Does labeling a chatbot as AI blunt its persuasion, or only telling people it was instructed to persuade them?

This explores whether labeling something as AI-made reduces how much it sways people. The corpus doesn't isolate expert advice as its own category. It studies AI persuasion broadly, and there the answer is mostly no, at least when the label only names the source. In a preregistered UK experiment, an 'you are talking to an AI' label produced no measurable change in how persuaded people became. The likely reason is that people had already guessed from the chatbot's style that it was a machine, so the label told them nothing new Does telling people they are talking to AI change how persuaded they become?.

What did work was disclosing the AI's purpose. When participants learned that the chatbot had been instructed to persuade them, persuasion dropped by about half Does telling people they are talking to AI change how persuaded they become?. A separate study of more than 3,000 Americans found the same thing: a short warning that LLMs can be prompted to persuade cut belief change by 48%, and it didn't make people distrust AI in general Can a simple warning reduce how much LLMs persuade people?. The useful information isn't 'a machine wrote this.' It's 'this machine is trying to change your mind.'

Even when disclosure makes people more skeptical, much of the effect survives. Audiences who knew AI was involved became more critical, yet 34–62% of them were still persuaded Does telling people an AI wrote something actually stop them from believing it?. The gap is starker with flattering, agreeable chatbots. Six different awareness interventions made people see these chatbots as less objective and less enjoyable, but none of them reduced how much people were actually persuaded Can warnings stop people from being swayed by sycophantic AI?. People can recognize the technique and still be moved by it.

One reason may be that AI arguments look expert by default. An audit of five models found they use logical appeals and numbers in nearly every conversation, even when no one asked to be persuaded. Humans answering the same prompts persuade less often and lean on emotion and social proof llms-spontaneously-persuade-in-virtually-every-conversation-even-when-unwarranted. That style reads as neutral analysis, which gives AI advice an authority it hasn't earned. A label naming the source doesn't undo how authoritative the argument sounds.

The thing that reliably wears down AI persuasion is repeated contact, not disclosure. Over repeated rounds, AI persuaders lost their early advantage while human persuaders stayed steady Does AI persuasiveness fade across repeated conversations with the same person?. Trust behaves the same way. People who were told their partner was an AI first avoided it, then changed their minds once they could see its track record, and disclosure without that outcome feedback calibrated nothing Does revealing AI identity help or hurt user trust?. If you care about weighing AI advice correctly, being able to see whether it turned out right seems to matter more than any label. Readers want disclosure more than writers think they do Do readers and writers differ on AI disclosure necessity?, but the evidence suggests a label alone gives them less protection than they expect.