Voters searching for guidance on candidates and ballot measures are increasingly turning to AI chatbots, a habit election officials say carries real risks. The warning is blunt: do not use AI chatbots to decide who to vote for.

What You Need to Know

AI chatbots generate answers from patterns in training data rather than from verified election sources. They can state falsehoods confidently, omit local races entirely or reflect the biases of their training material. Election administrators say the only reliable sources for ballot information are official state and county election offices. Voters who rely on a chatbot to decide who to vote for may act on information that is outdated, incomplete or simply wrong.

The Gap Between Fluent Answers and Verified Facts

Chatbots are designed to produce plausible text, not to verify claims. When asked about a candidate or a ballot question, a model may blend accurate details with invented ones and present both in the same confident tone.

That fluency is the problem. A voter who receives a clear, well-written answer has little reason to doubt it, even when the underlying facts are wrong. Election officials say the risk grows in down-ballot races, where public information is thinner and models have less reliable material to draw from.

Where the Errors Show Up

Common failure modes reported by researchers include invented candidate positions, outdated filing information and confusion between similarly named races. Local contests are especially vulnerable because training data often skews toward national politics.

The result is a tool that can sound authoritative about a school board race it knows almost nothing about. Voters, however, have no easy way to tell the difference between a grounded answer and a fabricated one.

Why Official Sources Still Outperform Chatbots

State and county election offices publish sample ballots, candidate statements and measure texts in formats that can be checked against a primary source. That verifiability is what chatbots lack.

Officials recommend a simple sequence for voters who want reliable guidance on how to vote:

  • Check the source: Confirm every candidate and measure against the official state or county election website.
  • Read the sample ballot: Use the exact wording voters will see in the voting booth.
  • Verify local races: Look for official candidate statements rather than summaries from any AI tool.
  • Treat chatbot answers as leads: Use them only as a starting point for verification, never as a final answer.

The same discipline applies to any AI-generated summary of a ballot measure. If the text cannot be traced to an official document, it should not shape a vote.

The Broader Push to Regulate Political AI

The warning lands amid a wider effort to govern how AI systems handle election content. Lawmakers and regulators have pressed model developers to disclose how their systems respond to political queries and to limit the spread of fabricated election material.

Those efforts remain uneven. Rules vary by jurisdiction, and enforcement against general-purpose chatbots is still limited. Election officials, meanwhile, are left to manage voter confusion with public messaging rather than technical controls.

The practical burden falls on voters. A chatbot that answers a question about who to vote for is performing a task it was never built to do reliably, and the consequences play out at the ballot box.

Why This Matters

The stakes extend beyond a single election cycle. If voters come to trust chatbot answers on candidates, the damage compounds: misinformation spreads faster, trust in official election information erodes and the cost of correcting false claims rises with every cycle.

Election administrators face a new communications burden, forced to compete with tools that answer instantly and sound certain. Voters who decide who to vote for based on an unverified chatbot response may never learn the answer was wrong.

Model developers, for their part, face growing pressure to add guardrails around political queries. Whether those guardrails arrive before the next major election remains an open question, and the answer will shape how millions of people decide who to vote for.