Maplebear Inc., doing business as Instacart, Intervenor-Plaintiff, v. LLC (2026)

OPINION

The City of Seattle enacted Ordinance 126878 (the “Ordinance”) in 2023. 1 It added a new chapter 8.40 to the Seattle Municipal Code, titled the “App-Based Worker Deactivation Rights Ordinance.” Applying to “network companies” that provide a platform for customers to hire temporary workers, like delivery drivers, the Ordinance prohibits “unwarranted deactivations” of these app-based workers' accounts. Seattle, Wash., Ordinance 126878, § 8.40 (Aug. 14, 2023). The law is Seattle's attempt to address what it perceives to be the instability of the gig economy and the consequent vulnerability of delivery workers.

Uber Technologies, Inc., including its wholly owned subsidiary Portier, LLC (“Uber”), and Maplebear Inc. (“Instacart”) (collectively, “Plaintiffs”) are two network companies that rely on app-based workers for their businesses. They claim that the Ordinance violates their constitutional rights in two ways: it impermissibly infringes on their protected speech, and it is void for vagueness. They appeal the district court's denial of their motion for a preliminary injunction to enjoin enforcement of the Ordinance against them.

We conclude that the Ordinance does not regulate speech subject to protection under the First Amendment because the Ordinance regulates nonexpressive conduct—the unwarranted deactivation of worker accounts. Even if the Ordinance is interpreted to regulate speech, that speech would be commercial speech, and the Ordinance's regulation of it would overcome the applicable lower level of First Amendment scrutiny. As for the vagueness challenge, we conclude that the Ordinance is not unconstitutionally vague, because a person of ordinary intelligence would understand its meaning. Accordingly, the district court did not abuse its discretion when it denied Plaintiffs' request for a preliminary injunction, and we affirm.

I. Background

Uber and Instacart are large companies in the growing gig economy. Plaintiffs and other companies like them connect workers with customers, then take a cut of payments made by the customers. Relevant to the Ordinance, Uber and Instacart act largely as delivery services. (The Ordinance does not apply to certain categories of app-based workers, like rideshare drivers, which is what many people may immediately think of as “Uber.” Ordinance § 8.40.020.) Their workers pick up orders from restaurants, or retailers, or even other customers on the app. The workers also can operate as personal shoppers by going to businesses like grocery stores, browsing for items, and delivering those items to customers.

Both workers and customers access those services through accounts on apps built by the network companies. While Uber and Instacart are the parties to bring this challenge, any app-based worker regulation affects more than just the users of those two platforms. As of 2021, sixteen percent of adults in the United States have earned money via an online gig platform. Monica Anderson et al., The State of Gig Work in 2021, Pew Rsch. Ctr. (Dec. 8, 2021), https://www.pewresearch.org/internet/2021/12/08/the-state-of-gig-work-in-2021/ (last visited Oct. 29, 2025).

Seattle's Ordinance prohibits deactivations of app-based workers for reasons that its drafters deemed unwarranted. Ordinance § 8.40.060.A. Plaintiffs do not challenge that prohibition but instead challenge the constitutionality of certain provisions of the Ordinance, in particular section 8.40.050, titled “Deactivation requirements.” That section requires (1) that a network company “inform” any app-based worker “in writing” of the company's deactivation policy and (2) that the deactivation policy be “reasonably related” to the company's “safe and efficient operations.” Id. § 8.40.050.A.1–2.

The Ordinance includes a list of eight example policies that are not reasonably related to safety and efficiency. Id. § 8.40.050.A.2.a–h. The list includes policies that would result in a deactivation “based solely on a quantitative metric derived from aggregate customer ratings” or “based on the results of a background check, consumer report, driver record, or record of traffic infractions, except in cases of egregious misconduct or where required by other applicable law.” Id. § 8.40.050.A.2.e, h. 2

The Ordinance applies only to internal communications with app-based workers. It does not require that the companies communicate any deactivation policies to the general public. Nor does it require that the companies communicate their agreement with such policies. The companies may describe the policies as being required by the Ordinance.

Seattle's stated purpose for enacting the Ordinance is to “protect[ ] and promote[ ] public health, safety, and welfare.” Id. § 1.C. Minimum labor standards are recognized as helping both employers, by improving worker performance and quality, and workers, by increasing job stability. Id. § 1.E, W. But network companies' growing reliance on algorithms leads, Seattle says, to “arbitrary evaluations and unwarranted deactivations.” Id. § 1.H. Establishing “clear performance expectations,” limitations on deactivations predicated on quantitative metrics and possibly inaccurate background checks, and a legal right to challenge unwarranted deactivations are all attempts to regulate reliance on automatic, algorithmic determinations of worker-eligibility. See id. § 1.I–L, P, U.

Uber filed this action seeking a temporary restraining order and preliminary injunction against the Ordinance shortly before it was set to take effect. It alleged three claims, two of which are pursued in this appeal: that the Ordinance compels speech in violation of the First Amendment, and that it is unconstitutionally vague.  3 Instacart moved to intervene and, in doing so, agreed to forego independent briefing on the temporary restraining order before the district court.

The district court held that the Ordinance regulates the deactivation of worker accounts, with any effects on speech or expression being “incidental to its worker-related conduct goals.” The court concluded that the Ordinance was not void because the term “reasonable” was not vague, and any discrepancies in interpretation were rooted in a “fundamental disagreement” between Uber and Seattle over what policies are reasonably related to safe and efficient operations. After determining that Uber did not show a likelihood of establishing any constitutional violation, the district court found neither irreparable harm nor a balance of equities or public interest tipping in Uber's favor. It denied the motion for a temporary restraining order and preliminary injunction.

Uber and Instacart appealed. A stay pending final resolution of the appeals, entered by the district court based on a stipulation of the parties, is currently in place.

We have jurisdiction under 28 U.S.C. § 1292(a)(1). A district court's denial of a preliminary injunction is reviewed for abuse of discretion, a “limited and deferential” standard. Harris v. Bd. of Supervisors, 366 F.3d 754, 766 (9th Cir. 2004); see also X Corp. v. Bonta, 116 F.4th 888, 897 (9th Cir. 2024). The underlying issues of law that led to that decision, such as whether the Ordinance compels speech or merely regulates conduct, are reviewed de novo. See Harris, 366 F.3d at 760.

II. First Amendment claims

Uber and Instacart argue that Seattle's Ordinance compels pure, content-based speech in violation of the First Amendment in two ways. First, the notice requirement compels speech by forcing the companies to communicate their deactivation policies and decisions in writing and make them accessible to workers. Second, the deactivation standards regulate that speech's content by requiring that the companies' policies be, in the words of the Ordinance, “reasonably related” to “safe and efficient operations.”

A. Conduct

As the parties bringing the challenge, Plaintiffs “bear[ ] the burden ‘to demonstrate that the First Amendment even applies.’ ” B & L Prods., Inc. v. Newsom, 104 F.4th 108, 112 (9th Cir. 2024) (quoting Clark v. Cmty. for Creative Non-Violence, 468 U.S. 288, 293 n.5, 104 S.Ct. 3065, 82 L.Ed.2d 221 (1984)), cert. denied, ––– U.S. ––––, 145 S. Ct. 1958, 221 L.Ed.2d 738 (2025). They fail to do so. The Ordinance regulates conduct, not speech.

While “freedom of speech prohibits the government from telling people what they must say,” Rumsfeld v. F. for Acad. & Institutional Rts., Inc., 547 U.S. 47, 61, 126 S.Ct. 1297, 164 L.Ed.2d 156 (2006), not every law that results in a burden on speech is subject to scrutiny as a regulation of speech. Restrictions on conduct are subject to scrutiny only if the conduct includes a “significant expressive element that drew the legal remedy in the first place.” Arcara v. Cloud Books, Inc., 478 U.S. 697, 706, 106 S.Ct. 3172, 92 L.Ed.2d 568 (1986). Here, the conduct that draws the remedy is the unwarranted deactivation of worker accounts. When conduct is nonexpressive, as is the case here, it is not subject to First Amendment scrutiny. B & L Prods., 104 F.4th at 110; see also Sorrell v. IMS Health Inc., 564 U.S. 552, 567, 131 S.Ct. 2653, 180 L.Ed.2d 544 (2011).

When determining whether a law regulates sufficiently expressive conduct, a court may consider both the “inevitable effect of a [law] on its face” and the law's “stated purpose.” HomeAway.com, Inc. v. City of Santa Monica, 918 F.3d 676, 685 (9th Cir. 2019) (quoting Sorrell, 564 U.S. at 565, 131 S.Ct. 2653). We consider both in turn.

The inevitable effect of the Ordinance is to ensure that network companies devise policies that do not result in unwarranted deactivations of app-based workers' accounts. At its core, then, the Ordinance regulates a business agreement between two parties: the network company and the worker who uses the company's platform to obtain gigs. A business agreement between two parties is not conduct with a significant expressive element. See B & L Prods., 104 F.4th at 114 (holding that acceptance of an offer to sell firearms on state property is not conduct with a significant expressive element); HomeAway.com, 918 F.3d at 685 (holding that booking a short-term rental through an online platform is not conduct with a significant expressive element); Int'l Franchise Ass'n, Inc. v. City of Seattle, 803 F.3d 389, 408 (9th Cir. 2015) (“A business agreement or business dealings between a franchisor and a franchisee is not conduct with a ‘significant expressive element.’ ”); see also Janus v. Am. Fed'n of State, Cnty., & Mun. Emps., Council 31, 585 U.S. 878, 946, 138 S.Ct. 2448, 201 L.Ed.2d 924 (2018) (Kagan, J., dissenting) (noting that the Supreme Court “has rejected all attempts” to make a First Amendment issue out of basic employment matters).

The Ordinance's stated purpose is to “protect[ ] and promote[ ] public health, safety, and welfare by establishing protections against unwarranted deactivations for app-based workers.” Ordinance § 1.C. It does so by requiring network companies to inform app-based workers of what actions might lead to deactivation, Ordinance § 8.40.050.A.1, presumably so they can conform their conduct to the companies' expectations, and by prohibiting policies that might lead to unwarranted deactivation, id. § 8.40.050.A.2. The Ordinance's goal is to keep app-based workers informed and employed while keeping the public safe. See id. § 1. Nothing in the text of the Ordinance suggests an intent to restrict speech. 4

Plaintiffs urge us to look beyond the law's stated purpose and to focus solely on the contested provisions. They argue that subsections 8.40.050.A.1–2 do not prevent them “from deactivating courier accounts on any ground, and the fact that a separate provision addresses account deactivations does not mean no First Amendment scrutiny applies.” But this reading ignores the placement of these provisions “in the context of the Ordinance as a whole.” Yim v. City of Seattle, 63 F.4th 783, 792 (9th Cir. 2023) (quoting First Resort, Inc. v. Herrera, 860 F.3d 1263, 1274 (9th Cir. 2017)). To read these provisions independently from the Ordinance would be to orphan them from their purpose: preventing unwarranted deactivations. 5

The fact that the Ordinance, in Plaintiffs' words, necessarily “compel[s] and dictate[s] the content of a written communication” does not transform a law regulating nonexpressive activity into one that infringes speech.  6 “[C]ompelled changes in conduct—which might incidentally compel changes in speech—are not reviewed as content-based speech restrictions.” Olympus Spa v. Armstrong, 138 F.4th 1204, 1216 (9th Cir. 2025). Plaintiffs can still voice their displeasure with the Ordinance's standards, even while complying with them. Nothing prevents Plaintiffs from releasing statements, posting notices, or engaging the press in discussions about what grounds for deactivation they believe are reasonably related to the safety and efficiency of their businesses, including those grounds that the Ordinance explicitly identifies as not related to safety and efficiency. They can even include in the communications to their workers, alongside the required statements of their deactivation policies, a statement that they disagree with the Ordinance's requirements and the City's assessment of what is reasonably related to safe and efficient operations. That is their right. But it is not their right to disregard a valid economic regulation that targets neither speech nor expressive conduct that is protected by the First Amendment. See Int'l Franchise Ass'n, 803 F.3d at 408.

Like a law that prohibits employers from discriminating on account of race, which requires removal of signs and advertisements saying “White Applicants Only,” the incidental speech requirement here “hardly means that the law should be analyzed as one regulating the employer's speech rather than conduct.” Rumsfeld, 547 U.S. at 62, 126 S.Ct. 1297; see also Olympus Spa, 138 F.4th at 1216 (holding that a law that prevented a spa from denying admission to transgender women, which required the spa to change its conflicting admissions policy, did not trigger First Amendment scrutiny). Instead, the Ordinance is like any well-established anti-discrimination or workplace safety law: it requires private companies to conform their internal policies to minimum standards and to communicate those policies. The Ordinance thus ensures that Plaintiffs are unable to carry out an unlawful course of conduct by creating and disseminating illegal deactivation policies. Rumsfeld, 547 U.S. at 62, 126 S.Ct. 1297. “When ‘the only inevitable effect, and the stated purpose’ of a statute is to regulate nonexpressive conduct, our inquiry is essentially complete.” B & L Prods., 104 F.4th at 116 (quoting HomeAway.com, 918 F.3d at 685). The Ordinance governs nonexpressive conduct. Any burden on speech is incidental.

B. Commercial Speech

In the alternative, Seattle argues that, if the Ordinance regulates speech, then it is commercial speech subject to lower scrutiny under the First Amendment. Plaintiffs insist that the regulated speech is pure speech about controversial issues and not commercial speech.

We agree with Seattle that even if the Ordinance regulates speech, it is speech that “simply propose[s] a commercial transaction.” 7 Va. State Bd. of Pharmacy v. Va. Citizens Consumer Council, Inc., 425 U.S. 748, 760, 96 S.Ct. 1817, 48 L.Ed.2d 346 (1976) (quoting Bigelow v. Virginia, 421 U.S. 809, 822, 95 S.Ct. 2222, 44 L.Ed.2d 600 (1975)). We have already held that speech related to hiring people who “perform temporary work” is commercial speech. Valle Del Sol Inc. v. Whiting, 709 F.3d 808, 817–18 (9th Cir. 2013); see also Pittsburgh Press Co. v. Pittsburgh Comm'n on Hum. Rels., 413 U.S. 376, 387–88, 93 S.Ct. 2553, 37 L.Ed.2d 669 (1973) (holding that employment advertisements are commercial speech); Greater Phila. Chamber of Com. v. City of Philadelphia, 949 F.3d 116, 137 (3d Cir. 2020) (holding that speech involving a “proposal of possible employment” is commercial speech).

Our precedents establishing that hiring communications can be commercial speech cover the purported speech at issue here. That speech is communication by Uber or Instacart to temporary app-based workers about the proposed terms of engagement, including the standards for deactivation or termination of those workers' accounts. The Ordinance's challenged provisions regulate how and when a platform may prevent a worker from accessing its service to seek employment. Whatever speech might be regulated by the Ordinance is directly related to employment, however temporary.

Plaintiffs invoke two recent decisions by this court to argue that the Ordinance should be subject to greater scrutiny than that applied to commercial speech. See NetChoice, LLC v. Bonta, 113 F.4th 1101, 1111, 1121 (9th Cir. 2024); X Corp., 116 F.4th at 894, 898. Both cases involved state laws that regulated the speech of social media companies by requiring that they submit reports to the government that identified harmful speech and described what the companies were doing to address the problem. That requirement distinguishes these cases from the one before us. In both NetChoice and X Corp., we held that the reports were not commercial speech subject to intermediate scrutiny, but instead were compelled speech subject to strict scrutiny, because they required businesses to provide their own conclusions about “highly controversial issues of public concern.” NetChoice, 113 F.4th at 1120; see X Corp., 116 F.4th at 901–03. We noted that these reports required businesses to go beyond “opining about their products or services,” NetChoice, 113 F.4th at 1120, and instead communicate their “opinions about and reasons for those policies,” X Corp., 116 F.4th at 901 (emphasis added).

Here, the Ordinance regulates Plaintiffs' deactivation policies and requires communication of the policies to the workers subject to them. Unlike the social media companies—which were required to speak on topics “disconnected from any economic transaction” like hate speech, online harassment, and content harmful to children, NetChoice, 113 F.4th at 1119—Plaintiffs are not subject to a directive to produce a report. Instead, they must modify their deactivation policies, if necessary to comply with the Ordinance, and communicate those policies to their app-based workers.

The reporting requirement differs in another key area. Under both laws at issue in NetChoice and X Corp., social media companies were required to produce communications for an outside party: the government. Here, the Ordinance applies only to internal communications between a company and its app-based workers. It is a modern version of established workplace disclosure laws—it makes no difference whether such a disclosure occurs on bulletin boards, via distribution of employee handbooks, or in an online portal that only app-based workers may access. See CompassCare v. Hochul, 125 F.4th 49, 65 (2d Cir. 2025) (“[S]uch notice requirements are part of ‘a longstanding tradition in this country’ supported by a ‘historical warrant.’ ” (quoting Brown v. Ent. Merchs. Ass'n, 564 U.S. 786, 795, 792, 131 S.Ct. 2729, 180 L.Ed.2d 708 (2011))).

Assuming that the Ordinance does regulate commercial speech, we conclude that the Ordinance satisfies the attending, lower level of scrutiny. Commercial speech is still subject to constitutional protections. We determine what protection that speech receives depending on whether the government is restricting or compelling speech.  8 Am. Beverage Ass'n v. City & County of San Francisco, 916 F.3d 749, 755 (9th Cir. 2019) (en banc).

Under our precedent, we apply one of two separate Supreme Court tests to laws regulating commercial speech. CTIA - The Wireless Ass'n v. City of Berkeley, 928 F.3d 832, 842 (9th Cir. 2019). When the government restricts or prohibits commercial speech, we apply the test established in Central Hudson Gas & Electric Corp. v. Public Service Commission of New York, 447 U.S. 557, 566, 100 S.Ct. 2343, 65 L.Ed.2d 341 (1980). When the government compels disclosure of commercial speech, we apply the standard from Zauderer v. Office of Disciplinary Counsel of the Supreme Court of Ohio, 471 U.S. 626, 651, 105 S.Ct. 2265, 85 L.Ed.2d 652 (1985). Here, because the Ordinance regulates deactivation standards and requires companies to communicate those standards to their workers, the appropriate test is the one articulated in Zauderer.

Subsection 8.40.050.A.1—which requires network companies to inform app-based workers of their deactivation policies—meets the test drawn from Zauderer. “[T]he government may compel truthful disclosure in commercial speech as long as the compelled disclosure is reasonably related to a substantial governmental interest and involves purely factual and uncontroversial information that relates to the service or product provided.” CTIA, 928 F.3d at 842 (citation modified). Additionally, if such compelled disclosure is “unjustified or unduly burdensome,” it might not pass constitutional muster. Zauderer, 471 U.S. at 651, 105 S.Ct. 2265.

“There is no question that protecting the health and safety of [workers] is a substantial governmental interest.” CTIA, 928 F.3d at 845. The provision requires network companies to inform workers of what situations may lead to account deactivation and resulting lost income, which is reasonably related to Seattle's goals. The notice is purely factual and concerns only the service provided, because it contains details of a company's deactivation policies. Because the disclosure does not “force [Plaintiffs] to take sides in a heated political controversy,” id. at 848, or purport to compel speech on—or analogous to—matters we have previously identified to be controversial, see, e.g., id. at 845 (noting abortion as a matter of controversy), it is noncontroversial.  9 Further, the notice requirement can be satisfied by a single webpage and is therefore not unduly burdensome.

Subsection 8.40.050.A.2—which requires that the deactivation policies be “reasonably related” to the companies' “safe and efficient operations”—regulates at most speech related to what Seattle has designated as unlawful activity: the unwarranted deactivations of worker accounts. Any challenge involving the restriction of commercial speech must first satisfy a threshold inquiry that the communication be “neither misleading nor related to unlawful activity.” World Wide Rush, LLC v. City of Los Angeles, 606 F.3d 676, 684 (9th Cir. 2010) (quoting Cent. Hudson, 447 U.S. at 564, 100 S.Ct. 2343). Even if we accept Plaintiffs' contention that the Ordinance regulates commercial speech, it is not an impermissible regulation of that speech. 10

Plaintiffs have not demonstrated that the Ordinance violates the First Amendment.

III. Vagueness claim

A municipal ordinance violates the Fourteenth Amendment if it either fails to provide a “person of ordinary intelligence a reasonable opportunity to know what is prohibited,” or if it is so standardless that it may be applied in an arbitrary or discriminatory manner. Vill. of Hoffman Ests. v. Flipside, Hoffman Ests., Inc., 455 U.S. 489, 498, 102 S.Ct. 1186, 71 L.Ed.2d 362 (1982) (quoting Grayned v. City of Rockford, 408 U.S. 104, 108–09, 92 S.Ct. 2294, 33 L.Ed.2d 222 (1972)). Plaintiffs make much of the fact that “law[s] interfere[ing] with the right of free speech” are subject to a “stringent” application of that test. Id. at 499, 102 S.Ct. 1186. As explained above, however, we reject Plaintiffs' premise that the Ordinance restricts speech. Instead, we are guided by the Supreme Court's standard for economic regulation, which is “subject to a less strict vagueness test ․ because businesses, which face economic demands to plan behavior carefully, can be expected to consult relevant legislation in advance of action.” Id. at 498, 102 S.Ct. 1186.

Plaintiffs argue that the Ordinance's deactivation standard provision in subsection 8.40.050.A.2 is void for two reasons. First, the word “reasonable” does not supply adequate guidance to follow the law. Second, the examples of what is not reasonably related to safety and efficiency are so contradictory as to be more confusing than clarifying. We are unpersuaded by both arguments.

The term “reasonable” is a well-established standard that is widely employed in both statutory and common law. As the district court pointed out, Plaintiffs must satisfy a reasonableness inquiry to succeed on their vagueness claim. Additionally, the Ordinance's text supplements the term by delineating between what is and is not a permissible reason to deactivate an account. See First Resort, 860 F.3d at 1274 (“[I]mprecise terms may avoid vagueness problems when used in combination with terms that provide sufficient clarity.” (quoting Human Life of Wash., Inc. v. Brumsickle, 624 F.3d 990, 1021 (9th Cir. 2010))). For example, the bar on deactivations based on customer ratings is limited by the word “solely.” Ordinance § 8.40.050.A.2.e. (“Any policy that would result in a deactivation based solely on a quantitative metric derived from aggregate customer ratings of an app-based worker's performance[.]” (emphasis added)).

Seattle passed the Ordinance out of its concern that companies like Uber and Instacart relied on algorithms to deactivate accounts automatically. Its concern was magnified by the City's belief that customer ratings are “highly likely” to be influenced by bias against workers of certain races and ethnicities. Id. § 1.L. For those reasons, the Ordinance dictates that an account deactivation may not be based “solely” on customer ratings. Id. § 8.40.050.A.2.e. Deactivation based on more than that—for instance, including review of comments that accompany customer ratings—is not prohibited.

Plaintiffs may be right that a person of ordinary intelligence could “reasonably think [that quantitative metrics] relate to safety and efficiency,” but that same person could also understand, once the concern for discrimination was identified, that deactivation based solely on customer ratings is unreasonable. We also note that the services covered by the Ordinance and at issue here are delivery services, not ridesharing. An Uber rideshare passenger has an obvious basis to evaluate the safety of a ride; the recipient of a delivery by Uber Eats does not. Because of this distinction, Seattle may have thought it wise to designate customer ratings as a less reliable indicator to assess safety and efficiency in the context of delivery services. Regardless of the policy considerations the City took into account when enacting the Ordinance, in the context of a vagueness challenge, what matters is not whether a person of ordinary intelligence would agree with the content of or motivations for a law, but rather whether they would understand its dictates. Here, given the clarity afforded to the word “reasonable” when that word is read within the context of neighboring text and provisions in the Ordinance, we conclude that they would.

Plaintiffs' arguments that the other examples are more confusing than clarifying fare no better. There are multiple throughlines of policies in the list of examples. One of them is safety. The bar on deactivations based on background checks creates a carveout for reports that raise evidence of “egregious misconduct,” id. § 8.40.050.A.2.h, or any “action or behavior” that involves physical, economic, or emotional harm, id. § 8.40.020. So, too, with the bar on deactivations related to driving records: the Ordinance contains a carveout for unsafe behavior, id. § 8.40.050.A.2.h, as defined by more than three “moving violations or at-fault collisions” in less than three years, id. § 8.40.020. These limitations help to make clear what is and is not permitted under the law.

The Ordinance provides fair notice to a person of ordinary intelligence as to what grounds for deactivation are reasonably related to safety and efficiency. Plaintiffs may disagree with standards imposed by the Ordinance or may wish for more unfettered discretion in terminating a worker or deactivating an account, but that does not make the regulation vague.

IV. Conclusion

The Supreme Court has said, “[P]laintiff[s] seeking a preliminary injunction must establish that [they are] likely to succeed on the merits, that [they are] likely to suffer irreparable harm in the absence of preliminary relief, that the balance of equities tips in [their] favor, and that an injunction is in the public interest.” Winter v. Nat. Res. Def. Council, Inc., 555 U.S. 7, 20, 129 S.Ct. 365, 172 L.Ed.2d 249 (2008). The first factor is “the most important.” Baird v. Bonta, 81 F.4th 1036, 1040 (9th Cir. 2023) (quoting Env't Prot. Info. Ctr. v. Carlson, 968 F.3d 985, 989 (9th Cir. 2020)). If a movant cannot show a likelihood of success on the merits, then a court “need not consider the other factors.” Id. (quoting Disney Enters., Inc. v. VidAngel, Inc., 869 F.3d 848, 856 (9th Cir. 2017)).

The Ordinance regulates the conduct of deactivating worker accounts, not speech. But even if we accept Plaintiffs' contention that the Ordinance regulates speech, it is commercial speech, and the Ordinance overcomes the appropriate level of scrutiny under the First Amendment. Plaintiffs also fail to show that the law is unconstitutionally vague. Because Plaintiffs are unlikely to succeed on the merits of their claim, the district court did not abuse its discretion when it denied the preliminary injunction.

By affirming the authority of Seattle to enact and enforce the Ordinance, we do not express a view on the merits of the policy. Seattle has decided to enact the Ordinance, and the policy judgments reflected in the Ordinance are the City's. Plaintiffs are not required to agree with this or any other regulation but remain subject to them.

AFFIRMED.

APPENDIX

Seattle, Wash., Ordinance 126878, § 8.40.050 (Aug. 14, 2023)11

8.40.050 Deactivation requirements

A. A network company shall adopt the following measures prior to deactivating an app-based worker, except as provided in subsections 8.40.050.C and 8.40.050.D:

1. Fair notice of deactivation policy. A network company must inform the app-based worker in writing of the network company's deactivation policy, defining what constitutes a violation that may result in deactivation. The network company's written deactivation policy must be specific enough for an app-based worker to understand what constitutes a violation and how to avoid violating the policy. The deactivation policy must be available to the app-based worker in English and any language that the network company knows or has reason to know is the primary language of the app-based worker. The deactivation policy must be accessible to the app-based worker at least three years after deactivation. The Director may issue rules governing the form and description of the deactivation policy, the manner of its distribution, and required languages for its translation.

2. Reasonable policy. The policy that may lead to a deactivation must be reasonably related to the network company's safe and efficient operations. Examples of policies that are not reasonably related to the network company's safe and efficient operations include, but are not limited to:

a. Any rule or policy that would result in a deactivation based on an app-based worker's availability to work or number of hours worked, consistent with subsection 8.37.080.A.1;

b. Any policy that would result in a deactivation based on an app-based worker's acceptance or rejection of any individual offer, any types of offers, or any number or proportion of offers, consistent with subsection 8.37.080.A.2;

c. Any policy that would result in a deactivation based on an app-based worker's cancellation of an offer with cause, consistent with subsection 8.37.080.C;

d. Any policy that would result in a deactivation based on an app-based worker contacting the network company;

e. Any policy that would result in a deactivation based solely on a quantitative metric derived from aggregate customer ratings of an app-based worker's performance;

f. Any policy that would result in a deactivation based on statements by an app-based worker regarding compensation and/or working conditions made to customers, other app-based workers, network companies, the media, public officials, and/or the public;

g. Any policy that would result in a deactivation based on an app-based worker asserting their legal rights, whether in court or via procedures provided by any local, state, or federal agency; and

h. Any policy that would deactivate a worker based on the results of a background check, consumer report, driver record, or record of traffic infractions, except in cases of egregious misconduct or where required by other applicable law.

3. Investigation. A network company must conduct a fair and objective investigation prior to deactivating an app-based worker. The investigation must be sufficiently thorough to justify the deactivation and demonstrate an unbiased and neutral view of facts collected. If the app-based worker does not participate in the investigation or provide relevant information, the network company may complete the investigation based on available sources of information.

4. Confirmation of violation. The network company must demonstrate by a preponderance of the evidence that the alleged violation of the network company's policy or rule occurred.

5. Consistent application. The network company must apply the rule or policy, and penalty for violations, in a consistent manner.

6. Proportionate penalty. The penalty of deactivation must be reasonably related to the offense, and account for mitigating circumstances, such as the app-based worker's past work history with the network company.

B. Deactivation of an app-based worker will be considered unwarranted if the action is intended to or results in discrimination or a discriminatory act.

C. Subject to the provisions of this Section 8.40.050 and rules issued by the Director, a network company may immediately deactivate an app-based worker if such action is required to comply with any applicable court order or local, state, or federal laws or regulations, or where an app-based worker has engaged in egregious misconduct.

D. In the case of allegations of egregious misconduct, the network company may deactivate the app-based worker before completing an investigation. Except in extraordinary circumstances, the investigation shall not take longer than 14 days. If the investigation is delayed due to extraordinary circumstances, the network company must provide the app-based worker with written notice that the investigation is delayed, the reason(s) for the delay, and the date on which the completion of the investigation is anticipated.

I agree with the Majority that the Ordinance is not unconstitutionally vague. Unlike the majority, however, I believe the Ordinance compels speech and is thus subject to the First Amendment. While I agree with the majority's alternative reasoning that the speech at issue is commercial speech and subject to intermediate scrutiny, I believe Plaintiffs have raised serious questions going to the merits. Here, the district court found that the Ordinance had no First Amendment implications, 1 and I believe that to be error. I would therefore vacate and remand to have the district court redo its analysis on the merits and reexamine the Winter factors  2 in light of a determination that the Ordinance is subject to intermediate scrutiny, and in light of the current facts. Thus, I respectfully dissent, in part.

i. The Ordinance compels speech.

The majority concludes that the Ordinance “regulates conduct, not speech,” and that “the conduct that draws the remedy is the unwarranted deactivation of worker accounts,” Maj. Op. at ––––. The Ordinance does regulate conduct by prohibiting unwarranted deactivations. Ordinance § 8.40.060.A.1. But it also compels speech, and that regulation of speech also draws a remedy. The Ordinance requires plaintiffs to draft and provide workers a written “deactivation policy, defining what constitutes a violation that may result in deactivation,” id. § 8.40.050.A.1, and which “must be reasonably related to the network company's safe and efficient operations,” id. § 8.40.050.A.2.

The drafting and publication of such a policy is speech. 3 In a recent case, we specifically noted that the publication of “existing [terms of service] and content moderation policies may be commercial speech,” if not pure compelled speech subject to even more rigorous scrutiny. See X Corp. v. Bonta, 116 F.4th 888, 901 (9th Cir. 2024). And the Supreme Court has held that “the creation and dissemination of information are speech within the meaning of the First Amendment.” Sorrell v. IMS Health, Inc., 564 U.S. 552, 570, 131 S.Ct. 2653, 180 L.Ed.2d 544 (2011). Indeed, “[i]f the acts of ‘disclosing’ and ‘publishing’ information do not constitute speech, it is hard to imagine what does fall within that category.” Bartnicki v. Vopper, 532 U.S. 514, 527, 121 S.Ct. 1753, 149 L.Ed.2d 787 (2001) (quoting Bartnicki v. Vopper, 200 F.3d 109, 120 (3d Cir. 1999)). Because it is “well-established that the forced disclosure of information, even purely commercial information, triggers First Amendment scrutiny,” NetChoice, LLC v. Bonta, 113 F.4th 1101, 1117 (9th Cir. 2024), the Ordinance's deactivation policy requirement compels speech subject to the First Amendment.

To some extent, the majority acknowledges that sections 8.40.050.A.1–2 of the Ordinance not only regulate conduct but also compel speech. The majority appears to concede “[t]he fact that the Ordinance, in Plaintiffs' words, necessarily ‘compel[s] and dictate[s] the content of a written communication.” See Maj. Op. at ––––. However, the majority finds that “this reading ignores the placement of these provisions ‘in the context of the Ordinance as a whole,’ ” and that “[t]o read these provisions independently from the Ordinance would be to orphan them from their purpose: preventing unwarranted deactivations.” Id. at –––– (quoting Yim v. City of Seattle, 63 F.4th 783, 792 (9th Cir. 2023)). The majority therefore concludes that these “[c]ompelled changes in conduct—which might incidentally compel changes in speech—are not reviewed as content-based speech restrictions.” Id. at –––– (quoting Olympus Spa v. Armstrong, 138 F.4th 1204, 1216 (9th Cir. 2025)). I disagree with this conclusion for three reasons.

First, the Ordinance's compulsion of speech is not incidental to its prohibition on unwarranted deactivations. The prohibition on unwarranted deactivations and the requirement to promulgate a deactivation policy of a certain kind are distinct and separately enforceable. See Ordinance § 8.40.140 (“The failure of any respondent to comply with any requirement imposed on the respondent under this Chapter 8.40 is a violation.”); id. § 8.40.170.E (empowering the Director of the Office of Labor Standards to “assess fines for a violation of this Chapter” including under § 8.40.050). In fact, while the Office of Labor Standards has been empowered since January 1, 2025, to “administer and enforce” the deactivation policy requirement, the Office is prohibited from enforcing the prohibition on unwarranted deactivations until June 1, 2027. Id. §§ 8.40.130.A, B. Thus, for the twenty-nine months following enactment, including now, the Ordinance empowers the Office to assess fines only for violations of the speech regulation, and not for violations of the conduct regulation. 4 And we have held that we cannot “ignore that the [challenged] requirement compels speech simply because other parts of the [challenged law] may primarily or exclusively regulate non-expressive conduct.” NetChoice, 113 F.4th at 1117. 5 That is because the government “cannot insulate a specific provision of law from a facial challenge under the First Amendment by bundling it with other, separate provisions that do not implicate the First Amendment.” Id.

Second, by concluding based on the Ordinance's “purpose” that the speech regulation is incidental to the conduct regulation, the majority, I believe, applies the wrong test. We consider a law's “stated purpose” when determining whether the conduct that the law regulates is sufficiently expressive to fall within the purview of the First Amendment. HomeAway.com, Inc. v. City of Santa Monica, 918 F.3d 676, 685 (9th Cir. 2019) (quoting Sorrell, 564 U.S. at 565, 131 S.Ct. 2653). That test is inapplicable here because the Ordinance's speech regulation is separate from its conduct regulation, and we can consider the statutory purpose only if the law regulates conduct rather than speech.

Third, the Ordinance's “stated purpose” is more ambiguous than the majority says. Quoting from the Ordinance's findings and declarations, the majority identifies the Ordinance's purpose as to “protect[ ] and promote[ ] public health, safety, and welfare by establishing protections against unwarranted deactivations for app-based workers.” Maj. Op. at ––––, –––– (quoting Ordinance § 1.C). The majority concludes that the purpose of the Ordinance is to prevent unwarranted deactivations and not to compel speech. But the same section of the Ordinance also reflects a purpose to compel plaintiffs' publication of deactivation policies, separate from the prevention of any deactivation decisions. See id. § 1.H (“App-based workers often do not have the information they need to know about how they will be evaluated.”). At least one purpose of the Ordinance is, therefore, to compel plaintiffs to share information about how app-based workers will be evaluated. Because disclosure of such information is speech subject to First Amendment scrutiny, Sorrell, 564 U.S. at 570, 131 S.Ct. 2653; NetChoice, 113 F.4th at 1117; X Corp., 116 F.4th at 901, the Ordinance's stated purpose was, in part, to compel speech. 6

I therefore disagree with the majority's conclusion that the Ordinance does not regulate speech.

ii. Plaintiffs raise serious questions on the merits.

I agree with the majority 1) that the speech at issue is commercial speech, and 2) that the so-called Zauderer standard applies to the commercial speech compelled by the Ordinance. See Zauderer v. Off. of Disciplinary Counsel, 471 U.S. 626, 105 S.Ct. 2265, 85 L.Ed.2d 652 (1985); see also Maj. Op. at –––– – ––––. But I disagree with the majority's decision to conduct the Zauderer analysis in the first instance. Because the district court concluded that the Ordinance does not compel speech, it “had no occasion to conduct a further inquiry into whether [the Ordinance], as a speech regulation, survived First Amendment scrutiny.” Expressions Hair Design v. Schneiderman, 581 U.S. 37, 48, 137 S.Ct. 1144, 197 L.Ed.2d 442 (2017). We should therefore remand for the district court to consider “whether the law can be upheld as a valid disclosure requirement under Zauderer.” Id.

Moreover, unlike the majority, I believe that plaintiffs at least raise a serious question on the merits as to whether the Ordinance satisfies the Zauderer standard. “Under Zauderer as we interpret it today, the government may compel truthful disclosure in commercial speech as long as the compelled disclosure is ‘reasonably related’ to a substantial governmental interest and involves ‘purely factual and uncontroversial information’ that relates to the service or product provided.” CTIA – The Wireless Ass'n v. City of Berkeley, 928 F.3d 832, 842 (9th Cir. 2019) (first quoting Zauderer, 471 U.S. at 651, 105 S.Ct. 2265, then quoting Nat'l Inst. of Fam. & Life Advocs. v. Becerra, 585 U.S. 755, 768, 138 S.Ct. 2361, 201 L.Ed.2d 835 (2018)).

The speech compelled here is not purely factual and uncontroversial information. The speech regulation requires more than a mere notice of rights—that is separately required by a distinct section of the Ordinance. See § 8.40.100. The speech regulation instead requires plaintiffs to formulate and publish their opinions on what behaviors or standards are reasonably related to safety or to efficiency. These are not purely factual determinations but are instead plaintiffs' opinions. And they are likely to be controversial opinions. This is especially so because safety and efficiency are often in tension, and plaintiffs' policies will have to balance these two values. Plaintiffs will likely also balance these two values against others, such as customer satisfaction, profitability, and administrability, because the Ordinance does not require that the policy be related solely to safety and efficiency. Any balance of these oft-competing values is likely to be controversial. Different stakeholders—including users, workers, employees, shareholders, and traffic authorities—are bound to have different perspectives on where the best balance lies.  7 Because the Ordinance compels commercial speech that is not or at least may not be “purely factual and uncontroversial information,” it potentially violates the First Amendment. See CTIA, 928 F.3d at 842.

The majority states that the speech regulation is nonetheless permissible because any challenge to it “must first satisfy a threshold inquiry that the communication be ‘neither misleading nor related to unlawful activity.’ ” Maj. Op. at –––– (quoting World Wide Rush, LLC v. City of Los Angeles, 606 F.3d 676, 684 (9th Cir. 2010)). The majority finds that the speech regulation “regulates at most speech related to ․ unlawful activity: the unwarranted deactivations of worker accounts.” Id. The majority concludes that plaintiffs' challenge thus fails this “threshold inquiry.” See id. But the majority's threshold inquiry does not apply to challenges to compelled speech—it governs only challenges to restrictions on speech. See World Wide Rush, 606 F.3d at 684–85. And, in any case, the speech here is unrelated to (or not solely related to) unlawful activity.

The majority's “threshold inquiry” is taken from World Wide Rush, a case involving a challenge to a restriction on speech, not to a compulsion of speech. See World Wide Rush, 606 F.3d at 680, 685. World Wide Rush, in turn, derived this test from Metro Lights, L.L.C. v. City of Los Angeles, which was also a speech restriction case. 551 F.3d 898 900, 903 (9th Cir. 2009). And Metro Lights identifies this “threshold matter” as an element of the Central Hudson test, not the Zauderer test. See Metro Lights, 551 F.3d at 903–04 (describing the requirement that “the communication [be] neither misleading nor related to unlawful activity” as an “element[ ] of the Central Hudson test” (citing Cent. Hudson Gas & Elec. Corp. v. Pub. Serv. Comm'n, 447 U.S. 557, 566, 100 S.Ct. 2343, 65 L.Ed.2d 341 (1980))). Because, as even the majority acknowledges, Maj. Op. at –––– n.10, the Central Hudson test is not applicable to this case, plaintiffs need not show that their speech is unrelated to unlawful activity.

This distinction matters because, even though Zauderer's rational basis standard is “more permissive” than Central Hudson's intermediate scrutiny standard, see Pharm. Rsch. & Mfrs. of Am. v. Stolfi, petition for cert. filed, No. 25-1018 (U.S. Feb. 24, 2026), 153 F.4th 795, 808 (9th Cir. 2025), “one size does not fit all in commercial speech cases,” CTIA, 928 F.3d at 842. The two tests have different triggering conditions, see id., and a regulation that survives Central Hudson review is therefore not guaranteed to survive Zauderer review.

In any event, the majority is wrong to conclude that the speech is related to unlawful activity. The majority concludes that by requiring that deactivation policies be reasonably related to safe and efficient operations, the Ordinance simply regulates speech related to unlawful activity—namely, speech related to unwarranted deactivations. I believe that this conclusion is both circular and incorrect. The Ordinance defines “unwarranted deactivation” as “a deactivation that does not comply with Section 8.40.050.” Ordinance § 8.40.020. But § 8.40.050 does not contain any freestanding limits on when deactivation is prohibited. In other words, it does not proscribe the firing of an app-worker for a reason unrelated to safe and efficient operations. Instead, it prohibits only certain forms of deactivation policy and describes when a policy would fail to be reasonably related to safe and efficient operations. Id. §§ 8.40.050.A.1–2. It then prohibits companies from deactivating a worker without the results of an investigation finding that the worker violated the company's deactivation policy. Id. §§ 8.40.050.A.3.1–5. In other words, it is not unlawful to deactivate workers for reasons unrelated to safe and efficient operations. It is unlawful to speak a policy that is unrelated to safe and efficient operations; and it is unlawful for a company to violate its own deactivation policy. Thus, this is not a regulation of speech related to unlawful activity, but a direct regulation of speech itself.

I think it is far from clear that Seattle is likely to prevail on the merits. At the very least, I believe that plaintiffs have raised serious questions going to the merits.

iii. We should remand for the district court to reanalyze the Ordinance under intermediate scrutiny and to reexamine the other Winter factors.

In addition to likelihood of success on the merits, a district court considering a motion for a preliminary injunction must consider whether the movant will likely suffer irreparable harm without the injunction, whether the balance of the equities favors the movant, and whether a preliminary injunction is in the public interest. Winter v. Nat. Res. Def. Council, Inc., 555 U.S. 7, 20, 129 S.Ct. 365, 172 L.Ed.2d 249 (2008). “It is well-established that the first factor is especially important when a plaintiff alleges a constitutional violation and injury.” Baird v. Bonta, 81 F.4th 1036, 1041 (9th Cir. 2023). A determination a plaintiff alleging such an injury is likely to prevail on the merits “usually demonstrates he is suffering irreparable harm” and “tips the public interest sharply in his favor.” Baird v. Bonta, 81 F.4th 1036, 1041 (9th Cir. 2023). “As a general matter, district courts ‘must consider’ all four Winter factors.” Id. at 1040 (quoting Vivid Ent., LLC v. Fielding, 774 F.3d 566, 577 (9th Cir. 2014)). But because the first factor is most important, if a movant fails to show likelihood of success on the merits, “the court need not consider the other factors in the absence of ‘serious questions going to the merits.’ ” Disney Enterps., Inc. v. VidAngel, Inc., 869 F.3d 848, 856 (9th Cir. 2017) (quoting All. for the Wild Rockies v. Cottrell, 632 F.3d 1127, 1134–35 (9th Cir. 2011)).

Plaintiffs have at least raised a serious question going to the merits. In my view, therefore, the district court must consider all four Winter factors. Baird, 81 F.4th at 1041. And I believe that the district court must decide in the first instance whether the Ordinance survives Zauderer review. Expressions Hair Design, 581 U.S. at 48, 137 S.Ct. 1144. Here, its review of the other three factors was cursory, dividing only four sentences between them. Further, the district court's analysis of these three factors was based on what I believe was its erroneous analysis of questions going to the merits. Moreover, since the district court's decision, plaintiffs have issued deactivation policies in light of the Ordinance. The district court has had no opportunity to consider how the issuance of these policies, and Seattle's response or lack thereof, has affected the equities or the existence of irreparable harm.

Accordingly, I would remand for reconsideration of all four Winter factors. I therefore respectfully dissent, in part.

FOOTNOTES

1.   Relevant portions of the Ordinance are attached as an appendix to this opinion.

2.   “Egregious misconduct” is defined as:[A]n action or behavior by an individual app-based worker that: (1) endangers the physical safety of the customer, or a third person, the network company, or an animal; or (2) intentionally causes economic harm to the customer, a third person, or the network company; or (3) is threatening, harassing, or abusive to the customer, a third party, or the network company.․ [It] includes but is not limited to ․ failing to maintain a valid state driver's license ․․ [It] shall not include conduct related to non-criminal moving violations ․ or traffic collisions unless the app-based worker has accumulated more than three non-criminal moving violations or at-fault collisions in the previous three years.Id. § 8.40.020.

3.   The third claim cited in the complaint, that the Ordinance restricts Uber's expressive associational rights in violation of the First Amendment, has not been argued in this appeal.

4.   The partial dissent takes issue with our reading of the Ordinance's stated purpose, arguing that it “is more ambiguous than the majority says.” Partial Dissent at ––––. The partial dissent appears to concede that “protect[ing] and promot[ing] public health, safety, and welfare,” Ordinance § 1.C, is the Ordinance's stated purpose, but it adds another purpose to the mix—“to compel [P]laintiffs' publication of deactivation policies.” Partial Dissent at ––––. We disagree with that reading for two reasons.First, the text of subsection 1.H, which the partial dissent cites in support of the additional purpose it proffers, simply reflects Seattle's finding that “[a]pp-based workers often do not have the information they need to know about how they will be evaluated.” That subsection does not displace the central purpose of the Ordinance, stated earlier in subsection 1.C. Rather, it merely serves as one among many findings that Seattle considered when enacting the Ordinance.Second, and more fundamentally, even assuming that subsection 1.H does announce a purpose of the Ordinance, the partial dissent's reading isolates a small part of the law, and, in doing so, misses the central purpose animating the Ordinance. In the First Amendment context, the fact that other purposes may be inferred from a law's text does not bar us from recognizing that the law has a central stated purpose. Cf. HomeAway.com, 918 F.3d at 685 (“The Ordinance itself makes clear that the City's ‘central and significant goal ․ is preservation of its housing stock and preserving the quality and nature of residential neighborhoods.’ ” (alteration in original)). Our reading that the Ordinance's central purpose is to “protect[ ] and promot[e] public health, safety, and welfare” is reinforced by the fact that, unlike any other provision in section 1, the purpose articulated in subsection 1.C is attributed to “[t]his ordinance,” that is, to the entirety of the law. Ordinance § 1.C. Additionally, the words “health,” “safety,” and “welfare” are repeated throughout the Ordinance with a regularity that further underscores their centrality and significance to the Ordinance as a whole. See, e.g., id. § 1.B (“public health, safety, and welfare”); id. § 1.C (same); id. § 1.V (“safety”); id. § 1.W (“welfare” and “health”).Moreover, the Ordinance does not require Plaintiffs to communicate a deactivation policy to anyone other than the workers affected by it and does not require Plaintiffs to state their agreement with or support for the limitations imposed by the City. Those points further demonstrate that the central purpose of the Ordinance is to regulate the terms of network companies' deactivation of app-based workers, not to compel network companies' speech.

5.   The partial dissent relies on the delayed ability of Seattle's Office of Labor Standards to fine network companies for violating the Ordinance's prohibition on unwarranted deactivations, see Ordinance § 8.40.130.B, to argue that the requirement that network companies promulgate deactivation policies that are reasonably related to the companies' safe and efficient operations is not incidental to the substantive provisions of the Ordinance, Partial Dissent at –––– – ––––. To the contrary, the delay demonstrates that the substantive provisions are what count. An affected app-based worker can sue a network company for any violation of any provision of the Ordinance. Id. §§ 8.40.060.C, 8.40.230. The Office also can enforce immediately the requirements that network companies create internal procedures for app-based workers to challenge deactivations, id. § 8.40.060.B.1, that network companies provide relevant records to workers who bring such challenges, id. § 8.40.080, that network companies provide a written notice of rights under the Ordinance to every app-based worker (a requirement that Plaintiffs do not challenge), id. § 8.40.100, and that network companies not retaliate against workers who exercise their rights under the Ordinance, id. § 8.40.120.B. The delay on which the partial dissent relies merely reflects Seattle's policy decision to punish violators for unwarranted deactivations, by fining them, only later. Forbearance of financial punishment does not logically suggest that informing app-based workers about the kinds of conduct that can lead to deactivation is more than incidental to establishing their right to be free from unwarranted deactivation. Instead, this delay suggests only that the City cares more about encouraging compliance with the substance of the new Ordinance than about punishing violators.

6.   Any suggestion in the partial dissent to the contrary is mistaken. The partial dissent observes that the Ordinance requires Plaintiffs “to draft and provide workers a written” deactivation policy. Partial Dissent at –––– – ––––. We agree that it does. See Ordinance § 8.40.050.A.1. The partial dissent then concludes that “[t]he drafting and publication of such a policy is speech” within the meaning of the First Amendment. Partial Dissent at ––––. We disagree.As a threshold matter, the fact that speech may ensue incidentally from a government regulation that otherwise concerns nonexpressive conduct—here, network companies' deactivation of app-based workers' accounts—does not mean that such “speech” receives protection under the First Amendment. The Supreme Court not only has concluded that nonexpressive conduct is undeserving of First Amendment protection, but also has held that not all activity conducted with an intent to express an idea amounts to speech within the meaning of the First Amendment. See, e.g., United States v. O'Brien, 391 U.S. 367, 376, 88 S.Ct. 1673, 20 L.Ed.2d 672 (1968) (rejecting the notion that, “whenever the person engaging in the conduct intends thereby to express an idea,” the person engages in speech under the First Amendment). We have followed suit. See, e.g., Edge v. City of Everett, 929 F.3d 657, 669 (9th Cir. 2019) (holding that the plaintiffs' wearing of bikinis while working was not protected by the First Amendment because there was not a “ ‘great likelihood’ that their intended messages related to empowerment and confidence will be understood by those who view them”). That is why, even if the Ordinance were understood to be “not wholly unrelated to a communicative component,” “that in itself [would] not trigger First Amendment scrutiny.” Int'l Franchise Ass'n, 803 F.3d at 408.Nor is the partial dissent's invocation of X Corp. persuasive in support of its conclusion that the Ordinance's requirement that network companies promulgate deactivation policies compels such companies to engage in commercial speech. Partial Dissent at ––––. In X Corp., as the partial dissent notes, we observed that a social media company's posting of its “existing [terms of service] and content moderation policies” was commercial speech. 116 F.4th at 901. But that observation was dictum. On appeal, the plaintiff challenged only the relevant statute's requirements that social media companies submit to the state semiannual reports detailing their content-moderation practices and the ways in which their terms of service address, inter alia, hate speech or racism, and that social media companies not materially omit or misrepresent required information in such reports. Id. at 894, 898. Accordingly, we did not consider the statute's requirement that social media companies publicly post their existing terms of service. Id. Moreover, in X Corp., we reaffirmed that commercial speech generally is “speech that does no more than propose a commercial transaction,” id. at 900 (quoting United States v. United Foods, Inc., 533 U.S. 405, 409, 121 S.Ct. 2334, 150 L.Ed.2d 438 (2001)), and that the “commercial speech analysis is fact-driven,” id. (quoting First Resort, Inc. v. Herrera, 860 F.3d 1263, 1272 (9th Cir. 2017)), thereby rejecting any sweeping suggestion that would have the unwarranted consequence of designating much, if not all, activity with commercial indicia as commercial speech that implicates the First Amendment. Regardless, as discussed in Part II.B of this opinion, even assuming that the Ordinance regulates commercial speech, it satisfies the applicable level of scrutiny.

7.   For purposes of this section of the opinion, we assume, without deciding, that Plaintiffs are correct in their assertion that the Ordinance regulates speech.

8.   During oral argument, both Plaintiffs requested that we resolve the case before us on the merits rather than remand to the district court. The record is sufficiently developed for us to decide how the law applies, so we do. See X Corp., 116 F.4th at 898 (rejecting the district court's application of Zauderer and instead applying strict scrutiny).

9.   The partial dissent asserts that the content of network companies' deactivation policies is not purely factual and uncontroversial because the Ordinance “requires more than a mere notice of rights”; the partial dissent claims that it also requires Plaintiffs to “publish their opinions on what behaviors or standards are reasonably related to safety or to efficiency.” Partial Dissent at ––––. We are unpersuaded by this characterization. For one thing, once a deactivation policy is issued, it will be “literally true.” Whether or not the network company issuing the deactivation policy would have devised its policy differently had the City enacted different labor regulations, or no regulations at all, the policy will truthfully and accurately identify the conduct of app-based workers that could lead to deactivation. The policy will reflect facts. CTIA, 928 F.3d at 847.More broadly, as we recently observed, “a purely factual statement does not become controversial simply because it ‘can be tied in some way to a controversial issue.’ ” Foothills Christian Ministries v. Johnson, 148 F.4th 1040, 1054 (9th Cir. 2025) (quoting CTIA, 928 F.3d at 845), petition for cert. filed, No. 25-802 (U.S. Jan. 5, 2026). That observation applies here. The recognition that the concepts of safety and efficiency could generate differing views among various stakeholders in the context of devising deactivation policies, see Partial Dissent at ––––, is insufficient to render the required disclosure of the deactivation policies controversial. See id. (“[A] compelled statement is controversial when it takes ‘sides in a heated political controversy,’ and forces the speaker ‘to convey a message fundamentally at odds with its mission.’ ” (quoting CTIA, 928 F.3d at 845)).It also bears emphasis that, if a network company disagrees with the Ordinance's examples of policies that are not reasonably related to safe and efficient operations, the company may elect to express its disagreement, even alongside the policy itself. This matters because the ability to register one's disagreement, even while complying with the Ordinance, dispels any “forced association” between the discloser (the network company) and the disclosure (the Ordinance-compliant deactivation policy), thereby undermining any suggestion of compulsion. See Pharm. Rsch. & Mfrs. of Am. v. Stolfi, 153 F.4th 795, 812 (9th Cir. 2025) (quoting Pharm. Care Mgmt. Ass'n v. Rowe, 429 F.3d 294, 316 (1st Cir. 2005)), petition for cert. filed, No. 25-1018 (U.S. Feb. 24, 2026).

10.   The partial dissent faults us for conflating the two tests for commercial speech. Partial Dissent at ––––. Specifically, it observes that the threshold inquiry of whether the purported speech relates to unlawful activity is an element of the Central Hudson test for regulation of commercial speech, not the Zauderer test for compulsion of commercial speech. Partial Dissent at ––––. The threshold inquiry is indeed an element of Central Hudson, and we do not purport to import it into Zauderer. Rather, in considering Plaintiffs' assertion that Central Hudson applies, we conclude that their argument falters at Central Hudson's doorstep. This matters because we have already observed that Zauderer sets forth a “more permissive standard” than that articulated in Central Hudson. See Pharm. Rsch. & Mfrs. of Am., 153 F.4th at 808; see also id. at 810–11 (explaining that, if a “law[ ] that compel[s] certain information to be communicated directly from one private entity to another” regulates commercial speech, “courts apply either intermediate scrutiny,” pursuant to Central Hudson, “or a lower level of scrutiny akin to rational basis review,” pursuant to Zauderer). Plaintiffs fail to show that the Ordinance violates the First Amendment even under the more rigorous test in Central Hudson; a fortiori, the Ordinance passes the more permissive standard in Zauderer.

1.   “The starting point of the Court's analysis is whether the Ordinance presents a restriction on speech. The Court finds that the Ordinance does not, and that the First Amendment claim is unlikely to succeed on the merits.”

2.   Because the district court found that the Ordinance had no First Amendment implications, its analysis of the other Winter factors was cursory.

3.   Contrary to the majority, I believe the present case is unlike O'Brien and City of Everett—it concerns speech itself, not expressive conduct. The drafting of a policy text, making trade-offs between competing and controversial values, and proclaiming that certain internal regulations are or are not “reasonably related to safety and efficiency,” are orders of magnitude more expressive than simply “wearing bikinis while working.” Edge v. City of Everett, 929 F.3d 657, 662 (9th Cir. 2019). And the O'Brien Court declined to hold that expressive conduct, such as the burning of a draft card, is outside the scope of the First Amendment. United States v. O'Brien, 391 U.S. 367, 376, 88 S.Ct. 1673, 20 L.Ed.2d 672 (1968).

4.   I am unconvinced by the majority's argument that “this delay suggests only that the City cares more about encouraging compliance with the substance of the new Ordinance than about punishing violators.” Maj. Op. –––– n.5. First, it does not matter whether Seattle cares more about encouraging compliance or punishing violators. Nor does it matter whether Seattle cares more about preventing unwarranted deactivations or promoting transparency. The Ordinance aims at both, and does not compel speech only incidentally to its regulation of conduct.

5.   Seattle argues that NetChoice is distinguishable on this point because, in that case, we found the regulation was subject to the First Amendment because the regulated speech was “disconnected from any economic transaction.” NetChoice, 113 F.4th at 1119. But Seattle misreads NetChoice. The absence of a connection to an economic transaction is the reason we applied strict scrutiny in NetChoice—it is not the reason we found the statute to regulate speech. Id. at 1119–21.

6.   The majority wrongly contends that my reading “isolates a small part of the law, and, in doing so, misses the central purpose animating the Ordinance.” Maj. Op. –––– n.4. The Ordinance aims not only at decreasing unwarranted deactivations, but also at increasing transparency. I do not distort the Ordinance by noting that it compels speech, §§ 8.40.050.A.1–2, based on separate policy findings, § 1.H, and that it establishes an enforcement regime separate from the regime regulating unwarranted terminations, §§ 8.40.140, 170.E. Rather, it is the majority that improperly cabins the Ordinance by focusing exclusively on its separate regulation of conduct.

7.   I do not agree with the majority that this material is purely factual and uncontroversial simply because “once a deactivation policy is issued, it will be ‘literally true.’ ” Maj. Op. –––– n.9. Yes, once the policy is published, there will be a “fact” as to what the policy says. But that is true of all speech. And there will remain, after publication, no “fact” as to the best balance of safety, efficiency, and other values, and no “fact” as to which internal regulations are “reasonably related” to these values. Nor, contrary to the majority's suggestion, could the injury to the companies' First Amendment rights be fully ameliorated by the company's ability “to express its disagreement, even alongside the policy itself.” Id. The Ordinance requires the companies to make and publish their judgment about what is “reasonably related.” §§ 8.40.050.A.1–2.

CLIFTON, Circuit Judge:

Opinion by Judge Clifton; Partial Dissent by Judge Bennett