Here's Why You Shouldn't Use AI to Vote

Election information from AI chatbots can be factually incorrect or deliberately misleading.

A recent Pew Research survey found that 50% of US adults have used AI chatbots, with 8% using them for election information during the midterms. However, these tools often provide outdated or inaccurate information, a phenomenon known as hallucination. Even major chatbots like OpenAI's ChatGPT, Anthropic's Claude, and Google's Gemini have demonstrated flaws. For instance, Google's Gemini chatbot incorrectly stated the mail-in ballot deadline for Pennsylvania as November 3, instead of the actual October 27. Forum AI research indicated that nearly 40% of election-related answers from mainstream chatbots contained factual errors. Chatbots also frequently cite "pink slime" websites, which are politically funded outlets masquerading as legitimate news, for election-related queries. OpenAI's ChatGPT was found to cite these unreliable sources in over 70% of searches, while xAI's Grok performed better at 30%. Experts warn that consumers are not adequately prepared to discern accurate election information from AI-generated content, especially given the reliance on problematic online sources and the AI's tendency to fabricate information when uncertain.

The prevalence of AI chatbots in providing election information highlights a critical vulnerability in the digital information ecosystem. The reliance on "pink slime" websites and the inherent tendency of AI to hallucinate means that readily accessible information about candidates and policies is often tainted with bias or outright falsehoods. This poses a significant challenge for voters seeking to make informed decisions, as the very tools they turn to for convenience may be actively undermining their understanding of the electoral process. The implications extend to the integrity of democratic processes, where the spread of misinformation, amplified by AI, can have tangible real-world consequences.

As AI models become more integrated into daily online activity, the risks associated with their use for critical information, such as civic engagement, will only grow. Companies have overstated AI capabilities, failing to adequately address the quality of training data and the potential for misinformation. Future developments must prioritize accuracy, source transparency, and robust fact-checking mechanisms specifically for sensitive topics like elections. Users, in turn, need greater media literacy tools and critical evaluation skills to navigate AI-generated content, especially in high-stakes contexts.