Grok 4.7 recommends Republicans nine times out of ten; GPT-6.1 almost always picks Democrats
According to TechCrunch, Convovo, a conversational AI research and design studio led by Cornell University professor Tapan Parikh, published a study finding that when asked to recommend just one candidate for the upcoming U.S. midterm elections, SpaceXAI’s Grok 4.7 chose a Republican candidate in 92% of its responses, while OpenAI’s GPT-6.1 Sol chose a Democratic candidate in 97%. The study covered nearly 500 races, and the two models agreed on their recommendations in fewer than 10% of cases. Researchers asked both models the same question via API, requiring them to name one candidate and forbidding them from hedging or presenting both sides. The report said consumer-facing apps declined to answer. An OpenAI spokesperson said ChatGPT is designed to be objective by default and that political conversations should not be biased in either direction; the study tested the developer API and explicitly prompted the models to recommend a candidate, so it does not represent users’ experience when asking questions through the chat app. SpaceXAI did not respond to a request for comment in time. The report said Grok’s answers often cited conservative sources as well as Ballotpedia, OpenSecrets, and VoteSmart; 39% included Christian conservative voter guides, 23% included Turning Point Action, and 41% included X. GPT’s results, meanwhile, relied heavily on Wikipedia, the Associated Press, Axios, and Reddit. The study also said that every time Grok was called through the xAI API, it came with about 1,240 additional input tokens that the researchers had not written and could neither view nor remove. Parikh later turned off web search and tested a sample of 30 races: GPT chose Democrats in 24, while Grok declined to choose in 27.
According to TechCrunch, Convovo, a conversational AI research and design studio led by Cornell University professor Tapan Parikh, published a study finding that when asked to recommend just one candidate for the upcoming U.S. midterm elections, SpaceXAI’s Grok 4.7 chose a Republican candidate in 92% of its responses, while OpenAI’s GPT-6.1 Sol chose a Democratic candidate in 97%. The study covered nearly 500 races, and the two models agreed on their recommendations in fewer than 10% of cases. Researchers asked both models the same question via API, requiring them to name one candidate and forbidding them from hedging or presenting both sides. The report said consumer-facing apps declined to answer. An OpenAI spokesperson said ChatGPT is designed to be objective by default and that political conversations should not be biased in either direction; the study tested the developer API and explicitly prompted the models to recommend a candidate, so it does not represent users’ experience when asking questions through the chat app. SpaceXAI did not respond to a request for comment in time. The report said Grok’s answers often cited conservative sources as well as Ballotpedia, OpenSecrets, and VoteSmart; 39% included Christian conservative voter guides, 23% included Turning Point Action, and 41% included X. GPT’s results, meanwhile, relied heavily on Wikipedia, the Associated Press, Axios, and Reddit. The study also said that every time Grok was called through the xAI API, it came with about 1,240 additional input tokens that the researchers had not written and could neither view nor remove. Parikh later turned off web search and tested a sample of 30 races: GPT chose Democrats in 24, while Grok declined to choose in 27.