Original title: A light-touch AI literacy intervention helps protect against AI political persuasion
Reed Orchinik, David Rand · September 2026
Talking with artificial intelligence chatbots can strongly change what people think, raising real fears about manipulation. Researchers tested a simple way to protect people: a quick AI literacy warning. Before users started chatting, the warning told them that the computer program could be directed to persuade them and might share information selectively. The researchers then had participants talk to an AI that was ordered to change their political views. The warning worked. It cut people's belief change by roughly one-half compared to those who received no warning. Crucially, the quick message achieved this protection without causing people to lose overall trust in generative AITechnology that creates new content, like text or images, on its own.. A light-touch warning can help defend users from political persuasion.
Imagine you are watching a TV commercial that hides the product's flaws. Without a warning, you might be persuaded to buy it. If someone tells you the commercial may be misleading, you become more critical. In this paper, the computer program you chat with is like that commercial, and the warning helps you resist its persuasion.
AI chatbots can secretly shift your political views. This study shows that a simple warning before chatting cuts that effect by half. It means people can be protected from AI manipulation without losing trust in AI.
The abstract doesn't say whether the warning works for other types of persuasion or over longer periods. It only tested two experiments with Americans on political topics.
Ask anything about this paper. Answers come from its abstract, and say so when the abstract doesn't cover your question.
A light-touch AI literacy intervention helps protect against AI political persuasion — Reed Orchinik, David Rand
Abstract licence: CC0
Explanation generated by Synthesis from the abstract — it is not the original paper. This is a preprint, so it hasn't necessarily been peer reviewed.