How many voters will actually turn to AI for political information this fall? And how do chatbots stack up against other sources of political information voters might turn to? To explore this, we collaborated with Change Research to ask two questions on a national poll of registered voters. We found that a small but potentially decisive share (15%) of voters plan to seek political information from AI chatbots–with young voters and voters of color the most likely to do so–and a much larger swath of the electorate is likely to encounter AI election information via Google Search and AI search overviews.
The findings below reflect Change Research’s survey of 1,892 respondents weighted to a registered voter population. The survey was fielded from May 5-10, 2026. Respondents were recruited using a combination of targeted digital ads and text-to-web SMS, and all respondents took the survey online.
AI as a source of political information
Asked which information sources they were likely to use to learn about candidates running for office in 2026, most Americans say they will rely on news articles (68% very or somewhat likely to use), discussions with friends, family, or colleagues (62%), candidate websites (57%), and social media (53%).
A much smaller percentage say they plan to use an AI chatbot – just 15% of respondents said they were very or somewhat likely to use an AI chatbot to learn about candidates and a majority (58%) actually reported that they were very unlikely to do so. However, 62% said they would rely on Google Search to learn about candidates. Many of these searchers may be exposed to AI-generated political information via the AI summaries Google displays on top of search results, although it is unclear what share of political Google Search queries will have AI overviews.
These results highlight two key points: 1) although only a small slice of the electorate will actively seek out AI-generated information about this year’s elections, even a small segment of active AI information-seekers is large enough to be decisive in close elections and 2) a huge proportion of the electorate may encounter AI-generated political information through passive consumption.
Who is most likely to turn to a chatbot?
Voters under age 35 are nearly three times more likely to say they will actively seek information from AI chatbots (21%) than voters over age 65 (8%). This is still far lower than the share of voters under 35 who say they will rely on social media (69%), but means that one fifth of young voters plan to rely on chatbots as an information source as they make voting decisions in November.
Voters of color are much more likely to say they will use chatbots for political information compared to white respondents (21% vs. 12%) and men are slightly more likely than women to do so (17% vs. 13%).
Broken out by partisanship, there are modest partisan differences on intent to use AI, but larger differences among those who say they are very unlikely to use AI – 60% of Democrats, 54% of Republicans, and 68% of Independents say they are very unlikely to use AI to learn about candidates this year.
Respondents who were undecided on the generic Congressional ballot were less likely to say they would use a chatbot (9% very or somewhat likely) than those voting for Democrats (15%) or Republicans (17%). These undecided respondents tend to be a hard-to-reach population and they were significantly less likely to say they would seek out information on all of the channels asked about above than voters already supporting Democrats and Republicans.
How much do voters trust AI-generated information about politics?
Next, we wanted to assess whether voters view AI-generated political information as more or less accurate than information from candidates themselves. We know that Americans are generally pessimistic about the accuracy of AI and AI’s impact on society. Americans are also pessimistic about politics and have low trust in elected officials. Between candidate- and AI-provided information, which source wins out? Candidate information wins out as more trustworthy (36% to 14%).
When asked, “When it comes to information about political candidates, which do you find more accurate, answers from AI tools (e.g., ChatGPT, AI search summaries) or information from the candidates themselves?”, more than a third of respondents (36%) said that neither are accurate. 14% said that AI-generated information is somewhat or much more accurate, while 36% said candidate-generated information is more accurate. 14% said AI- and candidate-generated information are equally accurate.
By subgroup, trust mirrors likelihood to use AI, with 31% of respondents under 35 saying AI-generated information is equally accurate to or more accurate than candidate information. 37% of voters of color (including nearly half of Black voters) said that AI is equally or more accurate.
An outright majority of Independents (56%) responded that neither source was accurate.
Implications for 2026 and beyond
A significant minority of voters (15%) will actively seek out information from AI chatbots when they go to the polls this Fall. Even more may be exposed to AI-generated election information via search. These voters are disproportionately young and Black, Hispanic, and Asian – some of the exact demographics that have been the among the most persuadable in recent elections. And while most voters distrust AI, the same groups using it are more likely to trust it than others. In close races, AI-generated information could be decisive.
AI is also here to stay. Americans increasingly use generative AI across their work and personal lives, with the pace of adoption tracking ahead of previous transformative technologies such as the personal computer and the internet. In 1996, just 4% of voters sought political information online – rising to 18% in 2000 and 29% in 2004. More than 20 years later, many Americans live their lives online. This fall, more than half of voters say they plan to seek out political information via Google Search and candidate websites (not to mention news articles that are largely online).
We are in the very early stages of observing what impact AI-generated information will have on American politics (and wider society). Understanding what chatbots tell voters is already critical and as the share of voters consulting AI sources grows, it will only become more important.
Thank you for this important research!