Artificial intelligence has become a fixture in how Americans seek political information, from asking chatbots about ballot propositions to relying on AI-generated summaries in search results. But the rapid integration of these tools into the democratic process is prompting new concerns about their reliability and the potential for misinformation.
Recent studies indicate that about half of U.S. adults now use chatbots, a significant jump from a third in 2024. A report by the left-leaning marketing firm Orchestra found that roughly 650,000 daily prompts relate to politics, with OpenAI's ChatGPT and Google's AI tools being the most common sources. Ben Worthen, Orchestra's chief product officer, noted that the surge in AI usage for general questions has been "staggering," and that even standard Google searches now often yield AI-generated overviews.
Despite improvements from earlier models, researchers warn that AI systems still produce troubling errors. The Institute for Strategic Dialogue (ISD) tested six major chatbots with election-related questions and found that 29% of responses were incomplete, inaccurate, or outdated. In 16% of cases, the chatbots provided correct but insufficient information, omitting crucial details like deadlines or ID requirements. Accuracy dropped by 16 percentage points when queries were made in Spanish.
Max Read, ISD's director of civic innovation, emphasized that while skepticism about AI is warranted, these tools are becoming an unavoidable part of the information ecosystem. "The reality is that outputs are going to be playing an increasing role," he said. He urged election officials and campaigns to ensure their online information is up to date, as chatbots often rely on outdated or incorrect sources.
The use of AI for political information is not just a novelty; it's shaping real decisions. In California, startup CEO Amrita Bhasin uses AI to decode complex ballot propositions, asking the models to explain bills in simple terms or compare candidates' policies. She appreciates the neutrality, saying that newspaper guides can be "biased in a certain way." However, she emphasizes that AI is a tool for gathering information, not making the decision for her.
An analysis by Anthropic's Economic Index found that political topics ranked 66th out of 189 in AI conversations, but usage spikes around elections. In states with May primaries, political queries jumped 20%. Andy Hall of Anthropic and Dan Thompson of UCLA predict that the November midterms will bring the "biggest wave of political AI use on record."
The growing reliance on AI comes as lawmakers and regulators grapple with its implications. Some have called for greater oversight, while others, including former President Trump, have proposed renaming AI to "super intelligence"—a move that critics say distracts from real policy debates. Meanwhile, voters are increasingly turning to AI for help with everything from filling out ballots to understanding complex issues, raising the stakes for ensuring these systems are accurate and unbiased.
As the midterm elections approach, the intersection of AI and democracy is becoming a critical issue. With more voters relying on these tools, the pressure is on for tech companies and election officials to address the flaws that could undermine trust in the electoral process.
