Does AI writing collapse the author-to-public relationship?

Writing for a public is a specific practice. An author takes responsibility for communicating ideas to an audience the author does not know personally but whose interests, prior commitments, and likely objections the author has internalized. The author writes for that internalized audience, not for any single interlocutor. This is what makes published text different from private correspondence: the audience-model is constitutive of the genre.

AI generation collapses this distinction. The response is produced for the prompter — shaped by the prompt, optimized by alignment for the prompter's satisfaction, calibrated to the prompter's context. There is no modeled audience beyond the prompter. When the prompter then publishes the output to a public, the text that reaches that public was never written for them. The author-to-public relationship, which traditionally carried the responsibility of audience-internalization, is performed by a system that cannot internalize an audience because it has no audience beyond its immediate user.

This is distinct from standard complaints about AI writing (too generic, too hedged, too verbose). Those are stylistic. This is structural: the relationship between speaker and public that makes public writing a practice has been reorganized. The prompter inherits the genre-role of "author" without performing the genre-work of "writing for a public." Can AI replicate the communicative work experts do? names the adjacent problem from the expertise side; this names it from the authorship side.

The consequence is not that published AI text is bad text. It is that publishing AI text is a different speech act than publishing written text, even when the surface output is similar. Publics relate to texts through the author-function they presume; that function is being silently replaced.

Source: Epistemic Inflation