Tell readers an AI wrote a piece and they read it more warily, but does it still change their minds?

This explores whether labeling something as AI-made weakens its power to change minds, or whether it still works once the reader knows. The short answer from the corpus is that it still works, and mostly as well as before. In one study, audiences who knew AI was involved read more critically, yet between a third and nearly two-thirds of them were still persuaded Does telling people an AI wrote something actually stop them from believing it?. Disclosure puts readers on alert, but being on alert isn't the same as resisting. The authors call disclosure 'necessary but insufficient' as a safeguard.

The penalty for disclosure is also smaller than you might expect. When an identical news article carried an AI-disclosure line, both human and LLM raters scored it lower, but by less than 0.15 points on a 7-point scale Does disclosing AI assistance make readers trust articles less?. The penalty gets bigger when the writing is personal. Readers' trust, sense of caring, and liking for the writer dropped most sharply in interpersonal writing, because people don't believe AI can mean empathy How does revealing AI authorship change reader trust?. So disclosure mostly changes how readers feel about the *author*. It does much less to the *argument*. Even that penalty shrinks for readers with more AI literacy, and some of them view AI use positively Does AI literacy reduce the damage from AI disclosure?. The penalty also fades over time: people who at first avoid a disclosed AI partner switch to preferring it after they repeatedly see good results Does revealing AI identity help or hurt user trust?.

The most useful finding here is about what kind of disclosure you give. Saying 'AI was involved' is a weak defense. Saying 'AI can be prompted to persuade you' is a much stronger one. In experiments with more than 3,000 Americans, a brief warning of that kind cut belief change from a persuasive AI conversation roughly in half, and it didn't reduce people's general trust in AI Can a simple warning reduce how much LLMs persuade people?. What seems to protect readers is knowing what the AI is trying to do, not just knowing where the text came from.

There's a less obvious problem: AI may shape persuasion even when nobody discloses anything. A study of nearly 3,000 writers and 11,000 readers found that AI writing assistance shifted how readers saw the writer on all 29 measured traits, toward more confident, more extreme, higher quality, and more agreeable Does AI writing assistance change how readers perceive the writer?. If AI makes writers sound more confident and authoritative, it may be adding persuasive force that a disclosure label can only partly undo. Suspicion also lands on the wrong people: readers who accuse comments of being AI-written tend to target human writers, based on features that don't actually tell AI and human text apart Do unfounded AI accusations harm human writers instead?.

One limit on the evidence: only two of these studies measure belief change directly. Most measure trust, ratings, or impressions of the writer. The pattern still holds across them. Disclosure works as a mild social signal, and it works best as a protection against persuasion when it tells readers what the AI was trying to do. If you want to see why disclosure norms are hard to settle, readers and writers disagree about when disclosure is even needed Do readers and writers differ on AI disclosure necessity?.