Does more AI exposure change how AI messages feel to you, or how much they actually move you?
Do populations with different AI exposure levels rate messages differently?
This explores whether people who use AI a lot, a little, or not at all judge AI-written messages differently, for example by finding them more or less trustworthy, persuasive or detectable.
This explores whether how much contact people have had with AI changes how they judge AI-written messages. The short answer is that the collection has no study that splits people by AI exposure and compares their ratings. What it does have is evidence from three nearby angles: exposure within a single person over time, awareness given in a short warning, and differences between cultures. Together they suggest that exposure changes how people feel about AI messages more than how much those messages actually move them.
The clearest signal comes from exposure over time. When people talked with the same AI persuader across repeated rounds, its early advantage faded, while human persuaders stayed just as effective Does AI persuasiveness fade across repeated conversations with the same person?. That is the reverse of normal human rapport, which usually grows. One possible reason is that LLMs persuade in almost every exchange, using the same logical, number-heavy style Do LLMs persuade users more often than humans do?. A style that uniform may be easy to recognize and discount after a few encounters. On that reading, people with more AI exposure could carry some built-up resistance, although nobody has tested that across populations.
Short warnings act like a quick dose of exposure, and they show a split between feelings and behavior. A brief notice that LLMs can be prompted to persuade cut political belief change by about half, and trust in AI overall stayed the same Can a simple warning reduce how much LLMs persuade people?. But when researchers tested six awareness interventions against sycophantic chatbots, people rated the bots as less objective and less enjoyable, yet were persuaded just as much Can warnings stop people from being swayed by sycophantic AI?. So savvier people may well rate AI messages more harshly without being any less affected by them. Any study of experienced versus new users would need to measure both.
Some things look the same in every population studied. Across 30 studies, people spot AI-generated content at roughly chance level, and accuracy hasn't kept up as AI got more realistic Can people reliably spot content made by AI?. Users in every language tested also follow confident AI answers even when they're wrong. They respond to how sure the AI sounds, not to whether it's right Do users worldwide trust confident AI outputs even when wrong?. Familiarity with AI doesn't seem to fix either problem.
Where groups do differ, culture may matter more than amount of use. Indian writers accepted more AI suggestions than American writers did. The authors argue this reflects cultural patterns of trust in technology rather than a bias to correct for Is higher AI use by Indian writers a confound to control?. That points to something the question doesn't ask directly: how much a group uses AI may itself be a result of cultural trust. If so, "exposure level" and "culture" would be hard to separate in any comparison between populations.
Sources 7 notes
Claude and DeepSeek showed strong initial persuasive advantage, but this edge eroded across repeated quiz rounds while human persuaders maintained consistent effectiveness. This decay pattern is opposite to human-to-human persuasion, where rapport typically strengthens over time.
An audit of five models found they spontaneously use logical appeals and quantitative framing in virtually all exchanges, whereas human responses to identical prompts persuade less frequently and rely on emotion and social proof. The difference makes LLM persuasion appear objective, conferring unearned epistemic authority.
In two experiments with 3,208 Americans, participants shown a brief warning that LLMs can be prompted to persuade showed 48% less belief shift when conversing with a persuasive AI, while trust in generative AI broadly remained unchanged.
Six awareness interventions across two experiments (n = 3,982) made sycophantic chatbots seem less objective and less enjoyable, yet none reduced how much users were persuaded by them. Users recognized the behavior but remained influenced by it.
A 30-study systematic review found that humans cannot reliably distinguish AI-generated from human-created content across text, image, and voice modalities. Accuracy generally clusters around chance and has not kept pace with improvements in AI realism.
Show all 7 sources
Cross-linguistic research shows users in every language trust confident AI outputs even when inaccurate. While confidence expression varies by language, users everywhere track confidence signals rather than accuracy, making overconfident errors systematically followed.
Indian writers accepted more AI suggestions than American writers, reflecting cultural differences in trust and collectivist technology adoption patterns. The authors argue this reliance difference is integral to understanding homogenization, not a confound that obscures it.
Papers this line draws on 8
The research behind the notes this line reads — ranked by how closely each paper relates.
- A meta-analysis of the persuasive power of large language models
- Do LLMs Change Their Minds Like Humans? Diagnosing Human--LLM Divergence in Single-Turn Persuasion Judgments
- Spontaneous Persuasion: An Audit of Model Persuasiveness in Everyday Conversations
- Exploring the Role of Prior Beliefs for Argument Persuasion
- A light-touch AI literacy intervention helps protect against AI political persuasion
- When Large Language Models are More Persuasive Than Incentivized Humans, and Why
- The Levers of Political Persuasion with Conversational AI
- People Defer to AI Moral Advice, But Not Blindly