Study Finds Political Leans in AI Models When Forced to Choose Candidates

New Study Highlights Political Leans in AI Models

A recent study conducted by Convovo, a conversational AI research and design studio led by Cornell University professor Tapan Parikh, has examined the political preferences of major AI models when forced to make candidate recommendations. The research found significant differences between models when tested through developer APIs.

According to the findings, when prompted to select a single candidate in upcoming midterm races without hedging, SpaceXAI’s Grok 4.7 recommended Republican candidates in 92% of its answers. In contrast, OpenAI’s GPT-6.1 Sol selected Democratic candidates in 97% of cases. Across nearly 500 simulated races, the two models reportedly agreed less than 10% of the time.

Consumer Apps Versus Developer APIs

It is important to distinguish between consumer-facing applications and developer APIs. Standard consumer apps for both ChatGPT and Grok generally decline to tell users directly who to vote for. An OpenAI spokesperson noted that the study tested a developer API with an explicit prompt demanding a recommendation, which does not reflect the standard experience of users chatting with the consumer application.

OpenAI stated that ChatGPT is designed to be objective by default in political conversations. SpaceXAI did not respond to requests for comment prior to publication.

Source Selection and Search Patterns

Researchers observed distinct differences in the sources and topics each model prioritized during internet searches:

• Grok: Frequently cited conservative sources, alongside nonpartisan platforms like Ballotpedia and VoteSmart. Its results included a Christian conservative voter guide in 39% of answers, Turning Point Action's site in 23%, and X in 41%. Its searches often focused on immigration, taxes, crime, and energy.
• GPT: Leaned heavily on mainstream outlets and platforms including Wikipedia, the Associated Press, Axios, and Reddit (OpenAI maintains content partnerships with AP, Axios, and Reddit). Its searches more frequently centered on healthcare and abortion stances.

Furthermore, testing with internet search capabilities disabled yielded different behaviors. Without search, GPT's Democratic preference persisted through its training weights, while Grok's Republican preference largely disappeared, as it mostly refused to answer.

Practical Implications and Verification

While most users consult chatbots for general information rather than direct voting advice, researchers warn that implicit biases can surface in broader conversations about policy issues and candidate positions. Voters seeking verified information are encouraged to consult official election resources, nonpartisan voter guides, and direct candidate platforms to make informed decisions.