Do readers and writers differ on AI disclosure necessity?

The vignette study (727 participants) finds that readers judge AI disclosure more necessary than writers do. The excerpt treats this gap as its central result, and it is measured on hypothetical reading and writing situations, not on what people actually disclose. Among the procedural factors, disclosure is regarded as more necessary when AI's contribution is irreplaceable and when it is directly incorporated into the writing. The abstract also lists low writer intentionality among the conditions that raise necessity, but the conclusion splits that effect by perspective, which the second note takes up. Effort, the authors write, "shows no significant effect on the perceived necessity." Purpose is reported as insignificant too, though the discussion says supplementary qualitative analysis found it mattered to some participants.

The authors frame the study as bottom-up. Existing guidelines are usually written top-down by policymakers, publishers and platform moderators, "who are not the ones directly affected by the writings and AI disclosure," so the study asks readers and writers what they think is necessary. The four procedural factors (replaceability, effortfulness, intentionality, directness) are synthesized from prior transparency work. For the reader-writer gap, the discussion offers two candidate explanations that the excerpt does not test. One is self-serving bias: writers may credit good outcomes to themselves and so underweight AI's share. The other is visibility: as AI becomes "an ambient environmental factor" in writing interfaces, writers may stop seeing it as something to disclose. The null effect of effort runs against the effort heuristic, under which more effort should lower the value assigned to AI's contribution.

The directness result lines up with Do writers actually edit AI-generated text before publishing?: this study treats direct incorporation as the case most in need of disclosure, and the edit-rate note describes AI paragraphs reaching readers with little revision. The writer-side gap sits beside Do users truly own the AI-generated content they produce?. Writers who claim authorship at a reflective level may find disclosure less necessary, but this excerpt measures disclosure judgments, not authorship claims or felt ownership. The writers' lower judgment also contrasts with the wish for AI-use visibility in Do writers want to see each other's AI prompts in shared editors?, though the audience differs: collaborators inside an editor, not readers of a published text. On the reader side, Does telling people an AI wrote something actually stop them from believing it? shows that knowing about AI does not fully block persuasion. This study shows readers judging disclosure necessary; it does not measure persuasion.

The excerpt gives no effect sizes, test statistics, vignette levels, sample composition or recruitment details, so the strength of each effect cannot be read from it. It also skips from the purpose subsection to the discussion, so the procedural hypotheses and the vignette design are not visible here. The limitations section adds that vignettes ask participants to imagine being readers or writers, that some may find the purposes unrelatable, and that writer-perspective participants may answer under social desirability bias, since disclosure carries moral weight. The implication, at the strength the evidence allows: the perception gap is a reason to treat writers' low necessity judgments as a target for guidance, as the conclusion suggests, not as a measure of how writers would disclose in practice.