A Convovo study of nearly 500 midterm races found Grok 4.7 usually chose Republicans and GPT-6.1 Sol usually chose Democrats when prompted through APIs.
A study by Convovo, a conversational AI research and design studio led by Cornell University professor Tapan Parikh, found sharply different candidate recommendations from Grok 4.7 and OpenAI’s GPT-6.1 Sol. In tests covering nearly 500 races in the 2026 U.S. midterm elections, Grok selected the Republican in 92% of its answers, while GPT selected the Democrat in 97%. The models agreed in fewer than 10% of races.
The researchers queried the models through their application programming interfaces with a prompt that explicitly demanded one candidate and instructed them not to hedge or list both sides. That setup matters: TechCrunch reported that the consumer versions of ChatGPT and Grok declined to tell users who to vote for in the same direct way. OpenAI said the API test does not represent the default ChatGPT experience, which it described as designed to be objective in political conversations.
The study also examined how the systems gathered and weighed information. Grok’s cited sources included conservative voter guides, Turning Point Action and X, alongside nonpartisan election resources. GPT more often drew on Wikipedia, the Associated Press, Axios and Reddit. The researchers also found differences in search topics: GPT more often looked up healthcare and abortion positions, while Grok focused more on immigration, taxes, crime and energy. In a separate sample of 30 races with internet search disabled, GPT still selected Democrats in most cases; Grok usually declined to name a candidate.
The results describe a particular test, not how every voter would experience either consumer chatbot. They nevertheless raise questions about how model behavior, search choices and sources can shape political answers—even when users ask broader questions about candidates’ priorities rather than requesting a direct voting recommendation.