When people learn an AI was involved, does that distrust fade with repeated use, or does seeing results make the difference?

This explores whether the trust hit that comes from revealing AI involvement wears off with familiarity. The short answer from the corpus: repetition alone doesn't seem to fade it, but repetition *with visible results* can. The collection has only one study that actually tracks people over repeated encounters. Most of the other evidence comes from single-moment snapshots, so the honest picture is part answer, part open question.

The clearest evidence of fading comes from studies of people working alongside an AI partner. When the AI's identity is revealed, people at first avoid it. After repeated rounds where they can see how things turned out, that preference flips Does revealing AI identity help or hurt user trust?. The detail worth noticing is what drives the change: it isn't exposure, it's feedback. People who are told 'this is an AI' but never see consistent outcomes don't recalibrate at all. So the penalty fades through learning, not habituation. People need evidence that the AI performs, not just more contact with it.

That matters because the most common form of the penalty may not involve any feedback loop. When the question is how someone who *used* AI comes across, rather than how well the AI performs, the damage looks stubborn. Across 13 experiments with more than 5,000 people, admitting AI reliance lowered how trustworthy someone seemed, even to tech-savvy evaluators who liked technology Does disclosing AI use damage how trustworthy you seem?. If enthusiasm and familiarity don't protect against it, simply getting used to AI may not either. The penalty is also steepest in personal, relational writing, where readers judge that AI can't genuinely care How does revealing AI authorship change reader trust?. No amount of repeated exposure produces an 'outcome' that disproves that kind of judgment. The corpus itself flags this as unresolved. No data yet tracks whether the social penalty fades as AI becomes ordinary, and part of what people object to may be the AI's agency, which won't go away with routine Does the social penalty for AI use fade as the tool becomes ordinary?.

One twist is that familiarity may cut the other way. Right now, readers trust unlabeled AI-assisted messages about as much as human-written ones. The authors suspect that baseline could erode as people realize how much text is AI-assisted, though their one-time study can't show it Does trust in unlabeled AI messages decline as awareness grows?. Hiding AI use is a risky bet in any case. When concealed use is discovered later, trust drops more steeply than it does after upfront disclosure Does hidden AI use cost more trust when exposed?. So the 'penalty' may be shifting away from disclosure and onto non-disclosure.

The lateral surprise is that a different kind of fading is already well documented: people stop *checking* AI output. Fluent answers and the cost of verifying them lead users into what one note calls 'cognitive surrender.' In the studies it describes, about 80% of AI outputs were adopted without challenge When do users stop checking whether AI output is actually backed?. Disclosure does make audiences more critical, yet between a third and nearly two-thirds of them are persuaded anyway Does telling people an AI wrote something actually stop them from believing it?. Put together, it's possible that how people *say* they feel about AI involvement stays negative while their actual vigilance quietly fades. That would be the opposite of what disclosure is meant to protect.