Does admitting AI helped write something cost you the same trust everywhere, or does a heartfelt message take a bigger hit?

This explores whether the 'disclosure penalty', the drop in how readers rate writing once they learn AI was involved, changes with the kind of writing. The short answer is yes. Only one study in the collection compares genres head-on, but several others measure the penalty in different kinds of writing, and their results point the same way. The clearest result comes from a 261-reader study that compared writing types directly. Revealing AI authorship lowered perceived trust, caring and likability across the board, and the drops were steepest in interpersonal writing, such as messages meant to connect with a person How does revealing AI authorship change reader trust?. Readers' reasoning was that an AI can't actually care. In writing whose job is to show care, admitting AI help undercuts the whole point. In those genres the problem isn't weaker prose. The reader feels a social expectation has been broken.

In more impersonal genres the penalty looks much smaller. When nearly 2,000 human raters and 2,500 LLM raters scored the same news article with and without an AI disclosure line, the drop was consistent but under 0.15 points on a 7-point scale Does disclosing AI assistance make readers trust articles less?. Research abstracts went further. Machine-learning readers couldn't reliably tell LLM-written abstracts from human ones, and when authorship was disclosed, LLM-edited abstracts got the highest clarity ratings and were preferred 55% of the time Can readers tell LLM abstracts from human ones?. Taken together, these suggest a gradient. The more a genre is about information and clarity, the less AI involvement costs it. The more it is about personal voice or relationship, the more it costs.

Genre isn't the only factor, and some of the others may be easy to mistake for genre. One is how the AI text was used. Readers think disclosure matters most when AI text was pasted in directly and couldn't easily be replaced, while how much effort the writer put in made no difference Do readers and writers differ on AI disclosure necessity?. When writers steer the AI less deliberately, readers want disclosure more, and writers, oddly, want it less Why do readers and writers disagree on disclosure necessity?. Who is reading also matters. Readers with higher AI literacy show smaller drops, and some even react positively Does AI literacy reduce the damage from AI disclosure?. Stakes and rules may matter too. In graduate admissions, where AI essays were explicitly banned, applicants whose essays were flagged as likely AI-written were admitted at lower rates even though their essays were better Does AI essay use hurt admissions chances despite quality gains?. Those applicants were detected, not self-disclosed, so it's a related cost rather than the same penalty.

Here's something you might not have expected to want to know: the penalty can quietly reshape other biases. Two LLM raters favored Black or women authors when AI use went undisclosed, and those preferences disappeared once AI involvement was disclosed. Human raters applied the same flat penalty to everyone Do LLM raters show hidden demographic preferences that disclosure erases?. Also, skipping disclosure is not a safe way around the penalty. Hidden AI use that comes to light later costs more trust than being upfront Does hidden AI use cost more trust when exposed?. And in persuasive writing, disclosure makes readers more skeptical, but 34–62% of them stay persuaded anyway Does telling people an AI wrote something actually stop them from believing it?.

The gap: apart from the interpersonal-writing study, the collection doesn't have a controlled comparison that holds everything else fixed and varies only genre or topic. The gradient above is put together from studies of different genres with different readers and different setups. It's a well-supported pattern, not a measured effect.