Nearly a third of AI chatbot replies to voting questions inaccurate, outdated: Study

The ISD study arrives at a pivotal moment: as tens of millions of Americans prepare to vote, AI chatbots have quietly become the new roadside attraction for fast answers. The finding that roughly 30% of replies were incomplete or just plain wrong tells us something deeper than a tech bug. It shows that the encyclopedic confidence of these systems is simply not a substitute for the fact-checking, editorial judgment, and local knowledge that voters need. There are two sides to this issue, as Poli Split sees. On the pro-AI side, chatbots already offer 24/7 access that overworked election clerks and belt-tightened newsrooms cannot always match. For recently naturalized voters facing language barriers, or shift workers too tired to visit a county clerk's website, an AI query may be their honest attempt to participate. And developers are improving fast—the study itself notes that prompt-based responses are a moving target. The other side is about the stakes. Even a 30% inaccuracy rate is far too high when the question asked is "Where do I vote?" or "What ID is required?" Outdated information can disenfranchise millions, and errors don't degrade evenly—they hurt groups whose turnout is most affected by confusion. The civic participation cost is not a math glitch, it is a representation gap. And the study notes only six bots tested with just 15 prompts, but those 90 interactions represent a snapshot of a far larger problem. What strikes me about the reporting of The Hill's article is that the described issue doesn't fit neatly into the right-left divide. Republicans worried about platform censorship may see some AI responses as the same persistent bias. Democrats worried about misinformation and suppression will see another in a long line. But more importantly, this is a reckoning for us as a society: we are increasingly living inside systems whose algorithms are opaque, and that handle our most foundational right—the right to self-govern. A practical note of responsibility for voters: the study's findings (https://thehill.com/policy/technology/6071757-ai-chatbots-election-info-inaccuracy/) do not call for a consumer migration away from AI, but they do exactly the opposite—they call for a dose of skepticism. Verify information from our nonpartisan, local election official. Check that URL twice. Press the script that says "you cannot sell the right to vote." In a world where a 30% inaccuracy rate isn't only a deal-breaker, it's a red flag. The time to audit your intelligence is now, before the next poll.