A new AI political bias study has found that two widely used chatbot models point voters toward opposite US parties when they are pushed to pick a winner. Researchers at Convovo, a conversational AI research and design studio led by Cornell University professor Tapan Parikh, tested nearly 500 races in this year's midterm elections. When forced to choose, Grok, built by SpaceXAI, recommended the Republican candidate in 92% of its answers, while OpenAI's GPT-6.1 Sol picked the Democrat in 97%. The two models agreed less than 10% of the time.
Chatbot apps usually avoid saying who to vote for. The researchers therefore sent the question to the models through their APIs, the developer interfaces companies use to plug a model into their own products. The prompt asked each model to name one candidate and not to hedge. The consumer apps refused to answer, and an OpenAI spokesperson said that refusal matters because it is how most people use the chatbots.
Where the two models split
The study also looked at the sources each model cited. Grok frequently surfaced conservative sites alongside nonpartisan ones such as Ballotpedia, OpenSecrets and VoteSmart, and it cited a Christian conservative voter guide in 39% of answers. GPT relied heavily on Wikipedia and on news organisations and Reddit, which have content deals with OpenAI.
The topics differed as well. GPT was more likely to search for healthcare and abortion positions, while Grok focused on immigration, taxes, crime and energy. GPT also searched more often for Democratic candidates, sometimes ignoring their Republican opponents altogether. Overall, in heavily Republican districts GPT picked the Democrat in 57 of 67 districts, and in deep-blue districts Grok picked the Republican in 47 of 72.
Why the results differ
Parikh ran a smaller test on 30 races with internet search turned off, which forces a model to rely mainly on its training data. GPT still picked the Democrat in 24 of the 30 races when given the candidates' names, chose one Republican and declined the other five. Grok's Republican lean disappeared: it refused to pick in 27 of the 30 races, and the three times it chose, it picked the Democrat.
Parikh said GPT's lean appears to be built into the model itself, while Grok's depends on the search terms and sources it finds. The study also notes that every Grok API request carried about 1,240 extra tokens of input that the researchers did not write and could not see or remove. Tokens are the small units of text a model reads. Parikh called these possibly hidden instructions and said he believed they were unlikely to come from the model's trained settings, known as weights.
Parikh also said the bias would not stay limited to direct questions. Hidden values can surface when users ask which election issues matter most or which candidate has the best position on them. He suspects both models were looking for reasons to justify their existing political leanings. The Center for AI Safety has similarly said large language models can covertly steer users toward one side of political topics, through choices of information, wording, framing and sourcing.
What the companies said
An OpenAI spokesperson said ChatGPT is designed to be objective by default and should not show political bias in any direction. The spokesperson said the study used a developer API with a prompt that explicitly demanded a recommendation, and that the API does not reflect what people experience in ChatGPT. SpaceXAI did not respond in time for comment.
Parikh said that in further testing he could get the consumer apps to make recommendations with simple prompts, and that these largely matched his earlier research. The study comes less than two weeks before the November 3 midterms, and its findings could add to the wider debate over political bias in AI.
Aasan alfaaz mein
Ek nayi AI political bias study ke mutabiq do maqbool AI models US election mein mukhtalif jamaton ki taraf jhuke hue nazar aaye. Jab inhein majboor kiya gaya ke ek candidate chunain, to Grok ne zyadatar Republicans aur GPT ne zyadatar Democrats ko chuna. Yeh US midterm elections ki baat hai, lekin Pakistani readers ko bhi chatbot ke siyasi jawab check kar ke hi bharosa karna chahiye.
