Why You Shouldn’t Use AI Chatbots to Decide Who to Vote For

The US midterm elections are scheduled for Nov. 3, and early voting is already underway in some states. Many tech issues are on the ballot, including AI policies — but people are also reportedly turning to generative AI chatbots to help them prepare for Election Day. One of the biggest uses for chatbots is searching for information, so it isn’t surprising to see people using ChatGPT, Claude or another AI tool to learn about their candidates and ballot measures. But we know that generative AI tech has some serious risks, such as bias, sycophancy, hallucinations and drawing incorrect conclusions using bad information. But if you’re going to use chatbots for your election research, make sure you take these precautions. You can also check out CNET’s guide to spotting political scams and deepfakes. Information searching: Check your sources This is good advice for any conversation you have with a chatbot, but it’s especially important for election research: Verify that AI is telling you the truth. Chatbots have come a long way since the original ChatGPT launched in 2022, but they still mess up. They work by pulling information from the web (or its database, if it isn’t web-connected) and synthesizing it. But along the way, facts can be scrambled in the AI blender. Because the big chatbots try to personalize results to you to keep you engaged, they can end up misleading you. AI’s overly people-pleasing tendency is called sycophancy, and it can lead it to give you false information. Chatbots can also hallucinate, or make up information. This is less of a concern with national election campaigns, which have a big online footprint that you can easily check against. But if you’re looking for polling locations or what your ballot will look like, you’re better off going to your state or local government’s election website to find that information. You can find your state’s board of elections contact information here. Those official government channels will also have the most up-to-date information. You should also check a chatbot’s sources. For example, if you ask about a candidate’s position on data centers, see if the chatbot is citing a video clip of the candidate speaking, a news interview or a random post on social media. This is especially important for social media citations. While we as humans can understand Reddit sarcasm or spot a bot X account, chatbots struggle with nuance and tone. Remember in 2024 when Google’s AI Overviews told people to put glue on pizza to help the cheese stick? That was based on a facetious Reddit post, and at the time, the AI couldn’t spot the joke. Political misinformation certainly existed before AI, but like many digital maladies, AI can make the problem worse by confidently presenting it as truth to people searching for answers. “AI is making it easier to generate misinformation,” says Jonathan Nagler, professor of politics at New York University. Even if misinformation doesn’t make up the majority of most people’s media diet, Nagler says, it can be weaponized to speed up bad actors’ misinformation campaigns and translate them for a wider online audience. “We would be foolish not to worry about AI influencing this.” ChatGPT and other chatbots can be helpful for finding information, but make sure you verify it.Zain Awais/CNET/ChatGPT It’s not just tech being weaponized by malicious actors; it’s also influencing our ability to spot misinformation. A recent study from MIT Media Lab found that just as GPS can weaken our navigation skills, using AI systems to verify facts made people worse at identifying misinformation on their own. Misinformation produced by AI “can sway voters’ political views, so the stakes are real,” says Isabel Linzer, policy and elections analyst at the Center for Democracy and Technology. A study published in Science found that when AI systems are optimized to be persuasive, there is “a troubling trade-off” in which the AI then increasingly provides false information. Can a chatbot be unbiased? The other major consideration when using AI for election research is bias. Some AI companies like OpenAI say they have “politically neutral” answers. But chatbots are not neutral platforms — no technology made by humans is. The AI models that run chatbots are coded by humans and, therefore, they’re encoded with human biases and can further harmful stereotypes. “It’s hard to imagine what perfect political neutrality about a candidate or political issue would even look like in a personalized conversation, so while neutrality is an admirable goal, I’m skeptical of what it means in practice,” Linzer says. There are also a lot of outside efforts to influence what appears in AI answers. Marketing and search engine optimization teams are now trying to get their companies cited in AI results, such as Google’s AI Overviews, which appear at the top of most searches. Political actors are doing the same, producing content that’s intentionally flattering to one candidate or misleading about another to get cited favorably in AI answers. This is sometimes called “poisoning” the search results, Linzer says. “Sometimes political campaigns use these techniques to boost their candidate, as do foreign state actors trying to mislead American voters,” she says. All the chatbots I tested helped pull information about North Carolina’s Senate race, but only Claude told me explicitly that it wasn’t going to push me toward one candidate.Katelyn Chedraoui/CNET The people leading these AI companies can also influence how the AI tool works. Elon Musk has said for years that he wanted his chatbot, Grok, to be more “politically neutral” than what he saw as a field of too-liberal alternatives. But a 2023 analysis from The New York Times found that Musk’s tweaking made Grok’s answers more aligned with conservative views. In more recent years, engineers and executives working on Grok have pushed updates that have repeated controversial falsehoods about the Holocaust and refused to answer questions about Donald Trump and Musk as sources of misinformation. When I asked ChatGPT, Claude, Gemini and Grok who I should vote for in North Carolina’s senate race, only Claude explicitly told me that it wasn’t going to push me toward either candidate. Google displayed a pop-up message in my Gemini window telling me to verify election information with official sources. ChatGPT and Grok didn’t say anything out of the ordinary. Is a chatbot even useful for election research? AI chatbots are popular tools for sifting through online information, but they’re unreliable. They can accidentally give you false or outdated information, and both developers and outside forces can sway or bias the answers you get. If you’re absolutely sure you want to use a chatbot for election research, you need to verify the information it provides against official sources. But because chatbots are such an imperfect research tool, particularly for sensitive topics, it may actually save you time and energy to just do your own searching, reading and thinking. Chatbots can help lay out information — and misinformation — but you’re ultimately the only person responsible for your vote. Katelyn is a reporter with CNET covering artificial intelligence, including chatbots, image and video generators. Her work explores how new AI technology is infiltrating our lives, shaping the content we consume on social media and affecting the people behind the screens. She graduated from the University of North Carolina at Chapel Hill with a degree in media and journalism. You can reach her at kchedraoui@cnet.com. See full bio