Algorithmic Speech Doctrine

The algorithmic speech doctrine concerns the First Amendment question of whether an algorithm that decides what users see is itself constitutionally protected expression. The answer determines the level of judicial scrutiny applied to platform regulation: if algorithmic curation is fully protected speech, architectural-regulation strategies face strict scrutiny; if it is not, those regulations face intermediate or rational-basis review. The Supreme Court has not resolved the question directly. The doctrine recurs in Farahany's Advanced Topics course (Weeks 4 and 11).

Pre-algorithmic baseline

Two earlier cases define the editorial-discretion baseline for human editorial choice. In Miami Herald v. Tornillo (SCOTUS, 1974), the Court held that editorial discretion — what a newspaper chooses to publish, where to place it, and how prominently — is core First Amendment expression, and that government cannot compel a newspaper to publish a candidate's rebuttal. In Turner Broadcasting (SCOTUS, 1994), the Court held that cable operators' decisions about which channels to carry are expression but receive intermediate scrutiny rather than strict scrutiny, because of the bottleneck nature of cable. Together these establish that human editorial choice over what to host or amplify is protected expression.

The algorithmic question

The central question is whether automated, machine-learning-driven recommendation ranks as editorial discretion. Two positions divide on the answer.

Under the platform position, the platform designed the algorithm, set its objectives, and deployed it to curate; the absence of human review of any specific decision is treated as irrelevant, because the platform expressed an editorial judgment by training and deploying the system. Zhang v. Baidu (S.D.N.Y., 2013) extended Tornillo to search-engine results on this reasoning.

Under the regulator position, the platform did not speak each ranking decision; the algorithm inferred it from behavioral signals optimized against an engagement objective, so there is no expressive intent at the granular level. This argument draws on the Austin & Levy "speech certainty" principle (Stanford L. Rev. 2025), which holds that speech is only speech if the speaker knows what they said when they said it. The argument is that machine-learning systems frequently cannot satisfy this standard, because engineers often cannot explain why a specific user received a specific recommendation, raising the question whether a message the speaker cannot account for qualifies as expression.

Distinctions Zhang does not resolve

Class 4.3 identifies two features of modern algorithmic curation that Zhang does not address. First, Zhang involved a human-designed editorial policy (filtering pro-democracy content) implemented through an algorithm, whereas modern recommendation feeds run trained ML systems in which no engineer decided that a given user receives given content. Second, Zhang involved a response to a query, whereas recommendation feeds run continuously, shaping the user's informational environment before any query. Whether these differences are constitutionally significant is the central unresolved question.

Consequences for regulation

The level of protection assigned to algorithmic curation determines the viability of platform regulation. If algorithmic curation is fully First Amendment-protected, child-safety laws including CAADCA, COPPA 2.0, KOSA, and EU DSA Article 25 (grouped in the course's Week 9) likely fail strict scrutiny; platform-must-host laws such as Texas HB 20 and Florida SB 7072 likely fail as well as the symmetric case; and almost no architectural regulation survives. If algorithmic curation is not fully protected, or is treated as commercial-speech-grade rather than fully protected, architectural regulation receives intermediate or rational-basis review, more regulatory room opens, and platforms lose their primary defensive doctrine.

Common-carrier and public-accommodations off-ramps

Justice Thomas's 2021 dicta in Biden v. Knight (a 12-page opinion on a moot case) proposed two historical frameworks that would bypass the algorithmic-speech question. Under the common-carrier doctrine, platforms with large market share resemble railroads, telegraphs, and telephones — public-obligation infrastructure that must serve all customers without discrimination. Under the public-accommodations doctrine, drawing on Pruneyard Shopping Center v. Robins (1980), states may impose nondiscrimination requirements on businesses dominant in public-facing communication. Texas and Florida legislated along these lines, producing the litigation in Moody v. NetChoice (2024). The Supreme Court declined to validate either doctrine as applied to platforms but did not foreclose them.

Key cases

Relationships

• related: AI and the First Amendment, Content vs Architecture Theory of Social Media Harm, Cognitive Liberty
• constrains: child-safety legislation (planned), platform-must-host laws (planned)
• load-bearing-in: Inside My Advanced Topics Class 4.3: Is an Algorithm Speech? (Farahany, February 2026), Inside My Advanced Topics Class 9.3: The Ninth Circuit Told California How to Fix the Law (Farahany, March 2026), Inside My Advanced Topics Class 11.1: The Wall That Cuts Both Ways, and Who Speaks (Farahany, April 2026)