WILKINSON, Circuit Judge:

A Maryland law requires newspapers, among other platforms, to publish on their websites, as well as retain for state inspection, certain information about the political ads they decide to carry. This case asks, at bottom, whether these terms can be squared with the First Amendment. For the reasons discussed below, we agree with the district court that they cannot. While Maryland's law tries to serve important aims, the state has gone about this task in too circuitous and burdensome a manner to satisfy constitutional scrutiny.

I.

In 2016, Donald Trump was elected the forty-fifth President of the United States. In the months that followed, a growing consensus emerged that Russian nationals had attempted to interfere in the presidential election through a sustained disinformation campaign carried out on social media and other online platforms. Prompted by these revelations, a number of states looked to amend their election laws to better protect against foreign meddling. Maryland was one of them and, in May 2018, the state passed the Online Electioneering Transparency and Accountability Act (the "Act").

Maryland has long regulated campaign-related speech. Before the Act took effect, though, the state principally regulated direct participants in the political arena. Most relevant here, Maryland imposed certain disclosure and recordkeeping obligations on political speakers looking to influence a given election cycle. Md. Code Ann., Elec. Law §§ 13-401, 13-404, 13-304, 13-306, 13-307, 13-221. First, the state required "campaign material" to include an "authority line" identifying the person or group behind a political advertisement.1 Second, Maryland mandated that "political committees," or similar persons making independent expenditures above a specific dollar amount, must collect and report to the Maryland Board of Elections basic information about their donors and related expenses.2 Before the Act, these provisions applied only to purchasers of TV, radio, and print advertising.

Following the 2016 election, however, Maryland concluded that this disclosureand-recordkeeping regime was inadequate. See Washington Post v. McManus, 355 F.Supp.3d 272, 281-282 (D. Md. 2019). In particular, legislators determined that Maryland's existing campaign finance regulations largely failed to cover the internet. And this presented a problematic blind spot. For one, the 2016 election marked an important shift in how campaigns were waged, with a surge in spending on online political advertising by candidates and political organizations. See J.A. 117 (detailing eightfold increase from 2012). Moreover, 2016 also involved pervasive attempts by foreign nationals to influence American elections by way of the internet. See J.A. 119-20 (describing internal investigations at Facebook, Twitter, and Google). These efforts were particularly prevalent in Maryland, which by some accounts was one of the top three most targeted states in 2016. See J.A. 118-19.

Against this backdrop, Maryland decided to develop legislation that would bolster the state's campaign finance regulations. With a particular eye toward combatting foreign meddling, the state made two main changes to its laws. See Washington Post, 355 F. Supp. 3d at 281-82. First, Maryland broadened its extant disclosure-and-recordkeeping regulations to include online advertisements. Accordingly, the campaign finance regime that previously applied only to TV, radio, and print was expanded to also include online political ads. Id. at 282. These provisions, which applied directly to ad purchasers, are not challenged here. Second, the Act extended Maryland's campaign finance laws to include for the first time "online platforms." Id. at 282-83. An "online platform" under the Act is defined in terms of both its size and its speech, picking up essentially any public website in the state that reaches a certain circulation (100,000 unique monthly visitors) and receives money for "qualifying paid digital communications" (which are, in short, political ads). Md. Code Ann., Elec. Law § 1-101(dd-1). The distinctive feature of these sections is that their onus falls on the websites themselves, not the political speakers. These provisions are the subject of this lawsuit.

The Act imposes two sets of disclosure obligations on "online platforms" operating in Maryland. First, there is a "publication requirement." Under this provision, online platforms must post certain information about the political ads on their websites. Id. § 13-405(b). In the main, within 48 hours of an ad being purchased, platforms must display somewhere on their site the identity of the purchaser, the individuals exercising control over the purchaser, and the total amount paid for the ad. They must keep that information online for at least a year following the relevant election. Second, there is an "inspection requirement." Under this part, platforms must collect records concerning their political ad purchasers and retain those records for at least a year after the election so that the Maryland Board of Elections can review them upon request. Id. § 13-405(c). As the district court explained, the "publication requirement and state inspection requirement are functionally distinct, but they operate as part of a single scheme." Washington Post, 355 F. Supp. 3d at 283. To that end, both requirements attach when (i) the buyer notifies a platform that its ad constitutes a "qualifying paid digital communication[ ]" under the Act, Md. Code Ann., Elec. Law § 13-405(a)(1), and (ii) supplies the platform with the necessary information that it will then have to post and retain as required by the publication and inspection parts of the Act, id. § 13-405(d)(1).

The Act took effect in July 2018 without the signature of Maryland Governor Larry Hogan. At the time, Governor Hogan feared the Act raised "serious constitutional concerns" because, in large part, it "would compel speech by news outlets." J.A. 111-12. That August, a collection of news outlets operating in the state (the "Publishers") filed for a preliminary injunction to prevent the platform-specific parts of the Act from going into effect.3 As relevant here, the Publishers argued that the Act's publication and inspection requirements violated the First Amendment —both facially, as a regulation that targets neutral third-party platforms, and also as-applied, as a law encompassing news outlets.

The district court concluded that the Publishers were likely to succeed on the merits of their challenge and that the other conditions for a preliminary injunction were also satisfied. Washington Post, 355 F. Supp. 3d at 305-306. It accordingly granted the Publishers' motion and preliminarily enjoined enforcement of the relevant portions of the Act as applied to them. Three points about the district court's opinion bear mention.

First, the court held that the Act should be evaluated under strict scrutiny rather than the more permissive framework of exacting scrutiny. Id. at 297. While the Supreme Court has applied exacting scrutiny to a number of campaign finance laws involving disclosure obligations, the court reasoned that those cases were inapposite here because the Maryland law burdened neutral third-party platforms rather than direct political participants. Id. at 293, 296-97. As such, strict scrutiny was appropriate since the Act was a content-based regulation that compelled political speech— features that would ordinarily require the most demanding form of judicial review. Id. at 297.

Second, the Act failed strict scrutiny. Although Maryland had certain compelling interests at stake, it failed to narrowly tailor its law in service of those ends. In particular, the court found that the Act was both over- and under-inclusive, ultimately failing to meaningfully address the core problem of foreign election interference that motivated the Act in the first place. Id. at 298-303.

Third, the district court also held that the Act would fail exacting scrutiny. This because, for similar reasons, there was such a mismatch between the Act's means and ends that it lacked the requisite substantial relationship to the governmental interests it claimed to further. Id. at 302-305.

This appeal followed. We review a district court's grant of a preliminary injunction for abuse of discretion. Pashby v. Delia, 709 F.3d 307, 319 (4th Cir. 2013). In so doing, we review factual findings for clear error and legal conclusions de novo. WV Ass'n of Club Owners & Fraternal Servs., Inc. v. Musgrave, 553 F.3d 292, 298 (4th Cir. 2009). Here, Maryland has effectively limited its appeal to the first Winter factor—that is, whether the Publishers have demonstrated that they are likely to succeed on the merits. See Winter v. Nat. Res. Def. Council, Inc., 555 U.S. 7, 20, 129 S.Ct. 365, 172 L.Ed.2d 249 (2008); Appellant's Opening Brief at 25, 54, 57. Our analysis below is accordingly limited to that issue.

Moreover, our holding is also limited by the posture of this case. While general First Amendment principles bear most definitely upon the resolution of the appeal, the ultimate issue before us is a narrower one, i.e., whether the Maryland Act as applied to these particular plaintiffs is unconstitutional. To that end, we do not expound upon the wide world of social media and all the issues that may be pertinent thereto. For the reasons that follow, we shall affirm the preliminary injunctive relief awarded by the district court.

II.

As the district court noted in a lengthy and thoughtful opinion, Washington Post, 355 F. Supp.3d at 272-306, the Act is a content-based law that targets political speech and compels newspapers, among other platforms, to carry certain messages on their websites. In other words, Maryland's law is a compendium of traditional First Amendment infirmities.

First, the Act is a content-based regulation on speech. It singles out one particular topic of speech—campaign-related speech—for regulatory attention. See R.A.V. v. City of St. Paul, 505 U.S. 377, 382-83, 112 S.Ct. 2538, 120 L.Ed.2d 305 (1992). The lodestar for the First Amendment is the preservation of the marketplace of ideas. Leathers v. Medlock, 499 U.S. 439, 448-49, 111 S.Ct. 1438, 113 L.Ed.2d 494 (1991). When the government seeks to favor or disfavor certain subject-matter because of the topic at issue, it compromises the integrity of our national discourse and risks bringing about a form of soft censorship. For this reason, content-based laws are "presumptively unconstitutional," Reed v. Town of Gilbert, ___ U.S. ___, 135 S.Ct. 2218, 2226, 192 L.Ed.2d 236 (2015); Cent. Radio Co. Inc. v. City of Norfolk, 811 F.3d 625, 633 (4th Cir. 2016), the presumption being necessary to ensure that the marketplace of ideas does not deteriorate into a forum for the subjects of state-favored speech.

Second, the Act singles out political speech. While generic content-based regulations strain our commitment to free speech, content-based regulations that target political speech are especially suspect. Because our democracy relies on free debate as the vehicle of dispute and the engine of electoral change, political speech occupies a distinctive place in First Amendment law. See, e.g., Burson v. Freeman, 504 U.S. 191, 196, 112 S.Ct. 1846, 119 L.Ed.2d 5 (1992) ("[T]he First Amendment has its fullest and most urgent application to speech uttered during a campaign for political office." (quoting Eu v. San Francisco Cty. Democratic Cent. Comm., 489 U.S. 214, 223, 109 S.Ct. 1013, 103 L.Ed.2d 271 (1989) (internal quotation omitted))). Regulations of political speech therefore "trench[ ] upon an area in which the importance of First Amendment protections is at its zenith." Meyer v. Grant, 486 U.S. 414, 425, 108 S.Ct. 1886, 100 L.Ed.2d 425 (1988) (internal markings omitted). The Act here aims directly at political speech. Indeed, its publication and inspection provisions apply exclusively to political speech. E.g., Md. Code Ann., Elec. Law § 1-101(k)(1) (defining covered "campaign material" as that "relat[ing] to a candidate, a prospective candidate, or the approval or rejection of a [ballot] question or prospective [ballot] question"). As such, the Act concerns content that is ordinarily shielded within "the heart of the First Amendment's protection." First Nat'l Bank of Boston v. Bellotti, 435 U.S. 765, 776, 98 S.Ct. 1407, 55 L.Ed.2d 707 (1978).

Third, the Act compels speech. And it does so in no small measure. Take, to start, the publication requirement. This provision requires online platforms that host political ads to post, in searchable format: (i) the ad purchaser's name and contact information; (ii) the identity of the treasurer of the political committee or the individuals exercising control over the ad purchaser; and (iii) the total amount paid for the ad. Md. Code Ann., Elec. Law §§ 13-405(b)(1), 13-405(b)(6)(i)-(ii). That is not all. The publication requirement also directs platforms to post all this information "in a clearly identifiable location on the online platform's website" within 48 hours of purchase, and to maintain this information on their websites for at least one year after the relevant election. Id. § 13-405(b)(3).

Furthermore, the Act's inspection requirement also compels speech. Under this provision, platforms must collect and retain records of the following information, to be disclosed to state regulators upon request:

Md. Code Ann., Elec. Law § 13-405(c)(3). Similar to the publication requirement, platforms must make these records available within 48 hours of the time an ad runs and retain them for at least one year after the relevant general election. Id. § 13-405(c)(2).

Critically, failure to comply with any of these provisions comes with real consequences. Noncompliance with either the Act's publication or inspection requirements is grounds for the Maryland Attorney General to seek injunctive relief to require removal of the ad. Md. Code Ann., Elec. Law § 13-405.1(b)(1)-(2). And failure to comply with the injunction is ultimately punishable by criminal penalties. Id. § 13-405.1(b)(4).

Taken together, the Act's publication and inspection requirements ultimately present compelled speech problems twice over. For one, they force elements of civil society to speak when they otherwise would have refrained. Time and again, the Supreme Court has made clear that it makes little difference for First Amendment purposes whether the government acts as censor or conductor. Indeed, the "freedom of speech `includes both the right to speak freely and the right to refrain from speaking at all.'" Janus v. Am. Fed'n of State, Cty. & Mun. Emps. Council 31, ___ U.S. ___, 138 S.Ct. 2448, 2463, 201 L.Ed.2d 924 (2018) (quoting Wooley v. Maynard, 430 U.S. 705, 714, 97 S.Ct. 1428, 51 L.Ed.2d 752 (1977) (internal quotation omitted)). It is the presence of compulsion from the state itself that compromises the First Amendment. The Amendment extends "not only to expressions of value, opinion, or endorsement, but equally to statements of fact the speaker would rather avoid." Hurley v. Irish-Am. Gay, Lesbian & Bisexual Grp. of Boston, 515 U.S. 557, 573, 115 S.Ct. 2338, 132 L.Ed.2d 487 (1995).

What's more, the fact that the Act compels third parties to disclose certain identifying information regarding political speakers implicates protections for anonymous speech. Requiring the press itself to disclose the identity or characteristics of political speakers is a problematic step. See Branzburg v. Hayes, 408 U.S. 665, 709, 92 S.Ct. 2646, 33 L.Ed.2d 626 (Powell, J., concurring) (noting that newsmen faced with a grand jury subpoena are not "without constitutional rights with respect to the gathering of news or in safeguarding their sources"); id. at 725, 92 S.Ct. 2646 (Stewart, J., dissenting) (requiring the press to disclose its sources risks "annex[ing] the journalistic profession as an investigative arm of government"). This country, moreover, has "a respected tradition of anonymity in the advocacy of political causes." McIntyre v. Ohio Elections Comm'n, 514 U.S. 334, 343, 115 S.Ct. 1511, 131 L.Ed.2d 426 (1995). Much as our forebears elected to hash out the architecture of this nation under the pseudonyms of "Publius" and "Agrippa," many political advocates today also opt for anonymity in hopes their arguments will be debated on their merits rather than their makers. See id. at 343, 343 n.6, 115 S.Ct. 1511. To be sure, this tradition of anonymity is anything but absolute. But revelations of executive misconduct throughout our history have often been anonymously sourced. And when the government enlists the press to disclose the sources of political speech, thus potentially exposing those speakers to identification and harassment, First Amendment protections and values come into play. See Pac. Gas & Elec. Co. v. Pub. Utils. Comm'n of Cal., 475 U.S. 1, 14-15, 106 S.Ct. 903, 89 L.Ed.2d 1 (1986) (plurality opinion).

In the end, each banner feature of the Act—the fact that it is content-based, targets political expression, and compels certain speech—poses a real risk of either chilling speech or manipulating the marketplace of ideas. Of course, these dangers, standing alone, may not be dispositive. But they certainly speak to the burden imposed by the Act, a burden that becomes all the more severe once we examine additional aspects of Maryland's law.

III.

A.

Maryland's law is different in kind from customary campaign finance regulations because the Act burdens platforms rather than political actors. So when "People for Jennifer Smith" want to place an online campaign advertisement with the Carroll County Times, it is the County Times that has to shoulder the bulk of the disclosure and recordkeeping obligations created by the sections of the Act challenged here.

This platform-oriented structure poses First Amendment problems of its own. See Brown v. Entm't Merchs. Ass'n, 564 U.S. 786, 802, 131 S.Ct. 2729, 180 L.Ed.2d 708 (2011) (noting "doubts [about] punishing third parties for conveying protected speech"). Of course, as Maryland is quick to point out, governments have long required, and the Supreme Court has long upheld, the publication and retention of certain information in connection with elections. See Buckley v. Valeo, 424 U.S. 1, 61-62, 96 S.Ct. 612, 46 L.Ed.2d 659 (1976) (describing history). And, to be sure, each of these cases involved laws that implicated the same traditional infirmities noted above; they were content-based, pegged to political expression, and compelled speech in some form. However, the Court's disclosure-related campaign finance case law has also consistently relied on a key premise: While "disclosure requirements may burden the ability to speak, [ ] they impose no ceiling on campaign-related activities and do not prevent anyone from speaking." Citizens United v. Fed. Election Comm'n, 558 U.S. 310, 366, 130 S.Ct. 876, 175 L.Ed.2d 753 (2010) (internal quotation and citation omitted).

The internal logic for this assumption makes sense for direct participants in the political process. Political groups, by design, have an organic desire to succeed at the ballot box. And this ambition generally offsets, at least in part, whatever burdens are posed by disclosure obligations. See Buckley, 424 U.S. at 69-72, 96 S.Ct. 612. In other words, it is not as if disclosure laws do not implicate traditional First Amendment infirmities—as noted, they well may. Rather, the point is that disclosure obligations are ordinarily less detrimental to our commitments to free speech because they do not necessarily censor speech like a direct limit on advocacy does. See The Real Truth About Abortion, Inc. v. Fed. Election Comm'n, 681 F.3d 544, 548-49 (4th Cir. 2012).

But this rationale falters when extended to neutral third-party platforms that view political ads no differently than any other. For sure, platforms are obviously attentive to what their advertisers are saying; the Boston Red Sox are unlikely to accept ads from a group extolling the virtues of the New York Yankees. Yet the predominant purpose of hosting ads is to raise revenue. And a core problem with Maryland's law is that it makes certain political speech more expensive to host than other speech because compliance costs attach to the former and not to the latter. Accordingly, when election-related political speech brings in less cash or carries more obligations than all the other advertising options, there is much less reason for platforms to host such speech. See City of Los Angeles v. Alameda Books, Inc., 535 U.S. 425, 466, 122 S.Ct. 1728, 152 L.Ed.2d 670 (2002) (Souter, J., dissenting).

Disclosure obligations applied to neutral-third party platforms are thus, from a First Amendment perspective, different in kind from conventional campaign finance regulations. First, platform-based campaign finance regulations like the one here make it financially irrational, generally speaking, for platforms to carry political speech when other, more profitable options are available. Second, platform-based campaign finance regulations create freestanding legal liabilities and compliance burdens that independently deter hosting political speech. For example, to avoid the Act's various sanctions, e.g., Md. Code Ann., Elec. Law § 13-405.1(b)(4), the Publishers here have claimed that they would have to acquire new software for data collection; publish additional web pages; and disclose proprietary pricing models. See, e.g., J.A. 41, 52, 58, 68-69, 76 (declarations of Publishers). Faced with this headache, there is good reason to suspect many platforms would simply conclude: Why bother?

In fact, the short history of Maryland's law shows that these chilling effects are not theoretical. Google, for instance, has already stopped hosting political advertisements in the state. See Michael Dresser, Google no longer accepting state, local election ads in Maryland as result of new law, The Balt. Sun (June 29, 2018). And several Publishers here have avowed that they will have to do the same if the Act is enforced against them. E.g., J.A. 77 ("Because of the burdens and potential liability imposed by the Act, many of our members are seriously considering refusing all digital political advertisements.") (declaration of Maryland-Delaware-D.C. Press Association); see also J.A. 42, 53, 59, 64, 70 (declarations of individual Publishers). Additionally, a candidate for Maryland's House of Delegates has alleged that Google's drop-off from political advertising harmed his campaign, and that he and other candidates for local and state elections would find it even more difficult to communicate with voters if newspaper websites followed suit. J.A. 150-154. All told, practice confirms what common sense would predict: While ordinary campaign finance disclosure requirements do not "necessarily reduce[ ] the quantity of expression," Buckley, 424 U.S. at 19, 96 S.Ct. 612, the same cannot be said for platform-based laws.4

In sum, it is apparent that Maryland's law creates a constitutional infirmity distinct from garden-variety campaign finance regulations. The First Amendment guards against "any action of the government by means of which it might prevent such free and general discussion of public matters as seems absolutely essential to prepare the people for an intelligent exercise of their rights as citizens." Grosjean v. Am. Press Co., 297 U.S. 233, 249-50, 56 S.Ct. 444, 80 S.Ct. 660 (1936) (quoting 2 Cooley's Constitutional Limitations 886 (8th ed. 1927)). Government policies that foreclose channels for political speech or simply crowd out too much political speech therefore pose especially serious constitutional dangers. See, e.g., Watchtower Bible & Tract Soc'y of N.Y., Inc. v. Vill. of Stratton, 536 U.S. 150, 167-68, 122 S.Ct. 2080, 153 L.Ed.2d 205 (2002). Because political actors and neutral third-party platforms operate under markedly different incentives, the consequences of a disclosure law vary starkly depending on where its burdens are placed. And when the onus is placed on platforms, we hazard giving government the ability to accomplish indirectly via market manipulation what it cannot do through direct regulation—control the available channels for political discussion. See Reed v. Town of Gilbert, ___ U.S. ___, 135 S.Ct. 2218, 2230, 192 L.Ed.2d 236 (2015).

B.

The First Amendment cautions that attend the Maryland Act are compounded by its application to the class of plaintiffs in this action. The Supreme Court has made clear that when the government tries to interfere with the content of a newspaper or the message of a news outlet, the constitutional difficulties mount. Miami Herald Publ'g Co. v. Tornillo, 418 U.S. 241, 258, 94 S.Ct. 2831, 41 L.Ed.2d 730 (1974). Maryland's law attempts to do just that. As noted, the Act forces news outlets to publish certain information on their websites and, if they fail to do so, empowers the state to seek a court order to have content pulled from these platforms. This brings with it yet another First Amendment infirmity.

For one, Maryland's law "intru[des] into the function of editors" and forces news publishers to speak in a way they would not otherwise. Tornillo, 418 U.S. at 258, 94 S.Ct. 2831. The First Amendment does not just protect a news outlet's editorial perspective or the way its beat reporters cover a given campaign or policy initiative. Rather, because the integrity of the newsroom does not readily permit mandated interaction with the government, the First Amendment applies in full force to all "news, comment, and advertising." Id. The Court has accordingly explained that "the simple selection of a paid noncommercial advertisement for inclusion in a daily paper" falls "squarely within the core of First Amendment security" just as much as any other piece of content. Hurley v. Irish-American Gay, Lesbian & Bisexual Grp. of Boston, 515 U.S. 557, 570, 115 S.Ct. 2338, 132 L.Ed.2d 487 (1995) (internal citation omitted). To put a finer point on it, the very "choice of material to go into a newspaper" garners independent constitutional protection, even if publishers "would not be forced to forgo publication of news or opinion" in the process. Tornillo, 418 U.S. at 258, 94 S.Ct. 2831.

Accordingly, Maryland's assurance that the publication requirement mandates little more than the disclosure of "a line or two of factual information" is of little solace. Appellant's Opening Brief at 53. The Supreme Court has emphasized that there is no constitutional difference between "compelled statements of opinion" and "compelled statements of fact" because "either form of compulsion burdens protected speech." Riley v. Nat'l Fed'n of the Blind of N.C., Inc., 487 U.S. 781, 797-98, 108 S.Ct. 2667, 101 L.Ed.2d 669 (1988) (internal quotation omitted). This because whenever the government "compel[s] individuals to speak a particular message," it "alter[s] the content of their speech." Nat'l Inst. of Family and Life Advocates v. Becerra, ___ U.S. ___, 138 S.Ct. 2361, 2371, 201 L.Ed.2d 835 (2018) (internal quotation omitted). Plainly, the "general rule[ ] that the speaker has the right to tailor the speech[ ] applies not only to expressions of value, opinion, or endorsement, but equally to statements of fact the speaker would rather avoid." Hurley, 515 U.S. at 573, 115 S.Ct. 2338.

Maryland's argument that the Act applies only to "advertisements that publishers have already chosen to accept" is also off the mark. Appellant's Opening Brief at 21. The state suggests that its law imposes a minimal hardship on impacted parties because they opted-in to the arrangement and retain the ability to opt-out by not posting qualifying political ads. Oral Arg. at 14:47-15:46, The Washington Post v. McManus (No. 19-1132). But the First Amendment does not condone this brand of a "bitter with the sweet" rationale. See, e.g., Sorrell v. IMS Health Inc., 564 U.S. 552, 573-74, 131 S.Ct. 2653, 180 L.Ed.2d 544 (2011). And for good reason: After all, another way of saying "opt out" is "stop speaking." Indeed, when a private entity, let alone a newspaper, decides to host political speech, its First Amendment protections are at their apex. To contend that news outlets forgo some of their free speech rights by accepting political speech turns the First Amendment on its head and does nothing to salvage the Act.

The inspection requirement implicates the same set of concerns and more. Not only does it compel the Publishers to turn over information to state regulators, it also brings the state into an unhealthy entanglement with news outlets. The core problem with this provision of the Act is that it lacks any readily discernable limits on the ability of government to supervise the operations of the newsroom. As it stands now, the Act requires news outlets to provide Maryland with no less than six separate disclosures, each assertedly justified by the state's interests in informing the electorate and enforcing its campaign finance laws. But with its foot now in the door, Maryland has offered no rationale for where these incursions might end. Today the state asks for information about the targeted audience; tomorrow perhaps the names and addresses of all officers or corporate affiliates of the ad purchaser; the day after the identities of donors to those purchasers. The absence of space and time limitations on the internet as opposed to traditional media outlets makes the absence of a limiting principle all the more unsettling. Without clear limits, the specter of a broad inspection authority, coupled with an expanded disclosure obligation, can chill speech and is a form of state power the Supreme Court would not countenance. See Nat'l Ass'n for Advancement of Colored People v. State of Ala. ex rel. Patterson, 357 U.S. 449, 462, 78 S.Ct. 1163, 2 L.Ed.2d 1488 (1958).

Maryland tries to avoid these infirmities by analogizing the Act to the third-party disclosure obligations that have been upheld in the broadcasting context. But this is an inapt comparison. The broadcast industry has always held a distinctive place in First Amendment law on the ground that "[b]roadcast frequencies are a scarce resource [that] must be portioned out among applicants." FCC v. League of Women Voters of Cal., 468 U.S. 364, 377, 104 S.Ct. 3106, 82 L.Ed.2d 278 (1984) (quoting Columbia Broad. Sys., Inc., v. Democratic Nat'l Comm., 412 U.S. 94, 101, 93 S.Ct. 2080, 36 L.Ed.2d 772 (1973)). That is, because broadcast licensees are given a federal grant to operate one of these limited channels, the Court has given the government wider latitude in regulating what is said on them. Red Lion Broad. Co. v. FCC, 395 U.S. 367, 399-400, 89 S.Ct. 1794, 23 L.Ed.2d 371 (1969). This justification, however, is inapposite for the virtually limitless canvas of the internet. And for that reason, "the vast democratic forums of the Internet [have never] been subject to the type of government supervision and regulation that has attended the broadcast industry." Reno v. Am. Civil Liberties Union, 521 U.S. 844, 868-69, 117 S.Ct. 2329, 138 L.Ed.2d 874 (1997). In short, what goes for broadcasters is too much a product of their technical circumstances to serve as a template for state regulation writ large. Turner Broad. Sys., Inc. v. FCC, 512 U.S. 622, 638-39, 114 S.Ct. 2445, 129 L.Ed.2d 497 (1994); see McConnell v. Fed. Election Comm'n, 540 U.S. 93, 234-38, 124 S.Ct. 619, 157 L.Ed.2d 491 (2003).

Moreover, compelling broadcasters to speak is not the same as altering the content of a news product. See Turner, 512 U.S. at 655, 114 S.Ct. 2445; see also Eugene Volokh, The Law of Compelled Speech, 97 Tex. L. Rev. 355, 361-66 (2018). The Court's decision in Hurley helps explain how this difference in medium informs First Amendment analysis. There, the Court held that Massachusetts could not force a private organization to include a gay-rights group in its St. Patrick's Day parade. Hurley, 515 U.S. at 580-81, 115 S.Ct. 2338. It based its decision on the view that a parade is a discrete expressive product such that any government intervention would necessarily conflict with the "general rule[ ] that the speaker has the right to tailor the speech." Id. at 573, 115 S.Ct. 2338. In so doing, the Court distinguished its decision in Turner Broadcasting, which presented a similar fact pattern on first blush. There, the Court evaluated the FCC's "must carry" rule, which required cable networks to carry certain channels. Id. at 576, 115 S.Ct. 2338. The Court upheld this regulation, however, on the ground that cable networks were composed of "individual, unrelated segments that happen to be transmitted together for individual selection by members of the audience" and thus lacked an expressive character that would be inherently impaired by the FCC's regulation. Id. at 576, 115 S.Ct. 2338; see Rumsfeld v. Forum for Acad. and Institutional Rights, Inc., 547 U.S. 47, 60, 126 S.Ct. 1297, 164 L.Ed.2d 156 (2006). At heart, the same state action —the forced inclusion of particular content—created a fatal First Amendment problem in one setting (parades) but was permissible in another (cable).

The same intuition governs here. The Court has made clear that news products are of a part with parades, political leaflets, fundraising pitches, or similar expressive endeavors. E.g., McIntyre v. Ohio Elections Comm'n, 514 U.S. 334, 348, 115 S.Ct. 1511, 131 L.Ed.2d 426 (1995); Riley, 487 U.S. at 797-98, 108 S.Ct. 2667. This is not to say that such mediums are immune from regulation. Rather, the point is that the integrity of these expressive commodities is presumptively at risk as soon as the government compels any alteration to their message.

IV.

The above collection of First Amendment infirmities underscores the difficulties Maryland faces in squaring its law with the First Amendment. The preceding discussion has illustrated the imperative of some form of heightened judicial scrutiny. We decline, though, to do more than is needed to resolve the case before us. On that front, we decline to decide whether strict or exacting scrutiny should apply to a disclosure law like the one here because we hold that the Act fails even the more forgiving standard of exacting scrutiny. To be sure, neither standard is deferential—both place high hurdles before the government. But strict scrutiny, in practice, is virtually impossible to satisfy, while exacting scrutiny is merely difficult. To declare an invariable reviewing standard of strict scrutiny would be an attempt to script the future in the face of novel challenges to electoral integrity that we know not of and cannot foresee. And because the disparity between Maryland's chosen means and purported ends is so pronounced, we need only apply the exacting scrutiny standard.

Under exacting scrutiny, there must be a "substantial relation" between an "important" government interest and "the information required to be disclosed." Buckley v. Valeo, 424 U.S. 1, 64-66, 96 S.Ct. 612, 46 L.Ed.2d 659 (1976). To start, there is no doubt that Maryland has asserted important government interests to sustain the Act. These interests fall into two buckets: one specific and one general. First, Maryland has principally justified the Act on the ground that it will help deter foreign interference in its elections. Washington Post v. McManus, 355 F.Supp.3d 272, 298-99 (D. Md. 2019). Election integrity is undoubtedly an important state interest. See, e.g., Indep. Inst. v. Fed. Election Comm'n, 216 F.Supp.3d 176, 191 (D.D.C. 2016), aff'd., ___ U.S. ___, 137 S.Ct. 1204, 197 L.Ed.2d 243 (2017). Second, Maryland has also claimed a set of secondary interests that are traditionally associated with disclosure-based laws: informing the electorate, deterring corruption, and enforcing the state's campaign finance laws. Washington Post, 355 F. Supp. 3d at 302-03. These too can be "sufficiently important" to justify certain campaign finance regulations. Buckley, 424 U.S. at 66-68, 96 S.Ct. 612.

But the fact that an interest is "important" in the abstract does not end the analysis. "In the First Amendment context, fit matters." McCutcheon v. Fed. Election Comm'n, 572 U.S. 185, 218, 134 S.Ct. 1434, 188 L.Ed.2d 468 (2014) (plurality opinion). Specifically, even under exacting scrutiny, a commitment to free speech requires governments to "employ[ ] not necessarily the least restrictive means but... a means narrowly tailored to achieve the desired objective." Id. (quoting Bd. of Trs. of State Univ. of N.Y. v. Fox, 492 U.S. 469, 480, 109 S.Ct. 3028, 106 L.Ed.2d 388 (1989)). And on this front, the Maryland Act falls short.

For one, Maryland's law does surprisingly little to further its chief objective of "combat[ting] foreign meddling in the state's elections." Washington Post, 355 F. Supp. 3d at 299. In essence, the Act regulates paid ads that "relate[ ] to" a candidate or ballot question and are placed on online platforms operating in Maryland. Md. Code Ann., Elec. Law § 1-101(k)(1). This approach has two flaws. First, even by Maryland's own reckoning, foreign nationals rarely, if ever, relied on paid content to try to influence the electorate. Instead, as the state concedes, "Russian influence was achieved `primarily through unpaid posts'" on social media. See Appellant's Opening Brief at 6 (quoting J.A. 144). The Act leaves this primary mechanism completely unaddressed. Second, the Act also fails to regulate even the narrow band of paid content used by foreign nationals. Of the small percentage of foreign-placed paid ads that reached Maryland voters, the vast majority did not urge people to choose a certain candidate or support a specific ballot initiative. Rather, their chief focus was to rouse passions on divisive questions such as those surrounding race or gun rights. J.A. 144-45; see also J.A. 294-95. Put otherwise, they were not "campaign material," as defined by the Act, and therefore are not affected by the Act.

Maryland seems to grant that the Act fails to combat the lion's share of tactics used by foreign operatives in 2016. But the state seeks an indulgence, claiming the Act is the best it can do in light of "constitutional questions as to whether [Maryland] can regulate the unpaid speech of anonymous commenters on the Internet." Appellant's Opening Brief at 56. This "something is better than nothing" argument, however, is unavailing. Indeed, Maryland has offered no support for the proposition that courts should place a thumb on the exacting scrutiny scale for laws that are the "least unconstitutional" among available options. Nor could it. The First Amendment makes plain that any law burdening free speech must rise or fall on its own merits.

What's more, while the Act strikes too narrowly in some respects, it also strikes too broadly in others. Two features stand out: the decision to include the press and the choice to draw even quite-small platforms within the Act's ambit.

First, Maryland has marshalled scant evidence to justify applying its Act to the press. As noted, when a government compels speech from news outlets, let alone political speech, it risks contravening the First Amendment many times over. To justify these intrusions, states must meaningfully demonstrate that a given law is impelled by the facts on the ground. Here, Maryland has not done so. In fact, the state "has not been able to identify so much as a single foreign-sourced paid political ad that ran on a news site, be it in 2016 or at any other time." Washington Post, 355 F. Supp. 3d at 301. Of course, states are owed a degree of deference in how they choose to pursue important governmental interests. But deference to ends does not obviate the need for concrete evidence showing the chosen means warrant the accompanying First Amendment burdens.

Maryland advances, however, a prophylactic rationale. On its view, the state "was not required to wait for foreign-sourced ads to appear via a particular method on plaintiffs' websites before acting prophylactically to prevent such misconduct." Appellant's Reply Brief at 17. And in support of this view, Maryland points to evidence that Russian operatives infiltrated Google's "DoubleClick" paid ad network during the 2016 election and that some newspaper websites, including those of some Publishers, use this network. As such, Maryland says there are sufficient grounds for it to regulate newspaper sites in anticipation of this possible new front in foreign interference.

This preventative justification fails to pass First Amendment muster. The Supreme Court has made clear that, when free speech values are at stake, states must supply rationales that are "far stronger than mere speculation about serious harms." Bartnicki v. Vopper, 532 U.S. 514, 531, 121 S.Ct. 1753, 149 L.Ed.2d 787 (2001) (quoting United States v. Treasury Emps., 513 U.S. 454, 475, 115 S.Ct. 1003, 130 L.Ed.2d 964 (1995)). The First Amendment does not permit states to broadly conjure hypotheticals in support of expressive burdens. If any evidence—no matter how indirect or futuristic—could concretize a purported harm, speech would be rendered substantially more vulnerable. Applied here, indirect evidence of Russian interference gets Maryland only so far. Without direct evidence (or anything close to it) of meddling on news sites, Maryland has failed to show that this purported threat is likely or imminent enough to justify the Act's intrusive preventative measures.

Second, the Act is also too broad because it fails to distinguish between platforms large and small. The Maryland law sweeps the spectrum of websites, covering both The Washington Post and Carroll County Times, as well as their equivalents in every industry operating in the state. Specifically, the Act applies to each "public-facing website, web application, or digital application, including a social network, ad network, or search engine, that: has 100,000 or more unique monthly United States visitors or users...." Md. Code Ann., Elec. Law § 1-101(dd-1). The law thus kicks in no matter how susceptible a website may be to foreign meddling or how influential it has been in a given election cycle.

As above, Maryland has failed to provide sufficient evidence to justify painting with such a broad brush. For instance, as the district court rightly observed, the clear bulk of foreign meddling took place on websites like Facebook, Instagram, or other social media platforms that each garner millions of visitors per month. See Washington Post, 355 F. Supp. 3d at 301. But the Act applies equally to The Cecil Whig and The Cumberland Times-News as it does to Facebook—notwithstanding the marked disparities between their respective reaches and past histories with foreign election interference. This is not to say that state regulations must always parse platforms based on their size. Rather, the point is that, in light of the First Amendment burdens here, Maryland must muster some concrete proof to justify the Act's capacious scope.

On a related note, the parties have discussed at length whether the Act's provisions duplicate disclosure requirements already in place under Maryland law. While Maryland is quick to point out certain discrete pieces of information that the Act requires for the first time, the Publishers respond by underscoring the substantial similarity between what is asked for by the Act and what is already required of ad purchasers operating in the state. But this is all mostly beside the point. The Act must stand or fall on its own merits, independent of whether it overlaps with other parts of Maryland's legal landscape. The judgment we have to make is whether this Act is or is not a constitutional one. And all the duplication in the world would not by itself condemn it, nor would the fact that the Act is wholly unique serve alone to sustain it. That being said, the duplication discussion does serve to illustrate that much of what Maryland wishes to accomplish through the Act can be done through better fitting means. Indeed, it seems plain that Maryland can apply the Act's substantive provisions to ad purchasers directly, rather than neutral third-party platforms, or expand its existing campaign finance laws to cover donors to the same effect.5

Taking a step back, it is important to bear in mind that while First Amendment analyses can get bogged down in terminology and tier-chasing, the touchstone for exacting scrutiny is whether there is "a fit that is not necessarily perfect, but reasonable." McCutcheon, 572 U.S. at 218, 134 S.Ct. 1434 (plurality opinion) (quoting Bd. of Trs. of State Univ. of N.Y. v. Fox, 492 U.S. 469, 480, 109 S.Ct. 3028, 106 L.Ed.2d 388 (1989)). For the reasons above, Maryland has failed to clear this bar. The Act, at heart, burdens too much and furthers too little, and this one-sided tradeoff falls short of what the First Amendment requires.

V.

Within our federal system, states are tasked with striking a difficult balance in administering elections. On the one hand, the marketplace of ideas resists governmental regulation. The First Amendment guarantees that all citizens shall be free to speak their piece on the issues of the day, and that government cannot meddle in the debate that takes place among the governed. On the other hand, for a democracy to reach its full potential, intervention is occasionally necessary to promote transparency, facilitate an educated populace, and deter corruption. As Justice Brandeis put it: "Sunlight is said to be the best of disinfectants." L. Brandeis, Other People's Money 62 (National Home Library Foundation ed. 1933). And sunlight also can serve First Amendment values.

The changing nature of elections and the novel technological challenges accompanying them have made the states' managerial tasks more difficult. How states choose to carry out their responsibilities has long merited our respect. But that respect has bounds—and here, Maryland has crossed them. Despite its admirable goals, the Act reveals a host of First Amendment infirmities: a legislative scheme with layer upon layer of expressive burdens, ultimately bereft of any coherent connection to an offsetting state interest of sufficient import. While we credit the aims of Maryland's legislators, we can in no way approve the state's chosen means. The most basic First Amendment principles compel as much.

For the foregoing reasons, the judgment of the district court is

AFFIRMED.