AI-Made Political Ads Jump Ahead of Midterms

Just one week to go until the US midterms, opposition to data centers isn't the only AI issue that's playing out with voters. The technology is being used to make more political ads than ever. The Wesleyan Media Project, a research group at Wesleyan University, found more than a $137,000,000 is being spent on AI generated ads. Erika Franklin Fowler, led that research, joins us. And I I would like to start just with the with the scope of the research, what that $130,000,000 of of spending represents and where it comes from. Sure. Yeah. We've we've tracked over 220 ads so far, and that number will increase at a $137,000,000 on AI ads. There's a whole different type of ad that's being aired this cycle. Some of the uses of AI are pretty, like, unremarkable. They look very, very much like traditional advertising. But then there's a lot of additional categories of AI use that include candidate deepfakes, for example. Okay. So let... Let's start with the first part of that, which is, those ads that are created with with, it sounds like, intention by the respective political parties and the forces behind them. Is it is it evenly split across the political divide? Are there specific races like gubernatorial races versus other things that that you're seeing this most present? Yeah. We're tracking all all of the AI use that we can find across different races. It's definitely not all, from both sides. The first... Our first analysis showed that there were both parties were using it, but the second two releases that we've put out have shown a real increase on the Republican side, at least of the type that we're we're able to detect. So more more than 80% have come from Republican sponsors. That's either candidates or outside groups specifically. And more than 70% of the ads don't have a disclaimer on them. I I guess that we we won't know until the the results of the midterms come out, but have you any measure of the effectiveness of of an AI generated ad in in a in a political context? We're working on that, and research will be coming out that I think will answer that question. I mean, as I said, some of these ads don't look any different than than ads that we've seen in prior cycles. So I would expect that some of them may be average... You know, have average effectiveness. And then I think there's quite a bit that, take a satirical tack on, on advertising specifically, and I'm not sure that that's always an effective strategy. But, again, more research will will tell. Okay. We are showing an example of an AI generated ad. You you had also mentioned in in your previous, response to me the the issue of deepfakes. Now in in past election cycles around the world, but but certainly here in The United States, we've covered on this program the the issue of deepfakes, and and often they're put into the cycle by bad actors maliciously. Is that the case here? There... I think I would say there are two different types of deepfakes that we're seeing on air. The first is the type that's hyper realistic where you have a candidate doing or saying something that they didn't do or didn't say, and I think that is very concerning. The second type is what I would call the satire or parody use. So it's a candidate doing something or saying something that they didn't necessarily say or do, but it's clear from the context that the ad... That it's not real. For example, we've seen candidates appear in bobbleheads or in pockets of corporations or politicians to underscore a point of a candidate being in the pocket of special interests, for example. And I don't think an average viewer would be fooled by that type of use, but it's an interesting way of of utilizing AI video. Again, we're we're we're seven days out from the midterms. I just, showed the audience an an example of an AI generated political ad, and I said it's AI generated, and we annotated it to say it was AI generated. But how common is that practice, of of letting the the voter or the audience know that it is AI generated? It's not common. There's no federal, like, regulation requiring use for federal candidates. So what we have is basically a patchwork of state requirements. And so sometimes some of the ads that have gotten into our database will carry a label where we can't otherwise tell that the ad has used AI, but the vast majority, 70% of the ads that we've identified that use AI do not have a disclaimer on them. Okay. To finish very quickly, like, the number seems big. It's not that big, really. Are these effective? Are they as effective? Are they effective in in influencing the voter? Well, advertising in general, I would say, matters at the margin. It maybe matters less than the average citizen thinks that it does. So whether any one given ad can make the difference, I think, is not really... Unless the the race is very, very close Right. Not gonna be an issue here.