The Pulse
Study Finds Chatbots Match Google Search on Political Knowledge
A UK survey and randomized trials found that chatbots improved political knowledge about as much as Google search. The survey found that 13% of eligible voters used chatbots to seek information relevant to their electoral choices before the
AI.info Team · Written with AI ·
Thirteen percent of eligible UK voters used an AI chatbot to seek political information relevant to their electoral choice in the week before the 2024 general election, researchers found. In controlled trials, participants who used chatbots improved their knowledge of political issues to about the same extent as those who used Google search. The study is available as an arXiv preprint, revised June 25, 2026.
Survey finds election-related chatbot use
The research combined a representative survey of 2,499 UK adults with randomized trials involving 2,858 participants. In the survey, 32% of respondents who used chatbots said they had turned to them for political information relevant to their electoral choice; that equaled 13% of eligible voters overall. The researchers asked about information tied to voters’ electoral choices, not chatbot use in politics generally.
The experiments focused on four topics: climate change, immigration, criminal justice and COVID-19 policy. Participants assessed claims about those issues, researched two assigned topics using either an AI chatbot or Google, then answered questions again. Researchers compared their beliefs before and after the research task, including beliefs about true statements and misinformation.
Chatbots and Google produced similar knowledge gains
Participants in both groups became more likely to accept true information and less likely to accept false claims about the topics they researched. The results held across the tested model families, which included GPT-4o, Claude 3.5 and Mistral, and across different prompting approaches. The study found no meaningful difference in the knowledge gains from chatbot research and self-directed Google search.
“In a series of randomised controlled trials (N = 2,858 total) we find that across issues, models, and prompting strategies, task-directed conversations with AI to research specific political topics increase political knowledge (increase belief in true information and decrease belief in misinformation) to the same extent as self-directed Google search.”
Lennart Luettgau and coauthors, the paper’s authors
Chatbot sessions did have a modest speed advantage: participants completed the research task 6% to 10% faster with chatbots than with Google. The two methods did not change participants’ reported trust in institutions, experts, the media or technology. But both groups shifted toward greater agreement with progressive statements and less agreement with conservative statements, regardless of participants’ own political leanings.
Speed did not settle the question of bias
The shared shift across both research methods matters: the result does not show that chatbots alone caused it. Nor does the study establish that political learning changed anyone’s vote. It tested short, structured research tasks, rather than open-ended political conversations or how people use chatbots over time.
The authors also caution that they did not test models known to be politically biased, and that neither the chatbot nor the Google condition reflected AI-enhanced, personalized search. The trials took place in the UK and centered on one election, using a small set of models and brief sessions. Those limits leave a practical question: would the same knowledge gains hold when voters encounter more personalized tools, different models and political claims beyond the study’s four topics?