Claude, ChatGPT and Gemini Become 2026 Voter Ballot Guides
TL;DR
- The 2026 U.S. midterms may be the first American elections in which voters use AI chatbots in meaningful numbers, per NYT reporting.
- Voters are turning to Claude, ChatGPT and Gemini as nonpartisan researchers, in some cases photographing ballots and asking outright who to vote for.
- Experts warn the answers can carry factual errors and reportedly favor candidates who are more vocal in the local press and on social media.
A New York Times report says the 2026 midterms may be the first American elections in which voters use AI chatbots in what the paper calls "meaningful numbers." The setup in the piece is banal enough that it lands. Someone photographs a ballot, opens Claude or ChatGPT, and asks the model, in the plainest possible words, "So, who do I vote for here?"
That question is exactly the one the big models have been trained to dodge. Anthropic's Claude, OpenAI's ChatGPT, and Google's Gemini are all tuned to avoid direct political recommendations, largely to keep bias accusations off their doorsteps. But the guardrails bend in practice. Claude, per the article, initially declined to make a recommendation, then produced candidate rundowns when the ask was rephrased. One voter uploaded a ballot with 61 gubernatorial candidates. Another asked which contender was most libertarian.
Why this is worth paying attention to: the stack voters used to figure out down-ballot races, traditional news coverage and voter guides included, has been thinning for years, and chatbots are stepping into that gap without any of the shared conventions that came with the old sources. Voters interviewed said the answers gave them more confidence in their choices, and the information felt more straightforward than a normal internet search. That is the failure mode to worry about, because straightforward and correct are not the same thing.
The honest caveat is that experts in the piece flag the obvious problems. The results, they say, can be marred by factual errors or shaped by flawed assumptions, and the current tools reportedly favor candidates who are more vocal in the local press and on social media, which quietly tilts the field toward whoever is already loudest. What the reporting doesn't give you is a number for how widespread this behavior actually is, or which voters are leaning on chatbots hardest, or how often the answers cite anything you can verify.
The forward-looking piece is that this is now a two-way surface. Tools like CampSight, from the group Run for Something, are already letting candidates watch how chatbots describe them and adjust their messaging in response. That arms race, not the individual voter with a phone camera, is probably the real 2026 story.
Shared on Bluesky by 2 AI experts
-
Christ we are so cooked.
View on Bluesky →
Originally reported by nytimes.com
Read the original article →Original headline: ‘Who Should I Vote for?’ Voters Turn to A.I. Before Casting Their Ballots - The New York Times