more thorough treatment of the record by both counsel and court. See 25 Cal. 3d, at 440-441, 600 P. 2d, at 1074-1075; id., at 445, 600 P. 2d, at 1077 (Clark, J., concurring in judg*284ment and dissenting in part). On the other hand, the McCoy procedure, unlike the Wende procedure, does assist the reviewing court by directing it to particular legal issues; as to those issues, this is presumably a good thing. But it is also possible that bad judgment by the attorney in selecting the issues to raise might divert the court’s attention from more meritorious, unmentioned, issues. This criticism is, of course, equally applicable to the Anders procedure. Moreover, as to the issues that counsel does raise in a McCoy brief, the one-sided briefing on why those issues are frivolous may predispose the court to reach the same conclusion. The Wende procedure reduces these risks, by omitting from the brief signals that may subtly undermine the independence and thoroughness of the second review of an indigent’s case.
Our purpose is not to resolve any of these arguments. The Constitution does not resolve them, nor does it require us to do so. “We address not what is prudent or appropriate, but only what is constitutionally compelled.” Cronic, 466 U. S., at 665, n. 38. It is enough to say that the Wende procedure, like the Anders and McCoy procedures, and unlike the ones in Ellis, Eskridge, Lane, Douglas, and Penson, affords adequate and effective appellate review for criminal indigents. Thus, there was no constitutional violation in this case simply because the Wende procedure was used.
IV
Since Robbins’ counsel complied with a valid procedure for determining when an indigent’s direct appeal is frivolous, we reverse the Ninth Circuit’s judgment that the Wende procedure fails adequately to serve the constitutional principles we identified in Anders. But our reversal does not necessarily mean that Robbins’ claim that his appellate counsel rendered constitutionally ineffective assistance fails. For it may be, as Robbins argues, that his appeal was not frivolous and that he was thus entitled to a merits brief rather than to a Wende brief. Indeed, both the District Court and the *285Ninth Circuit found that there were two arguable issues on direct appeal. The meaning of “arguable issue” as used in the opinions below, however, is far from clear. The courts below most likely used the phrase in the unusual way that we used it in Anders — an issue arguably supporting the appeal even though the appeal was wholly frivolous. See 152 F. 3d, at 1067 (discussing arguable issues in context of requirements of Anders); App. 48 (District Court opinion) (same). Such an issue does not warrant a merits brief. But the courts below may have used the term to signify issues that were “arguable” in the more normal sense of being non-frivolous and thus warranting a merits brief. See id., at 49, and n. 3 (District Court, considering arguable issues to determine “whether Anders was violated,” but also defining arguable issue as one that counsel could argue “in good faith with some potential for prevailing”). Further, the courts below, in determining whether there were arguable issues, did not address petitioner’s argument that, at least with regard to the adequacy of the prison law library, Robbins waived the issue for appeal by failing to object at trial. Thus, it will be necessary on remand to clarify just how strong these two issues are.
On remand, the proper standard for evaluating Robbins’ claim that appellate counsel was ineffective in neglecting to file a merits brief is that enunciated in Strickland v. Washington, 466 U. S. 668 (1984). See Smith v. Murray, 477 U. S. 527, 535-536 (1986) (applying Strickland to claim of attorney error on appeal). Respondent must first show that his counsel was objectively unreasonable, see Strickland, 466 U. S., at 687-691, in failing to find arguable issues to appeal — that is, that counsel unreasonably failed to discover nonfrivolous issues and to file a merits brief raising them. If Robbins succeeds in such a showing, he then has the burden of demonstrating prejudice. That is, he must show a reasonable probability that, but for his counsel’s unreasonable failure to file a merits brief, he would have prevailed on his appeal. *286See id., at 694 (defendant must show “a reasonable probability that, but for counsel’s unprofessional errors, the result of the proceeding would have been different”).14
The applicability of Strickland’s actual-prejudice prong to Robbins’ claim of ineffective assistance follows from Pen-son, where we distinguished denial of counsel altogether on appeal, which warrants a presumption of prejudice, from mere ineffective assistance of counsel on appeal, which does not. See 488 U. S., at 88-89. The defendant in Penson faced a denial of counsel because, as we have discussed, supra, at 280, not only was an invalid state procedure followed, but that procedure was clearly invalid insofar as it denied the defendant his right to appellate counsel under Douglas, see 488 U. S., at 83, 88. Our holding in Penson was consistent with Strickland itself, where we said that we would presume prejudice when a defendant had suffered an “[ajctual or constructive denial of the assistance of counsel altogether.” 466 U. S., at 692; see also Cronic, supra, at 659, and n. 25. In other words, while we normally apply a “strong presumption of reliability” to judicial proceedings and require a defendant to overcome that presumption, Strickland, supra, at 696, when, as in Penson, there has been a complete denial of counsel, we understandably presume the opposite, see Strickland, supra, at 692.
But where, as here, the defendant has received appellate counsel who has complied with a valid state procedure for determining whether the defendant’s appeal is frivolous, and the State has not at any time left the defendant without counsel on appeal, there is no reason to presume that the defendant has been prejudiced. In Penson, we worried that requiring the defendant to establish prejudice would leave him “without any of the protections afforded by Anders.” *287488 U. S., at 86. Here, by contrast, counsel followed a procedure that is constitutional under Anders and our other precedents in this area, and Robbins therefore received all the procedural protection that the Constitution requires. We thus presume that the result of the proceedings on appeal is reliable, and we require Robbins to prove the presumption incorrect in his particular case. See Strickland, 466 U. S., at 694.
Further, the ineffective-assistance claim that Robbins presses does not fall within any of the three categories of cases, described in Strickland, in which we presume prejudice rather than require a defendant to demonstrate it. First, as noted, we presume prejudice in a case of denial of counsel. Second, “various kinds of state interference with counsel’s assistance” can warrant a presumption of prejudice. Id., at 692; see Cronic, 466 U. S., at 659, and n. 25. Third, “prejudice is presumed when counsel is burdened by an actual conflict of interest,” Strickland, 466 U. S., at 692, although in such a ease we do require the defendant to show that the conflict adversely affected his counsel’s performance, ibid. None of these three categories applies to a case such as Robbins’. Nor does the policy reason that we offered in Strickland for the first two categories apply here, for it is not the case that, if an attorney unreasonably chooses to follow a procedure such as Anders or Wende instead of filing a merits brief, prejudice “is so likely that case-by-case inquiry into prejudice is not worth the cost.” 466 U. S., at 692; see Cronic, supra, at 658.15 On the contrary, in most eases in which a defendant’s appeal has been found, pursuant to a valid state procedure, to be frivolous, it will in fact be frivolous.
It is no harder for a court to apply Strickland in this area than it is when a defendant claims that he received ineffec*288tive assistance of appellate counsel because his counsel, although filing a merits brief, failed to raise a particular claim. It will likely be easier to do so. In Jones v. Barnes, 463 U. S. 745 (1983), we held that appellate counsel who files a merits brief need not (and should not) raise every nonfrivo-lous claim, but rather may select from among them in order to maximize the likelihood of success on appeal. Notwithstanding Barnes, it is still possible to bring a Strickland claim based on counsel’s failure to raise a particular claim, but it is difficult to demonstrate that counsel was incompetent. See, e. g., Gray v. Greer, 800 F. 2d 644, 646 (CA7 1986) (“Generally, only when ignored issues are clearly stronger than those presented, will the presumption of effective assistance of counsel be overcome”). With a claim that counsel erroneously failed to file a merits brief, it will be easier for a defendant-appellant to satisfy the first part of the Strickland test, for it is only necessary for him to show that a reasonably competent attorney would have found one non-frivolous issue warranting a merits brief, rather than showing that a particular nonfrivolous issue was clearly stronger than issues that counsel did present. In both eases, however, the prejudice analysis will be the same.16
*289In sum, Robbins must satisfy both prongs of the Strickland test in order to prevail on his claim of ineffective assistance of appellate counsel. The judgment of the Court of Appeals is reversed, and the ease is remanded for further proceedings consistent with this opinion.
It is so ordered.
In addition to this double review and double determination of frivolity, California affords a third layer of review, through the California Appellate Projects, described in a recent opinion by the California Court of Appeal for the First District:
“[The appellate projects] are under contract to the court; their contractual duties include review of the records to assist court-appointed counsel in identifying issues to brief. If the court-appointed counsel can find no meritorious issues to raise and decides to file a Wende brief, an appellate project staff attorney reviews the record again to determine whether a Wende brief is appropriate. Thus, by the time the Wende brief is filed in the Court of Appeal, the record in the ease has been reviewed both by the court-appointed counsel (who is presumably well qualified to handle the case) and by an experienced attorney on the staff of [the appellate project].” People v. Hackett, 36 Cal. App. 4th 1297, 1311, 43 Cal. Rptr. 2d 219, 228 (1995).
Before filing his Wende brief, counsel consulted with the California Appellate Project for the Second District Court of Appeal and received its permission to file such a brief App. 48.
The Antiterrorism and Effective Death Penalty Act of 1996, 110 Stat. 1214, which amended §2254 and related provisions, does not apply to respondent’s habeas petition, since he filed his petition before that Act’s effective date of April 24, 1996. See Lindh v. Murphy, 521 U. S. 320 (1997).
In subsequent eases, the Ninth Circuit has reiterated its view that the Wende procedure is unconstitutional because it differs from the Anders procedure. See Delgado v. Lewis, 181 F. 3d 1087, 1090, 1093, stay granted pending disposition of pet. for cert., 527 U. S. 1066 (1999); Davis v. Kramer, 167 F. 3d 494, 496, 497-498 (1999), cert. pending, No. 98-1427.
The Constitution does not, however, require States to create appellate review in the first place. See, e. g., Ross v. Moffitt, 417 U. S. 600, 606 (1974) (citing McKane v. Durston, 153 U. S. 684, 687 (1894)).
The same was true in Ellis itself. See Ellis v. United States, 249 F. 2d 478, 480-481 (CADC 1957) (Washington, J., dissenting) (“Counsel... concluded that the rulings of the District Court were not ‘so clearly erro*271neous as to constitute probable error.’ . . . Where, as here, there was a fairly arguable question, counsel should have proceeded to present argument”), vacated and remanded, 356 U. S. 674 (1958) (per curiam).
States have, in fact, already been doing this to some degree. See Warner, Anders in the Fifty States: Some Appellants’ Equal Protection is More Equal Than Others’, 23 Fla. St. U. L. Rev. 625, 642-662 (1996); Arizona v. Clark, 196 Ariz. 530, 536-539, 2 P. 3d 89, 95-98 (App. 1999).
Of course, no procedure can eliminate all risk of error. E. g., Walters v. National Assn. of Radiation Survivors, 473 U. S. 305, 320-321 (1985).
Although we have said that an indigent must receive “substantial equality” compared to the legal assistance that a defendant with paid counsel would receive, McCoy v. Court of Appeals of Wis., Dist. 1, 486 U. S. 429, 438 (1988), we have also emphasized that “[albsolute equality is not required; lines can be and are drawn and we often sustain them,” Douglas v. California, 372 U. S. 353, 357 (1963).
This distinction gives meaning to our previous emphasis on an indigent appellant’s right to “advocacy.” Although an indigent whose appeal is frivolous has no right to have an advocate make his case to the appellate court, such an indigent does, in all cases, have the right to have an attorney, zealous for the indigent’s interests, evaluate his case and attempt to discern nonfrivolous arguments. See Ellis, 356 U. S., at 675; Anders v. California, 386 U. S. 738, 741-743 (1967).
As one former public defender has explained, “an attorney confronted with the Anders situation has to do something that the Code of Professional Responsibility describes as unethical; the only choice is as to which canon he or she prefers to violate.” Pengilly, Never Cry Anders: The Ethical Dilemma of Counsel Appointed to Pursue a Frivolous Criminal Appeal, 9 Crim. Justice J. 45, 64 (1986). See also, e. g., Commonwealth v. Moffett, 383 Mass. 201, 206, 418 N. E. 2d 585, 590 (1981) (Anders requires a “Janus-faced approach” by counsel); Hermann, Frivolous Criminal Appeals, 47 N. Y. U. L. Rev. 701, 711 (1972).
Justice Stewart, in his dissent in Anders, was the first to make this criticism of the procedure set out by the Anders majority: “[I]f the record did present any such 'arguable’ issues, the appeal would not be frivolous.” 386 U. S., at 746; see id., at 746, n. See also, e. g., C. Wolfram, Modem Legal Ethics 817 (1986) (“The Anders directives are confusing, if not contradictory”).
See supra, at 279-280. A further criticism of Anders has been that it is unjust. More particularly, critics have claimed that, in setting out the Anders procedure, we were oblivious to the problem of scarce resources (with regard to both counsel and courts) and, as a result, crafted a rule that diverts attention from meritorious appeals of indigents and ensures poor representation for all indigents. See, e. g., Pritchard, Auc*283tioning Justice: Legal and Market Mechanisms for Allocating Criminal Appellate Counsel, 34 Am. Crim. L. Rev. 1161, 1167-1168 (1997) (Anders has created a "tragedy of the commons” that, “far from guaranteeing adequate appellate representation for all criminal defendants, instead ensures that indigent criminal defendants will receive mediocre appellate representation, whether their claims are good or bad” (footnote omitted)); Pritchard, supra, at 1169 (noting Anders’ similar effect on appellate courts); Pritch-ard, supra, at 1162 (“[J]udicial fiat cannot cure scarcity; it merely disguises the symptoms of the disease”); Doherty, Wolf! Wolf! — -The Ramifications of Frivolous Appeals, 59 J. Crim. L., C. & P. S. 1, 2 (1968) (“[T]he people who will suffer the most are the indigent prisoners who have been unjustly convicted; they will languish in prison while lawyers devote time and energy to hopeless causes on a first come-first served basis” (footnote omitted)). We cannot say whether the Wende procedure is better or worse than the Anders procedure in this regard (although we are aware of policy-based arguments that it is worse as to appellate courts, see People v. Williams, 59 Cal. App. 4th 1202, 1205-1206, 69 Cal. Rptr. 2d 690, 692 (1997); Brief for Retired Justice Armand Arabian et al. as Amici Curiae), but it is clear that, to the extent this criticism has merit, our holding today that the Anders procedure is not exclusive will enable States to continue to experiment with solutions to this problem.
The performance component need not be addressed first. "If it is easier to dispose of an ineffectiveness claim on the ground of lack of sufficient prejudice, which we expect will often be so, that course should be followed.” Strickland v. Washington, 466 U. S., at 697.
Moreover, such an error by counsel is neither “easy to identify” (since it is necessary to evaluate a defendant’s ease in order to find the error) nor attributable to the prosecution. See Strickland, supra, at 692.
Federal judges are, of course, fully capable of assessing prejudice in this area, including for the very sorts of daims that Robbins has raised. See, e. g., Duhamel v. Collins, 955 F. 2d 962, 967 (CA5 1992) (defendant not prejudiced by appellate counsel’s failure to challenge sufficiency of the evidence); Banks v. Reynolds, 54 F. 3d 1508, 1515-1516 (CA10 1995) (finding both parts of Strickland test satisfied where appellate counsel failed to raise claim of violation of Brady v. Maryland, 373 U. S. 83 (1963)); Cross v. United States, 893 F. 2d 1287, 1290-1291, 1292 (CA11) (rejecting challenge to appellate counsel’s failure to raise claim of violation of Faretta v. California, 422 U. S. 806 (1975), by determining that there was no prejudice), cert. denied, 498 U. S. 849 (1990). Since Robbins was convicted in state court, we have no occasion to consider whether a per se prejudice approach, in lieu of Strickland’s actual-prejudice requirement, might be appropriate in the context of challenges to federal convictions where counsel was deficient in failing to file a merits brief on direct appeal. See Goeke v. Branch, 514 U. S. 115, 119 (1995) (per curiam) (distinguishing *289rules established pursuant to this Court’s supervisory power to administer federal court system from constitutional rules applicable to States); United States v. Cronic, 466 U. S. 648, 665, n. 38 (1984) (same).