“Used to Manipulate You” - AI Persuasion Bots Persuading Voters in Election Cycles

**SPEAKER_1** (0:01)
This episode is brought to you by Google Chrome. You think you know a browser, but Gemini and Chrome, that's new. It can help you with practically anything on the web, like restoring a vintage motorcycle from a 50-page restoration block, or finally break down that long article you've had open for weeks. Gemini and Chrome is here for it. Ready to make anything online make sense? There's no place like Chrome. Check responses set up required compatibility and availability varies 18 plus.

**Patrick Bet-David** (0:26)
They're saying 2028 will be the first AI election. And this is on page 5 if you guys want to go to it.
2028 will be the first AI election. This is an Axios story, which Tom, I'm going to come to you first on this one. I'm curious what you're saying. And then, Antonio, I'll come to you. So we're headed to the first true AI election. This will be the first election cycle where AI is both a dominant policy debate and a wildly deployed political weapon. AI allows both candidates and bad actors to persuade and perplex with unprecedented speed, scale, and precision in our election. Watch what works as those same techniques will likely be used against you.
Number one, persuasion bots. Campaigns are training text bots to hold back and forth conversations in their candidate's voice. Some voters engage for hours, granting precise knowledge of what ideas and language moves them the most. Yale researchers found a bot that admits it's a bot, it's just as persuasive to people as one posing as human. Powerful. Number two, simulated voters. Political operatives are running AI agents that mass produce synthetic voters and test ads against them. This helps fine tune the perfect message, including any words and image before it ever reaches any actual human voters. Number three, AI firehose. Those Yale researchers found AI conversational persuasion doesn't practically scale and noted that flooding social media could be the cheaper play. My bet, amplifying pre-existing views will win out and in turn, that'll shape how LLM language learning models answer future questions asked by voters. A huge new battleground will be attempting to mold what the most popular LLM, Claude Gemini, Grock, and Chad JBT say about candidates and their views. Tom, thoughts?

**Tom** (2:12)
And what's interesting is they left perplexity out of that list. And I think that's probably the best deep researcher, according to people I talk with. So the first thing they bring about is how AI can be used as a fair tool, a persuasion bot.

**Patrick Bet-David** (2:26)
While you're doing that, hey, guys in the back, can you do me a favor? I'm sorry, Tom, to interrupt you. Can you do me a favor? Ask the same exact question on four different LLMs. Ask who is more likely to win 2028 election?
JD Vance, Newsome, Rubio, or AOC, and why? Ask the same exact questions on the four and ask for stats. And let's see how they each answer the question. Go for it, Tom.

**Tom** (2:54)
So first of all, is the two ways that AI can be used as a tool. Persuasion blot, simulated voters. That's using it in a fair way. Here are the three areas that scare the hell out of me. The AI fire hose. When they say AI fire hose, what they're talking about is overwhelming.
People check out the likes and the comments before they go into the substance of the likes and the comments. It's like, Vinny, Adam, you're both up on social media, so we'll see something, maybe a comment about the WNBA by a notable person that we know is a real person. And we go, wow, 2100 likes and 14,000 comments. Holy crap. Right? And so we see that, and right now, AI can overflood that with bots, making a comment look bigger than it actually is to real actual voters. So that is the fake amplification I worry about, Pat. But here's the other two things I worry about. Speed. The speed of which a fake news or a fake story can flood social media, be mistaken as real, and then flooded with likes and comments is going to mean that voters, that the opportunity for voter deception is very, very high.
And that is the part that worries me, Pat, is how quickly can the social platforms, it's basically ex-Facebook and Insta, those are the big three, how are they going to be able to monitor if there are fake news stories going out in an election and somebody is being unfairly smeared, one way or the other? That's the one I worry about, Pat. I don't worry about the first two, which are tools. I worry about the fake news, filtering it by the socials, and then responding so that the voters can see, oh my gosh, that was a BS story. I can't believe that.