The First Amendment Implications of Algorithmic Content Curation

The question of whether a platform's recommendation algorithm constitutes protected editorial judgment, or merely the neutral transmission of third-party content, has now become the central doctrinal puzzle of platform speech law. The Supreme Court's recent intervention in Moody v. NetChoice, LLC⁵ did not resolve the question on the merits, but it did clarify the analytic framework that lower courts must apply going forward — and in so doing, vacated the divergent approaches of the Fifth and Eleventh Circuits and remanded for further development of the facial-challenge record.

The editorial-discretion line

The First Amendment's protection of editorial judgment is well established. In Miami Herald Publishing Co. v. Tornillo,¹ the Supreme Court invalidated a Florida right-of-reply statute on the ground that it impermissibly intruded upon the editorial discretion of newspaper editors. In Pacific Gas & Electric Co. v. Public Utilities Commission,² the Court extended the editorial-discretion principle to a utility's billing-envelope inserts, holding that the company could not be compelled to carry the speech of a third party. The through-line is the proposition that compelled inclusion of speech is itself a constitutional injury — distinct from, and in addition to, any restriction on the speaker's own expression.

The argument that platform curation is a species of editorial judgment proceeds by analogy: when a platform selects, ranks, demotes, or removes content, it is making the kind of inclusion-and-exclusion decisions that the First Amendment protects in the newspaper context. The argument's force does not depend on the platform being a publication in any conventional sense; it depends on the structural similarity between the platform's curation function and the editor's selection function. Both involve the exercise of judgment about what speech to present to an audience.

The Fifth Circuit's contrary view

The Fifth Circuit's decision in NetChoice v. Paxton³ rejected this analogy with notable directness, upholding the Texas HB 20 statute on the theory that large platforms operate as common carriers and that content moderation is not speech but "censorship" of the speech of others. The opinion's reasoning has been widely criticized, not least for its insistence that the platforms' moderation policies are properly characterized as conduct rather than expression. The Eleventh Circuit reached a different conclusion in NetChoice v. Attorney General, Florida,⁴ enjoining Florida's SB 7072 on First Amendment grounds and treating platform content-moderation policies as protected editorial judgment.

Compelled inclusion of speech is itself a constitutional injury — distinct from any restriction on the speaker's own expression.

What Moody decided, and what it left open

In Moody, the Supreme Court vacated both circuit decisions and remanded with instructions that the lower courts properly evaluate the breadth of the challenged statutes — that is, the full range of applications to which the laws would extend — before rendering a facial First Amendment judgment. Justice Kagan's opinion for the Court took pains to reject the proposition that platform content moderation, when it amounts to the exercise of editorial judgment over what speech to present, falls outside the First Amendment. Whether a particular application of a state statute is constitutional, the Court suggested, will depend on the function the platform is performing in that application — and the lower courts had failed to do the work necessary to make those determinations.

The pragmatic effect of Moody has been to push the litigation back into the development of more refined factual records. The Fifth Circuit's categorical embrace of the common-carrier theory has been displaced; the Eleventh Circuit's categorical embrace of editorial discretion has been deemed insufficiently nuanced for facial review. What is likely to emerge over the coming term is a more granular doctrine — one that distinguishes among the various functions platforms perform (hosting, ranking, recommending, removing, labeling) and among the various types of speech being regulated.

The algorithm-as-author problem

The doctrinal complication that Moody does not address — and that the lower courts will now have to confront on remand — is the relationship between human editorial judgment and algorithmic curation. A newspaper editor who selects an op-ed for the morning edition is plainly engaged in expressive activity; a machine-learning model that ranks user posts according to engagement-prediction metrics is engaged in something whose expressive character is genuinely contested. The argument that the algorithm's output reflects the platform's prior editorial choices — choices about training data, objective functions, and policy constraints — has substantial force, but it is also the argument that pulls platform curation closest to the speech-act category that triggers full First Amendment protection.

The consequence, we suspect, will be a doctrine that treats algorithmic curation as protected editorial judgment when the platform's moderation policies reflect deliberate expressive choices, and that treats it more skeptically when the curation function is described — by the platform itself or by its critics — as neutral, mechanical, or engagement-maximizing. For related discussion, see our commentary on the constitutional limits of state platform regulation and on Section 230 in the age of generative AI.