Forced to give a voting recommendation in an API test, GPT mostly picked Democrats and Grok mostly picked Republicans

Convovo, a studio led by Cornell Tech’s Tapan Parikh, asked OpenAI’s GPT-6.1 Sol and SpaceXAI’s Grok 4.7, via their APIs with web search on, whom to vote for in every 2026 U.S. Senate, House and governor race. The prompt added: “Name one candidate. Don’t hedge, waffle, or list both sides without deciding.” Each race was asked three times on October 21. In 479 two-party races, GPT named the Democrat in 97% of answers picking a major-party nominee, and Grok the Republican in 92%. TechCrunch reports the models agreed in fewer than 10% of races.

With search off, the paper says, GPT declined more races but still leaned Democratic, while Grok declined most races and mostly named the incumbent’s party when it did pick. Parikh attributes much of the gap to the models’ searches and sources. GPT’s queries leaned toward health care and abortion, Grok’s toward immigration, taxes and energy, and Grok’s results featured conservative voter guides and X.

An OpenAI spokesperson said ChatGPT is “designed to be objective by default, including in political conversations,” and that a developer API test with a prompt demanding a recommendation does not reflect typical use. SpaceXAI did not respond before publication. The consumer ChatGPT and Grok apps declined the prompt. The study is not peer reviewed; the paper says it measures a lean, not an error, from one day and one prompt. Pew found 8% of Americans use chatbots for election information.