When you admit you used AI, the penalty falls on you, and even knowing that makes people less willing to say so.

This explores who bears the cost when someone reveals they used AI, and whether some people or situations feel it more than others. The clearest finding is that the penalty falls on the person who discloses. Across 13 experiments with more than 5,000 participants, saying you relied on AI consistently made you seem less trustworthy, and this held even among tech-savvy evaluators and people who liked technology Does disclosing AI use damage how trustworthy you seem?. The people doing the work see this coming. In four experiments with 4,439 participants, AI users expected to be rated as less competent and less diligent, and as a result said they were less willing to tell managers and colleagues they had used it Do people fear judgment when they use AI at work?.

That expectation sets a trap. Hiding AI use is the worst option if it comes out later: AI use that was kept quiet and then discovered caused a steeper drop in trust than upfront disclosure did Does hidden AI use cost more trust when exposed?. So the people hit hardest are those who try to dodge the penalty by concealing AI use and then get caught. Readers and writers also disagree about when disclosure is owed. Readers rate it as more necessary than writers do, especially when AI text is pasted in directly and couldn't easily be replaced. How much effort the writer put in made no difference to that judgment Do readers and writers differ on AI disclosure necessity?. Writers who think a little AI help doesn't count may be misjudging their audience.

On the audience side, the corpus pulls in two directions. One study found that readers with higher self-reported AI literacy had smaller drops in their opinion after learning AI was used, and some even saw it positively Does AI literacy reduce the damage from AI disclosure?. That cuts against the finding that tech-savvy evaluators still penalize disclosure. A reasonable reading is that literacy softens the penalty but doesn't remove it. Also, feeling positive about technology is not the same as understanding how AI is actually used. Time matters as well. When people can watch an AI partner's results over repeated rounds, their initial bias against it reverses. Disclosure without that feedback calibrates nothing Does revealing AI identity help or hurt user trust?. The penalty may be steepest in one-off judgments, such as a single email, application, or article, where the audience never gets to see results pile up.

Research on AI as the speaker adds a useful contrast. Labeling a chatbot as AI did not change how persuaded people were, but telling them its persuasive intent cut persuasion roughly in half Does telling people they are talking to AI change how persuaded they become?. Even when audiences know AI was involved, 34–62% are still persuaded Does telling people an AI wrote something actually stop them from believing it?. The lesson is that disclosure tends to cost the discloser more than it protects the audience. What makes disclosure useful is what it tells people about purpose, not just the presence of AI.

One limit: this corpus doesn't break the penalty down by demographics such as age, profession, gender, or seniority. So it can't say whether, for example, junior workers or certain professions are hit harder. What it does show is a pattern based on situation. The penalty is largest for people disclosing in one-off judgments to less AI-literate audiences, and largest of all for people who hide AI use and are found out.