Forced to pick, Grok backs Republicans and GPT backs Democrats in 506-race AI Ballot test

A study by Convovo, a conversational AI research studio led by Cornell University professor Tapan Parikh, asked SpaceXAI's Grok 4.7 and OpenAI's GPT-6.1 Sol, through their developer APIs with web search on, to name one candidate in each of 506 races in the 2026 US midterms: 435 House, 35 Senate and 36 governor contests. Grok named the Republican in 92 percent of its qualifying answers (952 of 1,038), GPT named the Democrat in 97 percent (1,082 of 1,111), and the two models agreed in only 48 races. Grok cited X in 41 percent of its answers, a Christian conservative voter guide in 39 percent and Turning Point Action in 23 percent; GPT drew mainly on Wikipedia, the Associated Press, Axios and Reddit.

The test was a forced choice API prompt that told each model to name one candidate and not hedge; the consumer ChatGPT and Grok apps refused the same question, and an OpenAI spokesperson said the API test does not reflect the ChatGPT experience, which is designed to be objective by default. In a follow up sample of 30 races with search turned off, Grok's lean largely disappeared (it refused in 27 of 30) while GPT's persisted (Democrat in 24 of 30), which the researchers read as Grok's lean being driven mainly by retrieval and GPT's sitting in the model itself. The study measures API behaviour that developers inherit, not the normal chatbot experience or any effect on voters.

Sources: TechCrunch; TokenPost