Admitting AI helped usually costs trust at first, but trust can rise when the alternative is getting caught hiding it.

This explores when telling people that AI was involved makes them trust it more, not less. The short answer from the corpus is that the first effect of disclosure is almost always a penalty. Trust goes up in two specific situations: when the alternative is being caught hiding AI use, and when people get to watch the AI deliver results over time. The library has nothing showing that a label alone raises trust the moment someone sees it, and that gap is worth knowing about.

Start with the baseline. Across 13 experiments with more than 5,000 people, admitting that you relied on AI made you seem less trustworthy, even to evaluators who liked technology Does disclosing AI use damage how trustworthy you seem?. The size of the penalty depends on the kind of writing. It hits hardest in personal, interpersonal writing, where readers expect real care and see AI as unable to provide it How does revealing AI authorship change reader trust?. In a news article it shrinks to almost nothing: less than 0.15 points on a 7-point scale, and LLM raters penalized it about as much as human raters did Does disclosing AI assistance make readers trust articles less?. So disclosure works less like a single trust switch and more like a test of whether AI fits what the reader expects from that genre.

The first way disclosure "raises" trust is relative. The same research program found that quietly using AI and then being found out causes a steeper trust drop than saying so upfront Does hidden AI use cost more trust when exposed?. Disclosure costs something, but it is the cheaper option once you account for the risk of being discovered. That risk is growing: unlabeled AI-assisted messages are currently trusted like human writing, but researchers expect that default to weaken as people learn how common AI help is Does trust in unlabeled AI messages decline as awareness grows?. Readers also rate disclosure as more necessary than writers do Do readers and writers differ on AI disclosure necessity?, so writers are probably underestimating how much hiding AI use will cost them.

The second way is over time. When people learn their partner is an AI, they first steer away from it. That bias reverses after repeated rounds in which they can see the outcomes Does revealing AI identity help or hurt user trust?. The important detail is that disclosure without visible results changed nothing. The label tells people what to pay attention to, and the track record is what actually adjusts their trust. A related finding: ChatGPT users build trust through how the conversation feels, meaning quick and responsive replies, rather than through checking accuracy Does conversational style actually make AI more trustworthy?. Trust in AI comes from the experience of using it, and the label only frames that experience.

The surprise is that the AI label may matter less than people think, because the audience often knows already. In a 1,500-person experiment, labeling a chatbot as AI had no effect on how persuasive it was, probably because its style gave it away. Disclosing its persuasive *intent*, though, cut its persuasive effect roughly in half Does telling people they are talking to AI change how persuaded they become?. Even when people know AI is involved, they become more skeptical but often stay persuaded: 34–62% did Does telling people an AI wrote something actually stop them from believing it?. So the disclosure that changes behavior is about what the system is *trying to do*, not what it *is*. Separately, people sometimes trust machines *more* for sensitive conversations, because nobody is judging them, which leads them to open up more Why do people share more openly with machines than humans?, How do people decide what to share with AI systems?. That is trust going up because of AI involvement, though it runs in the opposite direction: it is about what people are willing to tell the AI, not how much they trust what it produces.