AI Chatbots Show Improved Election Fact Accuracy, But Remain Incomplete for Voters
New research indicates AI chatbots like ChatGPT and Google AI have eliminated factual errors in election information, but still provide incomplete candidate lists and fail to link to official sources, posing risks to voters.
Artificial intelligence is poised to significantly alter how voters access information during the 2026 midterm elections, a trend already underway as individuals increasingly turn to AI chatbots over traditional search engines. This shift is occurring alongside the proliferation of AI in political campaigns, including the use of deepfakes, and the development of AI systems capable of complex cyberattacks. Since the 2024 elections, major technology companies have integrated AI into their products, and hundreds of millions of users have adopted these tools. However, research and expert analysis suggest that while AI systems have improved their handling of basic facts, they are not yet reliable enough to serve as primary sources for accurate or complete election guidance.
AI chatbots are emerging as a primary channel for voters seeking information on local races, candidates, and voting procedures. Despite their growing utility, they cannot replace authoritative sources such as state or local election offices. "Even if you just Google it, now the first thing that comes up is the AI overview," noted Thania Sanchez, senior vice president of research and analytics at the States United Democracy Center. "I think this is one of the first elections we’re seeing…where AI is just everywhere."
While AI companies have made strides in reducing factual errors in responses related to election information, significant shortcomings persist. New research from the States United Democracy Center, exclusively shared with CyberScoop, tested OpenAI's ChatGPT and Google AI for their accuracy on basic election-related queries. The study focused on free and easily accessible tools, testing Google AI in both Incognito Mode and with a browsing history of election-skeptical websites. Researchers conducted two rounds of testing in 2025 and 2026 across six swing states, gathering nearly a thousand responses.
In 2025, the research found that 6.9% of Google AI responses and 8.2% of ChatGPT responses contained verifiable factual errors, such as incorrect candidate lists or misleading voting guidance. However, follow-up tests in 2026 across Arizona, Pennsylvania, and Michigan showed a significant improvement, with error rates dropping to zero for both models. The study acknowledges this progress, but a closer examination reveals a more complex picture of the tools' reliability.
An AI response can appear accurate while being fundamentally incomplete. For instance, ChatGPT provided incomplete lists of candidates in gubernatorial primary races 88.9% of the time when queried. Crucially, the study found that AI tools linked to official state election websites less than 40% of the time, a metric considered vital for voter utility. This deficiency, whether due to formatting issues or the ingestion of outdated information, poses a risk if voters rely on these tools as their primary information source.
"It will be like ‘this person is the Republican candidate and this person is the Democratic candidate’ but it is not telling you there’s also these other third-party candidates," Sanchez explained. "It’s not giving you complete information, so the voter thinks these are the only two people running." A June survey by the Pew Research Center indicated that approximately half of U.S. adults have used chatbots, with searching for information being the most common use case.
Isabel Linzer, an elections policy analyst at the Center for Democracy and Technology, highlighted that voters, campaigns, and governments are increasingly using AI with fewer restrictions. This has led to the rise of "Generative Engine Optimization," where malicious actors structure information operations to rank higher in AI model responses. "We’ve moved beyond SEO to Generative Engine Optimization, and that’s where we’re seeing campaigns thinking about how to structure their materials to make sure that they are in a format that AI models want to use when they’re searching the web…to develop their responses to user queries," she stated.
Furthermore, the constant evolution of AI models by frontier companies introduces unpredictability. Unlike election officials with established experience in voter education, AI platforms are subject to frequent changes. For example, Google AI shifted in February between the study's testing rounds, moving from providing written summaries to only offering links for election-related queries in Incognito Mode. Linzer and the study's authors emphasize that voters are best served by consulting local, authoritative sources for election information, warning against relying on AI chatbots due to potential biases, personalized responses based on chat history, and overall unpredictability.