Justice Kennedy
delivered the opinion of the Court.
This case presents a challenge to a statute enacted by Congress to protect minors from exposure to sexually explicit materials on the Internet, the Child Online Protection Act [660]*660(COPA), 112 Stat. 2681-736, codified at 47 U.'S. C. § 231. We must decide whether the Court of Appeals was correct to affirm a ruling by the District Court that enforcement of COPA should be enjoined because the statute likely violates the First Amendment.
In enacting COPA, Congress gave consideration to our earlier decisions on this subject, in particular the decision in Reno v. American Civil Liberties Union, 521 U. S. 844 (1997). For that reason, “the Judiciary must proceed with caution and... with care before invalidating the Act.” Ashcroft v. American Civil Liberties Union, 535 U. S. 564, 592 (2002) (Ashcroft I) (Kennedy, J., concurring in judgment). The imperative of according respect to the Congress, however, does not permit us to depart from well-established First Amendment principles. Instead, we must hold the Government to its constitutional burden of proof.
Content-based prohibitions, enforced by severe criminal penalties, have the constant potential to be a repressive force in the lives and thoughts of a free people. To guard against that threat the Constitution demands that content-based restrictions on speech be presumed invalid, R. A. V. v. St Paul, 505 U. S. 377, 382 (1992), and that the Government bear the burden of showing their constitutionality, United States v. Playboy Entertainment Group, Inc., 529 U. S. 803, 817 (2000). This is true even when Congress twice has attempted to find a constitutional means to restrict, and punish, the speech in question.
This case comes to the Court on certiorari review of an appeal from the decision of the District Court granting a preliminary injunction. The Court of Appeals reviewed the decision of the District Court for abuse of discretion. Under that standard, the Court of Appeals was correct to conclude that the District Court did not abuse its discretion in granting the preliminary injunction. The Government has failed, at this point, to rebut the plaintiffs’ contention that there are plausible, less restrictive alternatives to the statute. Sub[661]*661stantial practical considerations, furthermore, argue in favor of upholding the injunction and allowing the case to proceed to trial. For those reasons, we affirm the decision of the Court of Appeals upholding the preliminary injunction, and we remand the case so that it may be returned to the District Court for trial on the issues presented.
I
A
COPA is the second attempt by Congress to make the Internet safe for minors by criminalizing certain Internet speech. The first attempt was the Communications Decency Act of 1996, Pub. L. 104-104, § 502, 110 Stat. 133, 47 U. S. C. § 223 (1994 ed., Supp. II). The Court held the CDA unconstitutional because it was not narrowly tailored to serve a compelling governmental interest and because less restrictive alternatives were available. Reno, supra.
In response to the Court’s decision in Reno, Congress passed COPA. COPA imposes criminal penalties of a $50,000 fine and six months in prison for the knowing posting, for “commercial purposes,” of World Wide Web content that is “harmful to minors.” § 231(a)(1). Material that is “harmful to minors” is defined as:
“any communication, picture, image, graphic image file, article, recording, writing, or other matter of any kind that is obscene or that—
“(A) the average person, applying contemporary community standards, would find, taking the material as a whole and with respect to minors, is designed to appeal to, or is designed to pander to, the prurient interest;
“(B) depicts, describes, or represents, in a manner patently offensive with respect to minors, an actual or simulated sexual act or sexual contact, an actual or simulated normal or perverted sexual act, or a lewd [662]*662exhibition of the genitals or post-pubescent female breast; and
“(C) taken as a whole, lacks serious literary, artistic, political, or scientific value for minors.” § 231(e)(6).
“Minor[s]” are defined as “any person under 17 years of age.” §231(e)(7). A person acts for “commercial purposes only if such person is engaged in the business of making such communications.” “Engaged in the business,” in turn,
“means that the person who makes a communication, or offers to make a communication, by means of the World Wide Web, that includes any material that is harmful to minors, devotes time, attention, or labor to such activities, as a regular course of such person’s trade or business, with the objective of earning a profit as a result of such activities (although it is not necessary that the person make a profit or that the making or offering to make such communications be the person’s sole or principal business or source of income).” § 231(e)(2).
While the statute labels all speech that falls within these definitions as criminal speech, it also provides an affirmative defense to those who employ specified means to prevent minors from gaining access to the prohibited materials on their Web site. A person may escape conviction under the statute by demonstrating that he
“has restricted access by minors to material that is harmful to minors—
“(A) by requiring use of a credit card, debit account, adult-access code, or adult personal identification number;
“(B) by accepting a digital certificate that verifies age; or
“(C) by any other reasonable measures that are feasible under available technology.” § 231(c)(1).
[663]*663Since the passage of COPA, Congress has enacted additional laws regulating the Internet in an attempt to protect minors. For example, it has enacted a prohibition on misleading Internet domain names, 18 U. S. C. §2252B (2000 ed., Supp. Ill), in order to prevent Web site owners from disguising pornographic Web sites in a way likely to cause uninterested persons to visit them. See Brief for Petitioner 7 (giving, as an example, the Web site “whitehouse.com”). It has also passed a statute creating a “Dot Kids” second-level Internet domain, the content of which is restricted to that which is fit for minors under the age of 13. 47 U. S. C. § 941 (2000 ed., Supp. II).
B
Respondents, Internet content providers and others concerned with protecting the freedom of speech, filed suit in the United States District Court for the Eastern District of Pennsylvania. They sought a preliminary injunction against enforcement of the statute. After considering testimony from witnesses presented by both respondents and the Government, the District Court issued an order granting the preliminary injunction. The court first noted that the statute would place a burden on some protected speech. American Civil Liberties Union v. Reno, 31 F. Supp. 2d 473, 495 (1999). The court then concluded that respondents were likely to prevail on their argument that there were less restrictive alternatives to the statute: “On the record to date, it is not apparent... that [petitioner] can meet its burden to prove that COPA is the least restrictive means available to achieve the goal of restricting the access of minors” to harmful material. Id., at 497. In particular, it noted that “[t]he record before the Court reveals that blocking or filtering technology may be at least as successful as COPA would be in restricting minors’ access to harmful material online without imposing the burden on constitutionally protected speech that COPA imposes on adult users or Web site operators.” Ibid.
[664]*664The Government appealed the District Court’s decision to the United States Court of Appeals for the Third Circuit. The Court of Appeals affirmed the preliminary injunction, but on a different ground. American Civil Liberties Union v. Reno, 217 F. 3d 162, 166 (2000). The court concluded that the “community standards” language in COPA by itself rendered the statute unconstitutionally overbroad. Ibid. We granted certiorari and reversed, holding that the community-standards language did not, standing alone, make the statute unconstitutionally overbroad. Ashcroft I, 535 U. S., at 585. We emphasized, however, that our decision was limited to that narrow issue. Ibid. We remanded the case to the Court of Appeals to reconsider whether the District Court had been correct to grant the preliminary injunction. On remand, the Court of Appeals again affirmed the District Court. 322 F. 3d 240 (2003). The Court of Appeals concluded that the statute was not narrowly tailored to serve a compelling Government interest, was overbroad, and was not the least restrictive means available for the Government to serve the interest of preventing minors from using the Internet to gain access to materials that are harmful to them. Id., at 266-271. The Government once again sought review from this Court, and we again granted certiorari. 540 U. S. 944 (2003).
II
“This Court, like other appellate courts, has always applied the abuse of discretion standard on review of a preliminary injunction.” Walters v. National Assn. of Radiation Survivors, 473 U. S. 305, 336 (1985) (O’Connor, J., concurring) (internal quotation marks omitted). “The grant of appellate jurisdiction under [28 U. S. C.] § 1252 does not give the Court license to depart from established standards of appellate review.” Ibid. If the underlying constitutional question is close, therefore, we should uphold the injunction [665]*665and remand for trial on the merits. Applying this mode of inquiry, we agree with the Court of Appeals that the District Court did not abuse its discretion in entering the preliminary injunction. Our reasoning in support of this conclusion, however, is based on narrower, more specific grounds than the rationale the Court of Appeals adopted. The Court of Appeals, in its opinion affirming the decision of the District Court, construed a number of terms in the statute, and held that COPA, so construed, was unconstitutional. None of those constructions of statutory terminology, however, were relied on by or necessary to the conclusions of the District Court. Instead, the District Court concluded only that the statute was likely to burden some speech that is protected for adults, 31 F. Supp. 2d, at 495, which petitioner does not dispute. As to the definitional disputes, the District Court concluded only that respondents’ interpretation was “not unreasonable,” and relied on their interpretation only to conclude that respondents had standing to challenge the statute, id., at 481, which, again, petitioner does not dispute. Because we affirm the District Court’s decision to grant the preliminary injunction for the reasons relied on by the District Court, we decline to consider the correctness of the other arguments relied on by the Court of Appeals.
The District Court, in deciding to grant the preliminary injunction, concentrated primarily on the argument that there are plausible, less restrictive alternatives to COPA. A statute that “effectively suppresses a large amount of speech that adults have a constitutional right to receive and to address to one another ... is unacceptable if less restrictive alternatives would be at léast as effective in achieving the legitimate purpose that the statute was enacted to serve.” Reno, 521 U. S., at 874. When plaintiffs challenge a content-based speech restriction, the burden is on the Government to prove that the proposed alternatives will not be as effective as the challenged statute. Ibid.
[666]*666In considering this question, a court assumes that certain protected speech may be regulated, and then asks what is the least restrictive alternative that can be used to achieve that goal. The purpose of the test is not to consider whether the challenged restriction has some effect in achieving Congress’ goal, regardless of the restriction it imposes. The purpose of the test is to ensure that speech is restricted no further than necessary to achieve the goal, for it is important to ensure that legitimate speech is not chilled or punished. For that reason, the test does not begin with the status quo of existing regulations, then ask whether the challenged restriction has some additional ability to achieve Congress’ legitimate interest. Any restriction on speech could be justified under that analysis. Instead, the court should ask whether the challenged regulation is the least restrictive means among available, effective alternatives.
In deciding whether to grant a preliminary injunction, a district court must consider whether the plaintiffs have demonstrated that they are likely to prevail on the merits. See, e. g., Doran v. Salem Inn, Inc., 422 U. S. 922, 931 (1975). (The court also considers whether the plaintiff has shown irreparable injury, see ibid., but the parties in this case do not contest the correctness of the District Court’s conclusion that a likelihood of irreparable injury had been established. See 31 F. Supp. 2d, at 497-498.) As the Government bears the burden of proof on the ultimate question of COPA’s constitutionality, respondents must be deemed likely to prevail unless the Government has shown that respondents’ proposed less restrictive alternatives are less effective than COPA. Applying that analysis, the District Court concluded that respondents were likely to prevail. Id., at 496497. That conclusion was not an abuse of discretion, because on this record there are a number of plausible, less restrictive alternatives to the statute.
' The primary alternative considered by the District Court was blocking and filtering software. Blocking and filtering [667]*667software is an alternative that is less restrictive than COPA, and, in addition, likely more effective as a means of restricting children’s access to materials harmful to them. The District Court, in granting the preliminary injunction, did so primarily because the plaintiffs had proposed that filters are a less restrictive alternative to COPA and the Government had not shown it would be likely to disprove the plaintiffs’ contention at trial. Ibid.
Filters are less restrictive than COPA. They impose selective restrictions on speech at the receiving end, not universal restrictions at the source. Under a filtering regime, adults without children may gain access to speech they have a right to see without having to identify themselves or provide their credit card information. Even adults with children may obtain access to the same speech on the same terms simply by turning off the filter on their home computers. Above all, promoting the use of filters does not condemn as criminal any category of speech, and so the potential chilling effect is eliminated, or at least much diminished. All of these things are true, moreover, regardless of how broadly or narrowly the definitions in COPA are construed.
Filters also may well be more effective than COPA. First, a filter can prevent minors from seeing all pornography, not just pornography posted to the Web from America. The District Court noted in its factfindings that one witness estimated that 40% of harmful-to-minors content comes from overseas. Id., at 484. COPA does not prevent minors from having access to those foreign harmful materials. That alone makes it possible that filtering software might be more effective in serving Congress’ goals. Effectiveness is likely to diminish even further if COPA is upheld, because the providers of the materials that would be covered by the statute simply can move their operations overseas. It is not an answer to say that COPA reaches some amount of materials that are harmful to minors; the question is whether it would reach more of them than less restrictive alternatives. In [668]*668addition, the District Court found that verification systems may be subject to evasion and circumvention, for example, by minors who have their own credit cards. See id., at 484, 496-497. Finally, filters also may be more effective because they can be applied to all forms of Internet communication, including e-mail, not just communications available via the World Wide Web.
That filtering software may well be more effective than COPA is confirmed by the findings of the Commission on Child Online Protection, a blue-ribbon Commission created by Congress in COPA itself. Congress directed the Commission to evaluate the relative merits of different means of restricting minors’ ability to gain access to harmful materials on the Internet. Note following 47 U. S. C. § 231. It unambiguously found, that filters are more effective than age-verification requirements. See Commission on Child Online Protection (COPA), Report to Congress 19-21, 23-25, 27 (Oct. 20, 2000) (assigning a score for “Effectiveness” of 7.4 for server-based filters and 6.5 for client-based filters, as compared to 5.9 for independent adult-IB verification, and 5.5 for credit card verification). Thus, not only has the Government failed to eárry its burden of showing the District Court that the proposed alternative is less effective, but also a Government Commission appointed to consider the question has concluded just the opposite. That finding supports our conclusion that the District Court did not abuse its discretion in enjoining the statute.
Filtering software, of course, is not a perfect solution to the problem of children gaining access to harmful-to-minors materials. It may block some materials that are not harmful to minors and fail to catch some that are. See 31F. Supp. 2d, at 492. Whatever the deficiencies of filters, however, the Government failed to introduce specific evidence proving that existing technologies are less effective than the restrictions in COPA. The District Court made a specific factfind-ing that “[n]o evidence was presented to the Court as to the [669]*669percentage of time that blocking and filtering technology is over- or underinclusive.” Ibid. In the absence of a showing as to the relative effectiveness of COPA and the alternatives proposed by respondents, it was not an abuse of discretion for the District Court to grant the preliminary injunction. The Government’s burden is not merely to show that a proposed less restrictive alternative has some flaws; its burden is to show that it is less effective. Reno, 521 U. S., at 874. It is not enough for the Government to show that COPA has some effect. Nor do respondents bear a burden to introduce, or offer to introduce, evidence that their proposed alternatives are more effective. The Government has the burden to show they are less so. The Government having failed to carry its burden, it was not an abuse of discretion for the District Court to grant the preliminary injunction.
One argument to the contrary is worth mentioning — the argument that filtering software is not an available alternative because Congress may not require it to be used. That argument carries little weight, because Congress undoubtedly may act to encourage the use of filters. We have held that Congress can give strong incentives to schools and libraries to use them. United States v. American Library Assn., Inc., 539 U. S. 194 (2003). It could also take steps to promote their development by industry, and their use by parents. It is incorrect, for that reason, to say that filters are part of the current regulatory status quo. The need for parental cooperation does not automatically disqualify a proposed less restrictive alternative. Playboy Entertainment Group, 529 U. S., at 824 (“A court should not assume a plausible, less restrictive alternative would be ineffective; and a court should not presume parents, given full information, will fail to act”). In enacting COPA, Congress said its goal was to prevent the “widespread availability of the Internet” from providing “opportunities for minors to access materials through the World Wide Web in a manner that can frustrate [670]*670parental supervision or control.” Congressional Findings, note following 47 U. S. C. §231 (quoting Pub. L. 105-277, Tit. XIV, § 1402(1), 112 Stat. 2681-736). COPA presumes that parents lack the ability, not the will, to monitor what then-children see. By enacting programs to promote use of filtering software, Congress could , give parents that ability without subjecting protected speech to severe penalties.
The closest precedent on the general point is our decision in Playboy Entertainment Group. Playboy Entertainment Group, like this case, involved a content-based restriction designed to protect minors from viewing harmful materials. The choice was between a blanket speech restriction and a more specific technological solution that was available to parents who chose to implement it. 529 U. S., at 825. Absent a showing that the proposed less restrictive alternative would not be as effective, we concluded, the more restrictive option preferred by Congress could not survive strict scrutiny. Id., at 826 (reversing because “[t]he record is silent as to the comparative effectiveness of the two alternatives”). In the instant case, too, the Government has failed to show, at this point, that the proposed less restrictive alternative will be less effective. The reasoning of Playboy Entertainment Group and the holdings and force of our precedents require us to affirm the preliminary injunction. To do otherwise would be to do less than the First Amendment commands. “The ‘starch’ in our constitutional standards cannot be sacrificed to accommodate the enforcement choices of the Government.” Id., at 830 (Thomas, J., concurring).
There are also important practical reasons to let the injunction stand pending a full trial on the merits. First, the potential harms from reversing the injunction outweigh those of leaving it in place by mistake. Where a prosecution is a likely possibility, yet only an affirmative defense is available, speakers may self-censor rather than risk the perils of [671]*671trial. There is a potential for extraordinary harm and a serious chill upon protected speech. Cf. id., at 817 (“Error in marking that line exacts an extraordinary cost”). The harm done from letting the injunction stand pending a. trial on the merits, in contrast, will not be extensive. No prosecutions have yet been undertaken under the law, so none will be disrupted if the injunction stands. Further, if the injunction is upheld, the Government in the interim can enforce obscenity laws already on the books.
Second, there are substantial factual disputes remaining in the case. As mentioned above, there is a serious gap in the evidence as to the effectiveness of filtering software. See supra, at 668. For us to assume, without proof, that filters are less effective than COPA would usurp the District Court’s factfinding role. By allowing the preliminary injunction to stand and remanding for trial, we require the Government to shoulder its full constitutional burden of proof respecting the less restrictive alternative argument, rather than excuse it from doing so.
Third, and on a related point, the factual record does not reflect current technological reality — a serious flaw in any case involving the Internet. The technology of the Internet evolves at a rapid pace. Yet the factfindings of the District Court were entered in February 1999, over five years ago. Since then, certain facts about the Internet are known to have changed. Compare, e. g., 31 F. Supp. 2d, at 481 (36.7 million Internet hosts as of July 1998), with Internet Systems Consortium, Internet Domain Survey, Jan. 2004, http:// www.isc.org/index.pl7/ops/ds (as visited June 22, 2004, and available in Clerk of Court’s case file) (233.1 million hosts as of Jan. 2004). It is reasonable to assume that other technological developments important to the First Amendment analysis have also occurred during that time. More and better filtering alternatives may exist than when the District Court entered its findings. Indeed, we know that after the District Court entered its factfindings, a congressionally ap[672]*672pointed commission issued a report that found that filters are more effective than verification screens. See swpm, at 668.
Delay between the time that a district court makes fact-findings and the time that a case reaches this Court is inevitable, with the necessary consequence that there will be some discrepancy between the facts as found and the facts at the time the appellate court takes up the question. See, e. g., Benjamin, Stepping into the Same River Twice: Rapidly Changing Facts and the Appellate Process, 78 Texas L. Rev. 269, 290-296 (1999) (noting the problems presented for appellate courts by changing facts in the context of cases involving the Internet, and giving as a specific example the Court’s decision in Reno, 521 U. S. 844). We do not mean, therefore, to set up an insuperable obstacle to fair review. Here, however, the usual gap has doubled because the ease has been through the Court of Appeals twice. The additional two years might make a difference. By affirming the preliminary injunction and remanding for trial, we allow the parties to update and supplement the factual record to reflect current technological realities.
Remand will also permit the District Court to take account of a changed legal landscape. Since the District Court made its factfindings, Congress has passed at least two further statutes that might qualify as less restrictive alternatives to COPA — a prohibition on misleading domain names, and a statute creating a minors-safe “Dot Kids” domain. See supra, at 663. Remanding for trial will allow the District Court to take into account those additional potential alternatives.
On a final point, it is important to note that this opinion does not hold that Congress is incapable of enacting any regulation of the Internet designed to prevent minors from gaining access to harmful materials. The parties, because of the conclusion of the Court of Appeals that the statute’s definitions rendered it unconstitutional, did not devote their atten[673]*673tion to the question whether further evidence might be introduced on the relative restrictiveness and effectiveness of alternatives to the statute. On remand, however, the parties will be able to introduce further evidence on this point. This opinion does not foreclose the District Court from concluding, upon a proper showing by the Government that meets the Government’s constitutional burden as defined in this opinion, that COPA is the least restrictive alternative available to accomplish Congress’ goal.
* * *
On this record, the Government has not shown that the less restrictive alternatives proposed by respondents should be disregarded. Those alternatives, indeed, may be more effective than the provisions of COPA. The District Court did not abuse its discretion when it entered the preliminary injunction. The judgment of the Court of Appeals is affirmed, and the case is remanded for further proceedings consistent with this opinion.
It is so ordered.