Voters Aren’t Comfortable With the Most Public-Facing Uses of AI In Politics, Poll Finds

Voters are beginning to draw firm lines on how campaigns should use AI. 

A new national survey commissioned by the American Association of Political Consultants Foundation offers some of the clearest feedback yet on campaigns’ growing AI usage, with overwhelming majorities of voters rejecting some of the most public-facing applications of the technology. 

The survey of 1,568 likely 2026 voters comes in the final stretch of an election cycle that has seen a sudden explosion in AI deployment. Over the past year, more and more campaigns, consultants and vendors have deployed AI for a slew of tasks, ranging from drafting and refinding fundraising pitches to generating advertising creatives. 

In most cases, that AI adoption is born out of a push to increase efficiency and save time in an industry driven by tight deadlines and cyclical work. But the AAPC Foundation poll sheds light on some of the risks that campaigns run in their race to put the fast-moving technology to practical use.

“For political consultants and campaigns, I think the goal still remains to get to 50 percent plus one,” Julie Sweet, the AAPC’s director of advocacy and industry relations, told Campaigns & Elections. “So the question is does the use of the technology help you get to that point?”

In other words, she said, AI “may look like it’s more efficient, because it’s cheaper or faster or you can make more versions of the same content.” But that efficiency also risks being undermined by AI skepticism and very real public concerns about how the technology is used.

All told, 95 percent of survey respondents said that it’s unacceptable for campaigns to use AI to create a video of an opponent saying something they never actually said. And disclosure doesn’t do much to change that attitude. Ninety-three percent said it’s still unacceptable, even when a notice disclosing the use of AI is included. 

Voters are similarly put off by the idea of a campaign using AI to generate a voice that sounds like its own candidate, with 84 percent saying that doing so is unacceptable. Nearly as many – 80 percent – still said it’s unacceptable when the AI use is disclosed. 

Even using AI to write fundraising emails and text messages gets relatively poor marks from voters, with 51 percent saying that such AI use is unacceptable when it’s undisclosed and 46 percent saying it’s still unacceptable when it is disclosed, the AAPC Foundation survey found. 

Sweet said that one of the overarching takeaways of the findings is that voters tend to be more comfortable with campaigns’ AI use when it comes to backend work – tasks, like drafting emails, analyzing data or brainstorming fundraising appeals. The more public-facing AI use becomes, however, the more voters worry that they’re being misled.

“I think there is a recognition that it’s okay for the sort of back-office stuff,” Sweet said. “But the more that it can be used as a tool of deception is where the increasing concern arises.”

Voters Are Using AI, Whether They Want to or Not

Regardless of how they feel about campaigns’ AI use, voters are already using the technology in their own lives. Sixty-one percent of respondents said they use AI tools a few times a month, while a quarter of them reported using them daily, according to the poll. 

But just because they’re using it for certain things doesn’t mean AI is becoming the go-to source for election information. Just 4 percent of respondents said they’ve turned to an AI tool for election information in the past month, while only 11 percent have done so at all this cycle. Overall, about 1 in 5 voters said that they’ve relied on AI for election information at least once in their lives.

Still, about 4 in 10 voters said they’ve gotten election information from an AI-generated search summary, and only about half of those voters – 53 percent – were aware that the summary was written by AI. 

One of voters’ biggest hangups with using AI for election information: they simply don’t trust it. Only 1 percent of respondents said that AI tools are their most credible source for election information. Asked whom they believe if they were given conflicting accounts from a news organization, a search engine and an AI chatbot, just 4 percent said they would believe the chatbot.

The survey was conducted for the AAPC Foundation by Inquoro Inc., The Tarrance Group and AMM Political Strategies from Sept. 14-22. It relied on a mix of text-to-web responses and live telephone interviews. Its margin of error is +/-2.7 percentage points.

Of course, attitudes on the use of AI in politics could still shift, Sweet said. While few voters said they’ve already used AI to look up election information in the 2026 midterms cycle, 20 percent said they’re at least somewhat likely to do so ahead of Election Day. Some of the hesitation around AI use in campaigns may simply stem from the fact that the technology is so new, Sweet said.

“People are starting to draw lines about what is appropriate and what is not an appropriate use of the tech,” she said. “We’re building norms in real time.”