Writers will show their AI use to collaborators, yet readers think it should be disclosed more than writers do.

This explores the gap between how writers treat their own AI use and how readers react when they find out about it. The corpus partly corrects the question's premise. In shared writing tools, writers don't seem to want to hide AI use from each other. Sixteen pairs of writers clearly preferred editors that showed when, where, and how a partner had prompted AI, because it helped them follow each other's thinking and check generated text. Some did find full visibility intrusive or felt self-conscious about it Do writers want to see each other's AI prompts in shared editors?. The real divide is between writers and the readers who eventually see the finished text.

That divide is well measured. In a 727-person study, readers consistently judged AI disclosure more necessary than writers did. Disclosure felt most necessary when AI text went straight into the piece and couldn't easily be replaced. How much effort the writer put in made no difference Do readers and writers differ on AI disclosure necessity?. The most revealing detail: when writers steered the AI less deliberately, readers thought disclosure mattered more, while writers thought it mattered less. The authors call this reversal surprising, and it comes from hypothetical scenarios Why do readers and writers disagree on disclosure necessity?. Still, the pattern fits a convenient blind spot. The less a writer actually shaped the text, the less they feel they owe anyone an explanation.

Writers have a practical reason to stay quiet, because the penalty for disclosing is real. Telling 261 readers that AI was involved lowered how trustworthy, caring, and likable they judged the writer. The steepest drops came in personal writing, where readers saw AI use as faking an empathy it can't have How does revealing AI authorship change reader trust?. The size of the penalty depends on context. On a news article, both human and LLM raters marked a disclosed version down by less than 0.15 points on a 7-point scale Does disclosing AI assistance make readers trust articles less?. So the cost is high for a condolence note and close to nothing for a news report. A writer who hides AI use may simply be reacting to the kind of writing it is.

The penalty also isn't fixed. Readers with more AI literacy shifted less after disclosure, and some even viewed it positively Does AI literacy reduce the damage from AI disclosure?. In a related setting, people at first avoided partners revealed to be AI. That preference reversed after repeated interactions where they could see the results Does revealing AI identity help or hurt user trust?. This suggests that much of today's disclosure penalty reflects unfamiliarity and could shrink once people have experience with AI-assisted work.

What you might not expect: hiding AI use doesn't keep it from changing how readers see you. AI-assisted writers were judged far more educated, wealthier, and more likely to be native English speakers than they were. Researchers call this "identity laundering," where a distinctive voice gets flattened into a generic, privileged persona Does AI writing make authors seem more privileged than they are?. Meanwhile, human writers are being falsely accused of using AI. Their accused comments don't actually differ from other human writing, so the accusations work more as gatekeeping than as detection Do unfounded AI accusations harm human writers instead?. Hiding AI use, disclosing it, and being accused of it each carry a different social cost, and none of them reliably tells readers how the text was actually made.