Eric Wilson 0:04

Welcome to the Campaign Trend Podcast, where you're joining in on a conversation with the entrepreneurs, operatives, and experts to make professional politics happen. I'm your host, Eric Wilson. Our guest today is BJ Martino, president and CEO of the Tarrence Group and one of our side's most respected pollsters. So I'm very grateful for him for making time for us today. It's a very busy season, as you can imagine. BJ was part of the bipartisan research team behind the AAPC Foundation's new national survey of likely voters on the topic of what else? Artificial intelligence. In this episode, we're digging into those findings and some insights from the poll. BJ, two things surprised me in this poll. And I want to start with the first. So for the last four years or so, AI in politics is the thing I spend most of my professional time on. So I'll admit that I'm in a bubble. But I was really shocked to see that just 4% of voters said they went to an AI tool for election information in the past month. But 38% said they got information from a Google search AI summary, and about half of those didn't even realize it was AI. So what's going on here?

B.J. Martino 1:23

Well, I think the difference is between purposeful use of AI tools to find election and campaign information versus sort of this we called it ambient use, meaning there are voters who are just Googling and they're reading summaries at the top of their Google search. They don't necessarily realize, many do, that those are AI-generated summaries. But with the high use of social media that we saw in the poll and the high use of all these other tools that are giving them AI-generated content, many voters, most voters, are being exposed to it. Fewer are actively using those AI tools to seek out that election information. And while we only saw 4% said that they looked up something in the last month, when we sort of broadened the aperture and asked voters, are you using AI? Have you ever used AI for election or campaign information? It's about 20%. So one in five have actively or purposely gone and sought out that sort of information. But we have to accept the fact that everyone's getting it now, effectively. They just maybe don't know, or it's not intentional that they're going to go ask the AI tool. They're just getting that AI generated content back at them.

Eric Wilson 2:49

Yeah, I suspect it's one of these things where they, you know, they just don't know what's going on. You know, the the old joke of, I can't believe they built um uh Shazam, the music lookup app, without AI. And it's like, well, no, that is AI. We just didn't really call it that back then. Or, or similarly, um, you know, if we ask voters, have you accessed any HTML slash CSS files in the last 30 days? And most of them would say, no, but well, we all know that that's what a website is. Uh so I think uh part of it is, you know, once the the technology sort of fades to the background, people uh understand how it's being used.

B.J. Martino 3:31

Which is why we wanted to ask in the survey the question different ways and get at different uses and not being always specific, have you used AI tools or have you sought AI-generated content, but just ask them generally about their sort of behavior, about again, use of social media, about Googling uh information online and reading the search summary, not telling them that that's AI. And that's where we discovered this chasm between the intentional use and the uh the ambient or unintentional use that most voters are receiving.

Eric Wilson 4:07

Yeah, and it's sort of um, you know, this it's so hard. I'm not telling you something that you don't already know, so hard to get inside voters' heads and ask them to like give us that information. Um talk to us about the the partisan and age gaps you found. So the where do Republicans and Democrats differ on AI? And then what about younger versus older voters?

B.J. Martino 4:32

Yeah, there is in terms of use in this study, we really didn't see a significant difference between Republicans and Democrats in the frequency of use of AI tools. We asked them how often on a daily basis or occasionally that sort of thing. And again, Republicans and Democrats, not that much difference. But when we asked them about do you think that AI is going to do more harm or more good, what we ended up finding is that Republicans are about split. They were net negative by only three points, whereas Democrats fully net 40 points saying that they believed that uh AI was going to do more harm than good. So a big partisan gap.

Eric Wilson 5:14

What why is that? Uh before we get to the younger voters, uh, because I get asked about that all the time by reporters, and I I have to I have to make some guesses. I think it's a few factors like labor, environmental concerns. I think who who's in the White House right now and promoting AI. Do you have anything to add to that, Brew?

B.J. Martino 5:33

Yeah, I think it's a it's it's difficult to distangle those views of AI with their views of who are the big promoters of it right now, and certainly the president's positions on and what he has said. By its nature, anything that the president says automatically becomes polarized, you see in the in the public opinion research data. So I do think that that is that is certainly a part of it, uh, of what's going on. Now, by age, it's interesting. Yeah, obviously, younger voters are more likely to say that they are users of these tools. Um, there are some seniors certainly who are using them, and they tend to use them, though, much more for everyday household or healthcare-related questions that they're asking. And that showed up in the data. What's interesting though is you see some of the youngest cohort of voters are some of those who, while they may be using it at a higher rate, they're also a little bit more skeptical uh of it, of it right now. And there's so there's this interesting sort of curvilinear relationship between your views of AI, where there's a point at which you're the youngest cohort and maybe the oldest. And if you're using it, you're a little better better about it. But if you're not, you're you're significantly more negative.

Eric Wilson 6:47

Yeah, that was an interesting cut too. I wonder if it's job market difficulty, because this is under 35, right? So it could be job market difficulties, it could be things that they're hearing from their professors in in college, right? Again, it's just fascinating to think about what are the inputs that are are going into these opinions.

B.J. Martino 7:08

Yeah, we didn't get into the qualitative in this study to understand what underlies that, but certainly for that cohort who are many of whom just entering the job market or have just entered the job market, to have this sense of here are these tools that do the job that I'm doing as well as I can at this point in my career, uh, I think create some of that negativity.

Eric Wilson 7:32

So now that we're in this final stretch, I I mean, I I expect the usage by voters is going to increase, you know, because we are now in early vote, getting get out the vote time. Um, what practical advice would you give to a candidate at this stage in the campaign uh uh about uh using AI or trying to shape what AI is saying about their campaigns? Care to make any recommendations there?

B.J. Martino 7:57

I think the the most important thing for campaigns right now is to pay attention to what the summaries are saying about you. Go into every single search engine and everywhere that someone might go and receive that ambient AI-generated content back at them and understand what does the Google search summary say about my campaign or my issue right now? And make sure that the foundational sources that they're drawing from uh are correct and you're optimizing for that. I think the other thing we have to be careful about is there is there was a bit of a glamour around uh using some of these AI tools to create video content. Uh, and a lot of the voters in in this survey told us they don't find that acceptable, especially when it comes to using that AI-generated content, video to put words in your opponent's mouth. False and misleading information is the biggest concern they told us about AI tools in the beginning. And when we got to that topic of what campaigns should be can and can do and what's acceptable and not acceptable, uh, they pushed back on that pretty hard. And I think ultimately it all falls back on our campaigns to just listen to and obey uh what our AAPC uh rules of the road are, which is we don't present false and misleading information uh to uh to voters and and make sure that that's not in the content that we deliver. Finally, I think that there deep fakes and and accusations of using AI and manywhere are going to become the new they they did a push pull. Now, that accusation that my industry has dealt with for years when we're doing legitimate message testing, and another site says, oh, they're doing a push pull. It's like, well, that's that's not a push pull. There's a difference between a legitimate message testing survey and a push-pull. I think here there's going to be these accusations of, well, that didn't actually happen. That video is AI generated. It's a deep fake. We've already seen some indications of that. And so for campaigns, be ready. Have your sources, have the original sources available. You'd be able to point to where those things are from. And on the other hand, be ready for those tools potentially being used against you. Yeah. And have everything. So it's just going to be a lot more doing the due diligence and being ready to show your work.

Eric Wilson 10:24

Yeah. Uh it you've just got to be really buttoned up right now. And and and BJ, I'm sure you have never had this happen, but I'd like to argue with the results of your poll that don't conform to my existing ideas and priors. It's never happened before. I'm ready for it. So 70% of voters said that it was unacceptable for a campaign to use AI to answer voters' questions on their website or by text if it wasn't disclosed. And the disclosure only moved that to 65%. So, two observations that I'd like to like you to react to. One, the term AI is very nebulous and it's a boogeyman in public discourse right now. So this strikes me as what I like to refer to as a do you want to be punched in the face question? Uh and two, I think we get more grace from a voter, we would get more grace from a voter if if we tell them the options are either to use AI and answer their questions or to not use AI and totally ignore them. How am I doing on getting better results?

B.J. Martino 11:27

I think if we change the nature of the question and said this is the only way you could get this information, that the campaign wouldn't be able to get back to you for a week or two, and but otherwise, or you can get an answer right now, I would certainly see uh and think we would see a little bit more uh acceptable measures to that. But at the same time, I do think voters are concerned about this increasing dehumanization uh of their everyday life. Everywhere they encounter now, they're uh on the phone uh with a with a company they're trying to reach, and they have to go through five, 10 minutes of dealing with the AI auto-generated voice that's telling them they shouldn't get an operator yet. They're dealing with all these things where they're not getting the human-to-human contact. Uh, and and I think that there, to some extent, in this pushback we see, is a pushback against this feeling of I'm being shoved off to the AI because someone doesn't really care about me.

Eric Wilson 12:33

Yeah, I I I I think that that's uh a piece of advice I constantly give to to candidates is you know, sometimes people ask me, well, you know, you know, it's really hard for me to make videos uh for social media. Could I get an AI avatar to do that? And I was like, well, you could, but that defeats the purpose because voters are hiring you for a job. Your name is on the ballot, and and so they're hiring a person, not not your AI. So I I think that's probably probably sound advice. You're listening to the campaign trend podcast. I'm speaking with BJ Martino from the Tarrence Group about the latest survey from the AAPC Foundation about voters and AI. So uh BJ, did you in these questions about acceptable, unacceptable, did you get a sense of how or why voters were drawing the line on certain things or or their intensity on these reactions? Like, you know, we we saw a really high level of people say, yeah, you can't use this to create videos. Is that like a strongly uh strong opinion loosely held? What what kind of data might we have on that?

B.J. Martino 13:44

I think we saw the greatest intensity when it came to content that was trying to create a reality that might be false or or misleading. And it's interesting we did a split sample asking voters, are you more concerned about yourself being fooled or about others being fooled? And of course, there were about half of voters who honestly said, I am concerned personally about being fooled by some of these, but it was almost universal. I'm concerned about everyone else being fooled. Not me. Yes. Not me. Uh, but there, you know, there were still some who said they were concerned themselves. And I think as you get closer to that, uh trying to reflect reality back to them, whether it be in video or audio content, that's where those concerns about uh AI-generated content in their faces becomes higher. What is less concerning is back-end sort of tools. There was far less pushback on the notion of using AI to help write some text for fundraising emails or or for outreach like that. And in many cases, uh many voters would not even know text is something that is uh impossible. Uh, that besides there are some AI tools that can claim they can detect it. Uh, but with an with a human and an AI agent working together to write some fundraising texts, you could reasonably deliver some messages to voters that uh aren't going to be seen as false or or misleading. They're just they're just words. Uh, but it's when we get beyond that and it's more voter-facing, uh, it's more about creating things that didn't exist previously. Even a campaign itself creating audio of their own candidate uh to put it in ad, that was broadly unacceptable. And then certainly when we got to the point of putting words or video images of your opponent that didn't exist previously uh using AI, that was effectively universal, saying that's not acceptable, with or without notice saying that this is AI generated.

Eric Wilson 15:50

So I want to get to the second finding that surprised me. And and in my notes at our conference where you were presenting this, there were some expletives next to this uh slide. Uh 11% of voters said social media was their most credible source of election information, but it was just 1% for AI. And and based on my own experience, again, in of one, uh using AI and using social media, if if this were like the prices right, I would have swapped those numbers. So what gives, BJ?

B.J. Martino 16:22

Well, this was a question where we provided a list of potential sources and asked them to pick the one that they said they thought was was most credible. And yeah, AI got 1%, rounding up to 1% actually in the survey. Uh, but social media at 11%. But this is a technology that voters have been around for some time now. They have lived with it for some time. And even though at 11% isn't a big number given everything else uh that was tested in the question, uh, there are some voters who now say, I know enough about social media, I've been around it long enough, I can detect what's real out. I am personally only following the things that I know are telling me what's true and what's real. And so you get to 11%. You know, we said, I think I said in the during the uh conference that AI use isn't necessarily in its infancy, but it's still a toddler for many of these voters, uh, right. And so they're not just walking in the door trusting what they have. There's no, we're starting off in the baseline of very low level of trust. Uh and when you compare, ask them to say, okay, well, if a a national news website or social media or AI tells you something and they conflict who wins this conflict, only 4% agree that AI won the conflict between these different differing sources. So that is a big hill that AI yet has yet to hurdle. And many of those who tell us that they're not using these tools for political or elections information are telling us their most important reason why is because I don't know if I can trust this yet. So it's just, in my opinion, something that's a matter of time as this continues to mature and use continues to grow.

Eric Wilson 18:17

Right. Yeah, I'm I'm reminded of conversations that were happening 20 years ago when social media was first coming onto the scene of, well, you know, it could be anybody, you know. Uh I why how do it how should I believe that they're a doctor or a polster or a member of Congress, right? There was a lot of skepticism there. Um, it was also, and you touched on this earlier, BJ, but it was interesting to see how the usage increased when specific tools were named. And I think, you know, more people are being exposed to this, whether it's in their workplace or you know, on social media. Um, is this kind of the usual pattern of like a voter saying, I hate Congress, but I like my congressman? Um, to the extent that that still exists anymore.

B.J. Martino 18:58

It's it's more, I think, the what our discussion earlier about do voters really realize this is AI? So we asked them these things, ChatGPT, Claude, and then we said, You're Alexa or Siri. Right. Uh, and then we got some uh additional respondents who said, Well, yes, I use those things. Uh, when they maybe previously said, I don't say I don't I don't use AI. And then they said, Oh, okay. I is a Google, I did a Google search and I read the summary. Yes. Oh, okay, then I guess I'm I'm using AI. So as you get specific, and again, it's very much how you ask the question, which is why, again, we asked so many questions that seem to get at the same thing multiple ways here to understand how voters uh are really thinking about it, if they're thinking about it, uh, that it's not that they and they have tools that they they may like and they may trust, uh, but they may not even realize, again, to the point that they don't realize necessarily immediately that this is AI that they're using. Yeah.

Eric Wilson 20:01

So we we certainly hope that we get to repeat this survey uh maybe post-election or in 2028. Where do you expect to see the the biggest shifts in in opinion in some of the data?

B.J. Martino 20:13

Well, I definitely expect to see uh usage continue to grow. I mean, there's no doubt that uh we've seen in polling over time that those numbers have gone up from month to month and from year to year. Um, I do expect the voters are going to become more savvy about AI generated content so that that that ambient crowd, who maybe doesn't realize that what they're getting uh back from their Google search is AI generated content, is going to become more aware of that over time. Uh, I do think it's going to become more acceptable to use AI tools in campaigns from the voters' view perspective as their own lives. They use more of these tools themselves in their own workplaces, in their own schools and homes, uh, that they're going to understand what's acceptable or not. I still think they're going to draw a line in many places, uh, especially when it comes to presenting what might be false, misleading information or creating a caricature of your one's opponent using AI tools, uh, I still think voters are going to push back against that. And I think there's going to be a real uh desire for authenticity, continued desire for authenticity from our candidates, which is something we, you know, as as campaigns should make sure we pay attention to. Um, and hopefully maybe the trust gap, you know, and I think that's remains to be seen. Will voters build that level of trust with what they're seeing? And yeah, it's nowhere to go but up. There's very little room down. Yes.

Eric Wilson 21:50

Uh well, my thanks to VJ Martino for a great conversation. I'm going to link to the survey um outputs, the the presentation, all of that in our show notes if you want to go look at uh additional numbers, and uh we'll include links so you can learn more about BJ and the Tarrence group. If this episode made you a little bit smarter or gave you something to think about, you know all we ask is that you share it with a friend or colleague. You look smart in the process, more people learn about the show. It's a win-win all around, and we like those around here this time of year. Uh remember to subscribe to the Campaign Trend Podcast wherever you listen or watch podcasts so you never miss an episode. And you can visit our website, campaigntrend.com, for more, including our newsletters that go out in tweet. With that, I'll say thanks for listening. We'll see you next time. The Campaign Trend Podcast is produced by Advocacy Content Kitchen, a media production company.