DOE v. BURLEW (2026)

OPINION

This case shows the “cost” that litigants impose on themselves when they pursue only facial constitutional claims. Moody v. NetChoice, LLC, 603 U.S. 707, 723, 144 S.Ct. 2383, 219 L.Ed.2d 1075 (2024). Kentucky passed a law requiring covered sex offenders to put their legal names on their qualifying social-media accounts. John Doe (a covered sex offender) sought to have the statute declared unconstitutional and moved for a preliminary injunction. He argued that the law violated the First Amendment because it barred regulated parties from speaking anonymously online. Yet he did not attempt to enjoin the law as applied to any specific speech or any specific social-media account. Rather, he sought broad relief by arguing that the law was facially overbroad. Although the district court agreed with Doe, it did not engage in the demanding comprehensive review that a facial challenge requires. See id. at 723–26, 144 S.Ct. 2383. Like the Supreme Court in NetChoice, then, we vacate the court's injunction and remand for further proceedings.

I

At a legislative hearing in early 2024, a Kentucky state senator recalled a conversation that she had with a constituent about a “loophole” in Kentucky's sex-offender registration laws. While the constituent was planning an event for children, an anonymous man reached out on social media asking about whether his business could participate. The constituent discovered only through additional research that Kentucky had placed this man on its sex-offender registry. The senator introduced a bill—SB 249—to close this loophole by requiring sex offenders to use their legal names on social media. Within months of this hearing, the Kentucky legislature unanimously passed the bill, and the Governor signed it into law. See Ky. Rev. Stat. § 17.544.

Kentucky law now prohibits a sex-offender “registrant who has committed a criminal offense against a victim who is a minor” from “creat[ing] or hav[ing] control of an account on a social media platform unless the account displays his or her full legal name.” Id. § 17.544(2). The law defines the phrase “criminal offense against a victim who is a minor” to include over a dozen crimes—including, for example, kidnapping, unlawful imprisonment, human trafficking involving commercial sexual activity, and sexual abuse—so long as the victim of these crimes was under the age of 18. Id. § 17.500(3)(a); see also id. § 17.500(5).

The law also contains an extended definition of the types of “social media platform[s]” that do (and do not) trigger its disclosure duties. Id. § 17.544(1). On the one hand, the definition covers “a website or application that is open to the public, allows a user to create an account, and enables users to” (among other things) “[i]nteract socially with other users” on the website or in the app, “share a social connection” with other users, and “[c]reate or post content viewable by others[.]” Id. § 17.544(1)(a). On the other hand, it excludes a “broadband internet access service,” an email account, a search engine, a “cloud storage or cloud computing service,” a service “in which interaction between users is limited to reviewing products offered for sale” or “commenting on” other users’ “reviews,” and a website “that consists primarily of” a provider's own selected content, even if readers may “chat” or post “comments” about this content. Id. § 17.544(1)(b).

Before this new disclosure law became effective, John Doe (a Kentucky citizen proceeding anonymously) sued John Burlew in his “official capacity as Daviess County Attorney and on behalf of all Kentucky County Attorneys” to enjoin the law's enforcement. Am. Compl., R.11, PageID 50. A state court had previously convicted Doe of multiple child-pornography offenses when he was 18 years old in 2015. Doe received a 5-year suspended sentence for these crimes and must register as a sex offender for the rest of his life. So the new law applies to him. Since his conviction, though, Doe has lived a law-abiding life. He has not been under any type of “criminal supervision” for many years. Id., PageID 53. He has also married and had children.

Doe regularly uses social media, including “X (formerly Twitter) and Reddit,” to obtain news, communicate with friends and family, and comment on other users’ content. Id. Although he keeps his social-media accounts anonymous, he does regularly disclose his prior conviction and status as a registered sex offender. He also states his (unpopular) political opinion that sex-offender registration laws do not deter crime and unfairly affect his children. When Doe has disclosed his prior crime or stated these opinions, other social-media users have harassed him online and threatened him and his family. He has also faced harassment for stating unrelated opinions. Doe fears that the new disclosure duty could lead to “severe harm” for his wife and children because the sex-offender laws elsewhere require him to publicly reveal his address. Id., PageID 54. Doe thus plans to “deactivate his social media and cease using it altogether” rather than put his legal name on the accounts. Id., PageID 55.

Doe sued on behalf of himself and brought a putative class action on behalf of similarly situated Kentuckians. His complaint alleged two counts. It first alleged that the disclosure law was “facially unconstitutional” because it violated the rights of Doe and similar Kentuckians to speak anonymously under the First Amendment. Id., PageID 60. This count sought “a declaration that the Statute is unconstitutional facially and as applied[.]” Id. The complaint next alleged that the disclosure law was “facially overbroad” under the First Amendment because it covered all speech on social-media platforms. Id., PageID 62. This count sought a “declaration that the Statute is facially unconstitutional[.]” Id., PageID 63.

Shortly after suing, Doe moved for a preliminary injunction and for certification of a class action under Federal Rule of Civil Procedure 23. The district court resolved these two motions in two opinions. The court first considered the injunction request. See Doe v. Burlew, 740 F. Supp. 3d 576, 579 (W.D. Ky. 2024). It found that Doe would likely succeed on his claim that the law infringed “his First Amendment right to post anonymously” and was “overbroad” because it regulated all social-media speech. Id. at 584–85. The court also concluded that it could grant class-wide relief even before ruling on Doe's motion to certify a class. Id. at 586–87. So it granted a preliminary injunction that barred “Burlew and all County Attorneys” from enforcing the disclosure law against anyone. Id. at 587.

The court next resolved Doe's class-certification motion. See Doe v. Burlew, 2024 WL 3463960, at *1 (W.D. Ky. July 18, 2024). It first held that Doe lacked standing to serve as a class representative for absent class members against any County Attorney other than Burlew. See id. at *2. It then held that Doe could not pursue a class action even against Burlew because he failed to show that the absent class members were too numerous to join a single suit and that his counsel could adequately represent a class. See id. at *3–4. The court's decision to deny class certification led it to narrow its preliminary injunction. See id. at *4. It limited the injunction to cover only Burlew (not other County Attorneys) and to protect only Doe (not other class members). See id.

The parties both appealed. Burlew appealed the court's decision to grant Doe a preliminary injunction. And Doe cross-appealed the court's refusal to grant broader relief.

II

We generally consider four factors when evaluating whether a district court properly granted a preliminary injunction. Winter v. Nat. Res. Def. Council, Inc., 555 U.S. 7, 20, 129 S.Ct. 365, 172 L.Ed.2d 249 (2008). In First Amendment cases, though, a plaintiff's likelihood of success on the merits typically drives the outcome. See Fischer v. Thomas, 52 F.4th 303, 307 (6th Cir. 2022) (per curiam); Liberty Coins, LLC v. Goodman, 748 F.3d 682, 689 (6th Cir. 2014). And here, the district court legally erred in the way it found a likelihood of success. So we remand for further proceedings.

A

1. To explain the district court's error, we find it helpful to disentangle different types of claims and remedies. We start with claims. The Supreme Court has long distinguished “facial” from “as-applied” constitutional claims. See NetChoice, 603 U.S. at 723–24, 144 S.Ct. 2383; Wash. State Grange v. Wash. State Republican Party, 552 U.S. 442, 449–51, 128 S.Ct. 1184, 170 L.Ed.2d 151 (2008). These claims differ in the showing of unconstitutionality that a challenger must make to succeed. See Bucklew v. Precythe, 587 U.S. 119, 138, 139 S.Ct. 1112, 203 L.Ed.2d 521 (2019). An as-applied claim requires a challenger to show only that a law violates the Constitution as applied to the specific conduct that the challenger seeks to undertake. See, e.g., FEC v. Wis. Right to Life, Inc., 551 U.S. 449, 456–57, 127 S.Ct. 2652, 168 L.Ed.2d 329 (2007) (Roberts, C.J., opinion). A facial claim, by contrast, requires a challenger to show that the law violates the Constitution in a much broader set of fact patterns. See Wash. State Grange, 552 U.S. at 449, 128 S.Ct. 1184. This broader theory seeks a holding that the law “could never be applied in a valid manner.” Members of City Council of L.A. v. Taxpayers for Vincent, 466 U.S. 789, 797–98, 104 S.Ct. 2118, 80 L.Ed.2d 772 (1984) (emphasis added).

Facial claims themselves come in two varieties. Outside the speech context, these claims require a challenger to show that a law violates a constitutional right in all its potential applications. See United States v. Hansen, 599 U.S. 762, 769, 143 S.Ct. 1932, 216 L.Ed.2d 692 (2023); United States v. Salerno, 481 U.S. 739, 745, 107 S.Ct. 2095, 95 L.Ed.2d 697 (1987). That type of challenge is “hard to win.” NetChoice, 603 U.S. at 723, 144 S.Ct. 2383. But the Supreme Court's so-called “overbreadth” doctrine adopts an easier-to-meet test under the First Amendment to provide “breathing room” for speech. Hansen, 599 U.S. at 769–70, 143 S.Ct. 1932. A free-speech challenger can succeed on a facial overbreadth challenge by showing that a law “prohibits a substantial amount of protected speech” as compared to its “plainly legitimate sweep.” United States v. Williams, 553 U.S. 285, 292, 128 S.Ct. 1830, 170 L.Ed.2d 650 (2008); see NetChoice, 603 U.S. at 723–24, 144 S.Ct. 2383.

We next turn to remedies. When an as-applied claim has succeeded, courts have typically granted a narrow injunction that bars the defendant from enforcing the law as applied to the specific plaintiff and the specific conduct found constitutionally protected. See Trump v. CASA, Inc., 606 U.S. 831, 843, 145 S.Ct. 2540, 222 L.Ed.2d 930 (2025); Ayotte v. Planned Parenthood of N. New England, 546 U.S. 320, 328–32, 126 S.Ct. 961, 163 L.Ed.2d 812 (2006); Scott v. Donald, 165 U.S. 107, 115–17, 17 S.Ct. 262, 41 L.Ed. 648 (1897). When a facial claim has succeeded, by contrast, courts have issued broader relief. In recent decades, they have granted “universal” injunctions that prohibited the defendant from enforcing the law against not just the plaintiff but also third parties not before the court. See CASA, 606 U.S. at 837, 840, 861, 145 S.Ct. 2540. That said, the Supreme Court recently indicated that a “universal injunction”—at least as applied to a federal law or executive order—“falls outside the bounds of a federal court's equitable authority under the Judiciary Act.” Id. at 847, 145 S.Ct. 2540. In that context, then, courts cannot grant remedies that are “broader than necessary to provide complete relief to” an individual plaintiff. Id. at 861, 145 S.Ct. 2540. We have yet to decide (and need not decide now) how CASA applies in the context of a state-law challenge.

Even still, courts also have granted what could be described as “facial” relief to a specific plaintiff. The “scope of injunctive relief” that a court may grant to a plaintiff depends on the scope of the constitutional “violation” that the plaintiff has “established” on the merits. Califano v. Yamasaki, 442 U.S. 682, 702–03, 99 S.Ct. 2545, 61 L.Ed.2d 176 (1979). Courts may not grant relief that goes beyond the violation “that produced the injury in fact that the plaintiff has established.” Lewis v. Casey, 518 U.S. 343, 357, 116 S.Ct. 2174, 135 L.Ed.2d 606 (1996). But they may grant “complete relief” to plaintiffs for violations that injure them. CASA, 606 U.S. at 851–52, 145 S.Ct. 2540. So if plaintiffs establish a law's unconstitutionality in the required number of applications, courts may grant injunctions that prohibit the government from enforcing the law against the plaintiff in any circumstance—not just as applied to specific conduct.

2. The parties debate how to characterize Doe's challenge. His complaint asserted two counts. All agree on the nature of the second count. This count alleged a facial claim: it attacked the Kentucky statute as “facially overbroad” because it “chill[ed] a substantial amount of protected First Amendment activity[.]” Am. Compl., R.11, PageID 61–62. And this count sought facial relief: it requested an “injunction precluding enforcement of the Statute” against any Kentuckian and any anonymous social-media account. Id., PageID 63.

But the parties dispute on appeal whether Doe's first count raised a facial or as-applied challenge. We agree with Burlew that this count also presented a facial attack. Count one alleged that the Kentucky statute was “facially unconstitutional” because it violated the right of Doe and similar Kentuckians “to speak anonymously” on social media. Id., PageID 59–60. The count thus treated every conceivable application of the statute as unconstitutional. It did not, by contrast, suggest that the law was unconstitutional as applied to a “defined subset” of Doe's speech. United States v. Stevens, 559 U.S. 460, 473 n.3, 130 S.Ct. 1577, 176 L.Ed.2d 435 (2010); cf. Wis. Right to Life, 551 U.S. at 460, 127 S.Ct. 2652 (Roberts, C.J., opinion). It, for example, made no effort to distinguish between anonymous political speech on X about a pending election and anonymous commercial speech on Facebook about a new product. Compare McIntyre v. Ohio Elections Comm'n, 514 U.S. 334, 357, 115 S.Ct. 1511, 131 L.Ed.2d 426 (1995), with Milavetz, Gallop & Milavetz, P.A. v. United States, 559 U.S. 229, 249, 130 S.Ct. 1324, 176 L.Ed.2d 79 (2010).

Count I also asked for an “injunction” identical to the one requested for count two that would enjoin “enforcement of the Statute” against any Kentuckian and any anonymous social-media account. Am. Compl., R.11, PageID 61. Admittedly, count one also sought a “declaration” that the law was “unconstitutional facially and as applied[.]” Id., PageID 60 (emphasis added). But such terminology does not transform a facial challenge into an as-applied one because “[t]he label is not what matters.” Bucklew, 587 U.S. at 139, 139 S.Ct. 1112 (citation omitted). Nor did the count identify any specific anonymous speech for protection. So we agree with the district court that the complaint's factual “allegations” did not “support[ ]” any request for an “as-applied” declaration. Doe, 740 F. Supp. 3d at 583 n.2.

Two other datapoints confirm that we should consider only Doe's facial challenge. For one thing, the district court treated Doe's request for a preliminary injunction as raising only a facial claim seeking a facial remedy. See id. at 583–86. The court thus evaluated Doe's likelihood of success on the merits using the Court's overbreadth standards for facial claims under the First Amendment. See id. at 583. Finding in Doe's favor, the district court initially granted broad facial relief, enjoining the law's enforcement as applied to anyone and any social-media account. See id. at 586–87. And although the court later limited its relief to Doe when denying Doe's request to certify a class action, it still relied on that initial facial analysis. See Doe, 2024 WL 3463960, at *4. It barred Burlew from enforcing the law against Doe in all circumstances, not just as applied to any specific speech that Doe wished to express anonymously or any specific social-media account. See id.

For another thing, Doe's appellate briefing confirms the facial nature of his claims and remedies. He argues that the law likely violates the First Amendment in the abstract without relying on specific examples of protected anonymous speech. Doe's Br. 17–27. He also argues that the law is facially overbroad. Id. at 27–41. And he argues that the law's many different applications to different facts make no difference because it triggers the “same First Amendment issues” for “everyone who is subject to it.” Id. at 39 (citation omitted).

In sum, Doe has not “preserved” any as-applied challenge in this appeal. Stevens, 559 U.S. at 473 n.3, 130 S.Ct. 1577. We thus must consider only whether he has shown that he will likely succeed on his “overbreadth challenge” to the Kentucky law. Hansen, 599 U.S. at 769, 143 S.Ct. 1932.

B

Doe's decision to pursue a facial claim “comes at a cost.” NetChoice, 603 U.S. at 723, 144 S.Ct. 2383. Courts do not favor these types of challenges. See Wash. State Grange, 552 U.S. at 450, 128 S.Ct. 1184; Connection Distrib. Co. v. Holder, 557 F.3d 321, 336 (6th Cir. 2009). The argument that a law facially violates the Constitution requires courts to engage in a lot of “speculation” about how the law might apply to unknown factual situations. Wash. State Grange, 552 U.S. at 450, 128 S.Ct. 1184. This speculation compels courts to prematurely identify a “law's coverage” even though prosecutors may never apply the law to the hypothesized facts. NetChoice, 603 U.S. at 723, 144 S.Ct. 2383. It also compels courts to prematurely resolve many difficult constitutional issues that may never arise in a real-world setting. See Wash. State Grange, 552 U.S. at 450, 128 S.Ct. 1184. Facial challenges thus undercut the “democratic” structure of our government and the judiciary's traditional role in it. Id. at 450–51, 128 S.Ct. 1184.

These judicial concerns with facial challenges do not go away even when a litigant sues to enjoin a law under the First Amendment's Free Speech Clause. See Connection Distrib., 557 F.3d at 336. To be sure, the Supreme Court's “overbreadth” doctrine reduces a free-speech challenger's burden to succeed on a facial claim as compared to the burden on a plaintiff invoking a different right. See NetChoice, 603 U.S. at 723, 144 S.Ct. 2383. Still, the Court has never “casually employed” the “strong medicine” that is an overbreadth finding. Williams, 553 U.S. at 293, 128 S.Ct. 1830 (quoting L.A. Police Dep't v. United Reporting Publ'g Corp., 528 U.S. 32, 39, 120 S.Ct. 483, 145 L.Ed.2d 451 (1999)). Rather, challengers must show that a law has “realistic” rather than “fanciful” applications that violate the Free Speech Clause. Hansen, 599 U.S. at 770, 143 S.Ct. 1932. And these applications “must be substantially disproportionate to the statute's lawful sweep.” Id.; see Ams. for Prosperity Found. v. Bonta, 594 U.S. 595, 615, 141 S.Ct. 2373, 210 L.Ed.2d 716 (2021).

In NetChoice, the Court recently described the method that courts should follow before finding a law unconstitutionally overbroad. See 603 U.S. at 724–25, 144 S.Ct. 2383. That case addressed state laws that regulated social-media “platforms” such as Facebook and YouTube. Id. at 720, 144 S.Ct. 2383. The laws regulated a social-media company's content moderation (the ways in which the company highlighted or minimized user-created videos, messages, and the like) and required the company to send an individual message to a user when restricting the user's content. Id. at 720–21, 144 S.Ct. 2383. Two circuit courts had reached different results on whether these laws violated the First Amendment. See id. at 722, 144 S.Ct. 2383. But the Court reversed both decisions because each considered the validity of the laws only as applied to the “best-known services” of the “giant social-media platforms” (such as “Facebook's News Feed and YouTube's homepage”). Id. at 724, 144 S.Ct. 2383. The courts thus wrongly treated the overbreadth challenges “more like as-applied claims than like facial ones.” Id.

The Court then explained the proper two-step approach for overbreadth challenges. Id. at 724–25, 144 S.Ct. 2383. At step one, a court must evaluate a law's “scope” by considering the “activities” it covers and the “actors” it regulates. Id. at 724, 144 S.Ct. 2383. The social-media laws in NetChoice, for example, might regulate not just a company's general news feeds but also its direct-messaging services. Id. At step two, a court must identify which of a law's many applications would violate the First Amendment and which would not. See id. at 725, 144 S.Ct. 2383. It then must “compare the two sets” of applications. Id. at 726, 144 S.Ct. 2383. Here again, the laws might raise greater constitutional concerns as applied to a news feed than to direct messages. Id.

The district court's evaluation of the Kentucky law in this case suffers from the same problems as the circuit courts’ evaluation of the social-media laws in NetChoice. While ostensibly applying overbreadth standards, the court narrowly considered the law's validity as applied to Doe's request to speak anonymously on social media. See Doe, 740 F. Supp. 3d at 583–85. The court nowhere identified the full scope of what the law “covers.” Hansen, 599 U.S. at 770, 143 S.Ct. 1932. Nor did it add up the law's constitutional applications and compare them to its unconstitutional ones. See NetChoice, 603 U.S. at 725–26, 144 S.Ct. 2383. The court understandably omitted this analysis because it issued its opinion around the same time as NetChoice. But we still must follow NetChoice’s lead as to the proper solution for its error: remand for reconsideration under the two-step approach. Id. at 726, 144 S.Ct. 2383. Simply stated, “a facial-challenge analysis requires more” than what the district court did here. Free Speech Coal., Inc. v. Skrmetti, 2025 WL 512049, at *2 (6th Cir. Jan. 13, 2025).

A review of both steps confirms this point. Starting with step one, the parties engage in several unresolved debates about the Kentucky law's scope. They, for example, debate when sex-offender registrants must disclose their identities. According to Doe, legislation raises greater First Amendment concerns when it requires parties to disclose their identities at the precise moment of their speech rather than separate in time from that speech. Doe's Br. 23 (citing Buckley v. Am. Const. Law Found., Inc., 525 U.S. 182, 119 S.Ct. 636, 142 L.Ed.2d 599 (1999)). In American Constitutional Law Foundation, for example, the Supreme Court held that the First Amendment did not permit a State to require petition circulators who seek signatures for ballot initiatives to wear name badges that disclose their identities while speaking to members of the public. See 525 U.S. at 197–200, 119 S.Ct. 636. But the Court did suggest that the State could require these circulators to submit an affidavit subject to the public-record laws that revealed their identities after this signature gathering. Id. at 198–99, 119 S.Ct. 636. And Doe argues that the Kentucky law requires a “contemporaneous identification” of his real name with every message that he sends on a social-media platform. Doe's Br. 23.

According to Burlew, by contrast, a sex-offender registrant may comply with the Kentucky law by ensuring that the social-media “account displays his or her full legal name” somewhere. Burlew's Reply Br. 7–8 (citing Ky. Rev. Stat. § 17.544(2)). Burlew thus suggests that sex-offender registrants may post content and send messages under anonymous usernames as long as they place their real names in, say, their account bios. Id. He adds that this reading would make the law more like the affidavit requirement that American Constitutional Law Foundation upheld: members of the public would have to take extra steps to determine the identity of the accountholders in this case—just as members of the public had to take extra steps to determine the identity of the petition circulators in American Constitutional Law Foundation. See 525 U.S. at 198–99, 119 S.Ct. 636. Is this factually true? For how many accounts? The record leaves these questions unclear.

Next, the parties debate what the law means when it says that the disclosure obligation applies to all social-media accounts that sex-offender registrants “control.” Ky. Rev. Stat. § 17.544(2). According to Doe, this obligation could require him to place his name on “his own children's social media accounts” because he monitors what they view and post online. Doe's Br. 29. According to Burlew, it does not reach this conduct. Resp., R.19, PageID 153. So what does the statute mean by the word “control”?

Turning to step two, we again see many unresolved questions. To start, the constitutional “answers might differ” depending on what sex-offender registrants plan to do with their social-media accounts. NetChoice, 603 U.S. at 725, 144 S.Ct. 2383. According to Burlew, statistics show that less than half of U.S. adult users post content on social media. Burlew's Br. 21–22. He thus suggests that a (conceivably large) number of sex-offender registrants who have social-media accounts will use those accounts to review content—not create it. And while the Supreme Court has noted that the Free Speech Clause includes a right to speak anonymously, see McIntyre, 514 U.S. at 342, 115 S.Ct. 1511, the Court has yet to hold that the Clause contains a right to peruse the internet anonymously. Burlew thus argues that the law covers “a great deal of nonexpressive conduct” that “does not implicate the First Amendment at all.” Burlew's Br. 23 (citing Hansen, 599 U.S. at 782, 143 S.Ct. 1932); see also Virginia v. Hicks, 539 U.S. 113, 123–24, 123 S.Ct. 2191, 156 L.Ed.2d 148 (2003). According to Doe, by contrast, “the Constitution protects the right to receive information and ideas.” Doe's Br. 36 (citing Stanley v. Georgia, 394 U.S. 557, 564, 89 S.Ct. 1243, 22 L.Ed.2d 542 (1969)). The Supreme Court, for example, recently found it problematic for a State to categorically prohibit sex offenders from visiting certain websites partly because the ban limited their ability to obtain information. See Packingham v. North Carolina, 582 U.S. 98, 107, 137 S.Ct. 1730, 198 L.Ed.2d 273 (2017). And Doe seeks to add an anonymity element to this distinct right.

The constitutional “answers” also “might differ” depending on the platforms at issue. NetChoice, 603 U.S. at 725, 144 S.Ct. 2383. Doe's verified complaint listed only X and Reddit as examples of platforms on which he has anonymous accounts. Am. Compl., R.11, PageID 53. Yet Burlew responded that other platforms (like LinkedIn) bar users from having an anonymous account and require them to show their legal names. Burlew's Br. 14. Do users have the same constitutional interest in anonymity as applied to platforms that prohibit this anonymity as a matter of private contract? The record also leaves unclear whether any covered social-media platforms tailor their services specifically to children. Would sex-offender registrants have the same alleged right to post anonymously on one of these children-focused platforms?

Likewise, these “answers might differ” depending on the speech at issue. NetChoice, 603 U.S. at 725, 144 S.Ct. 2383. Doe alleges that he engages in pure political speech on X and Reddit. Am. Compl., R.11, PageID 53–54. And the Supreme Court has referred to “a respected tradition of anonymity in the advocacy of political causes.” McIntyre, 514 U.S. at 343, 115 S.Ct. 1511. But Doe seeks to enjoin the law as applied to all social-media speech, not just political speech. Take the “commercial” speech that both parties flag. Burlew's Br. 24; Doe's Br. 28. A “disclosure requirement” in that context generally demands less-exacting scrutiny. Milavetz, 559 U.S. at 249, 130 S.Ct. 1324. Recall the origins for the Kentucky law: Its sponsor noted that a constituent had received a social-media inquiry from an anonymous sex offender who had asked if his business could participate in an event for children. Does the First Amendment really prohibit Kentucky from requiring this man to reveal his identity on whatever social-media account he used to make this inquiry? Since the overbreadth doctrine “does not normally apply to commercial speech” anyway, how do these covered applications play into the overbreadth calculus? Bd. of Trs. of State Univ. of N.Y. v. Fox, 492 U.S. 469, 481, 109 S.Ct. 3028, 106 L.Ed.2d 388 (1989). And should the constitutional analysis differ depending on whether a person on the sex-offender registry posts generic content for all to see or sends these types of “direct messages” to a specific person? NetChoice, 603 U.S. at 725–26, 144 S.Ct. 2383.

Even for political expression, the Court has never adopted a “freewheeling right” to speak anonymously. John Doe No. 1 v. Reed, 561 U.S. 186, 218 n.4, 130 S.Ct. 2811, 177 L.Ed.2d 493 (2010) (Stevens, J., concurring in part and concurring in the judgment). To the contrary, it has seemingly adopted different rules for different mediums. So it found unconstitutional an Ohio law that barred individuals from distributing anonymous campaign leaflets. See McIntyre, 514 U.S. at 341–57, 115 S.Ct. 1511. But it found constitutional a federal law that required speakers to disclose their identities in election ads on television. See Citizens United v. FEC, 558 U.S. 310, 366–71, 130 S.Ct. 876, 175 L.Ed.2d 753 (2010); McConnell v. FEC, 540 U.S. 93, 231–32, 124 S.Ct. 619, 157 L.Ed.2d 491 (2003). As others have noted, these decisions sit in substantial tension with each other. See Majors v. Abell, 361 F.3d 349, 356–57 (7th Cir. 2004) (Easterbrook, J., dubitante). This fact-specific approach would seemingly make facial resolution all the more difficult.

Lastly, the relevant constitutional “answers might differ” depending on the speaker at issue. NetChoice, 603 U.S. at 725, 144 S.Ct. 2383. The Supreme Court has held that speaker-based regulations of expression do not trigger heightened scrutiny if the government can justify them based on “some special characteristic” that the regulated speaker possesses—rather than the content of the speech itself. TikTok Inc. v. Garland, 604 U.S. 56, 72–73, 145 S.Ct. 57, 220 L.Ed.2d 319 (2025) (per curiam) (quoting Turner Broad. Sys., Inc. v. FCC, 512 U.S. 622, 660–61, 114 S.Ct. 2445, 129 L.Ed.2d 497 (1994)). Burlew might argue on remand that the regulated persons here have this type of “special characteristic” because they all have committed a qualifying “criminal offense against a victim who is a minor.” See Ky. Rev. Stat. §§ 17.500(3)(a), 17.544(2); cf. Conn. Dep't of Pub. Safety v. Doe, 538 U.S. 1, 4, 123 S.Ct. 1160, 155 L.Ed.2d 98 (2003). And might the strength of Kentucky's interests differ depending on the specific circumstances of each covered offender? Doe asserts that he committed child-pornography offenses years ago, that these offenses did not involve contacting minors, that he did not serve time in prison, and that he has committed no other crimes. Am. Compl., R.11, PageID 53. But the law would also cover, for example, a just-released child abuser who used social media to stalk his past victims. A court's answer to the question whether Kentucky has a strong enough interest to apply this law to Doe does not necessarily answer the separate question whether it has a strong enough interest to apply the law to other sex offenders. Cf. Reed, 561 U.S. at 200–01, 130 S.Ct. 2811; Holder v. Humanitarian L. Project, 561 U.S. 1, 39, 130 S.Ct. 2705, 177 L.Ed.2d 355 (2010).

Doe's contrary arguments lack merit. He first compares the Kentucky law to the North Carolina law found unconstitutional in Packingham. Yet the two laws differ. The North Carolina law in Packingham affirmatively prohibited sex offenders from using social-media accounts. 582 U.S. at 101–02, 137 S.Ct. 1730. The Kentucky law merely requires a disclosure of the sex offender's legal name on these accounts. See Ky. Rev. Stat. § 17.544(2). The Supreme Court has long viewed disclosure requirements (like the one in this case) as “less restrictive” on speech than absolute prohibitions (like the one in Packingham). Citizens United, 558 U.S. at 369, 130 S.Ct. 876; see Reed, 561 U.S. at 196, 130 S.Ct. 2811.

Doe fares no better with his reliance on our own decision in Speet v. Schuette, 726 F.3d 867 (6th Cir. 2013). There, we found a Michigan law that criminalized begging overbroad under the First Amendment. Id. at 870–80. Yet the law “ban[ned] an entire category of” speech—soliciting money—that we found “the First Amendment protects.” Id. at 879. Here, by contrast, the law regulates the creation or control of a social-media account. See Ky. Rev. Stat. § 17.544(2). This conduct may (or may not) lead to protected speech. Unlike the Michigan law in Speet, then, the Kentucky law here may well cover substantial “nonexpressive conduct” that does not implicate the First Amendment. Hansen, 599 U.S. at 782, 143 S.Ct. 1932; see Hicks, 539 U.S. at 123–24, 123 S.Ct. 2191.

All told, “[t]he online world is variegated and complex, encompassing an ever-growing number of apps, services, functionalities, and methods for communication and connection.” NetChoice, 603 U.S. at 725, 144 S.Ct. 2383; see id. at 745, 144 S.Ct. 2383 (Barrett, J., concurring). And a party raising a facial challenge must account for this variety. Neither Doe nor the district court has done so.

We vacate the preliminary injunction and remand this case to the district court for proceedings consistent with this opinion.

MURPHY, Circuit Judge.