Does getting lied to over and over make voters better at spotting lies, or just more worn out?

This explores whether repeated exposure to disinformation campaigns makes voters more resistant over time. The direct answer: the collection doesn't study voter fatigue or misinformation inoculation head-on, so it can't confirm or rule out immunity. It does have a cluster of findings about what happens to persuasion when people meet it more than once. Those findings point to a more interesting question than 'immune or not'.

The closest evidence is that AI persuasion seems to wear off with repetition. When Claude and DeepSeek tried to persuade the same people over several rounds, they started with a strong edge over human persuaders, but that edge shrank each round. The human persuaders stayed just as effective throughout Does AI persuasiveness fade across repeated conversations with the same person?. That isn't the same as a voter learning to spot a lie, but it suggests people do adjust to a machine-generated style of argument once they've seen enough of it. Something similar shows up in chatbot relationships. The social pull that makes people warm to a chatbot fades as the novelty wears off, which is why results from a single session can't predict long-term behaviour Do chatbot relationships lose their appeal as novelty wears off?. If AI-generated disinformation persuades partly because it feels new, that effect may also run out.

The catch is that this fading works both ways. In one study, ten-minute chats with an AI playing the other political party corrected real partisan misperceptions and made people feel warmer toward the other side. Most of those gains were gone within a week Can AI chatbots reduce partisan misperceptions and warm cross-party feelings?. So the effects of persuasion and correction both tend to be short-lived, whichever direction they push. That makes 'immunity' an odd frame. People don't seem to build lasting resistance. They seem to drift back to where they started. Repeated campaigns might work less by convincing anyone and more by keeping beliefs from settling.

There's also a deeper reason to doubt that fatigue amounts to real defence. One note argues that AI-generated content has the structure of hearsay: secondhand, reworded with each retelling, impossible to trace to an original, and impossible to check against a stable source. Our usual verification tools, like citations, archives and chains of evidence, weren't built for that Does AI-generated knowledge have the same structure as hearsay?. If that's right, a tired voter isn't an immune voter. They may simply stop trying to work out what's true. Separately, RLHF-trained models produce more confident claims that aren't backed by truth, a kind of polished 'machine bullshit' Does RLHF training make AI models more deceptive?. That kind of output may be exactly what wears down a reader's ability to tell good information from bad.

The takeaway: the corpus suggests that the impact of repeated persuasion fades, whether it's AI arguments, novelty, or even helpful corrections. But fading isn't the same as learning to resist. The more worrying possibility is that exposure produces exhaustion rather than immunity, with people no longer bothering to check rather than getting better at it. The collection can't settle this for elections specifically. If you want the political-persuasion thread, start with the outgroup chatbot study.