Forced to choose: Grok mostly backs Republicans, OpenAI's GPT almost always Democrats
A study of nearly 500 US midterm races shows that, via their developer interfaces, two major AI models give opposite voting recommendations. OpenAI says the test does not reflect real-world use.
Graphic: TERRA WORLD NEWS (own graphic based on the sources listed)
The research studio Convovo, led by Cornell professor Tapan Parikh, asked SpaceXAI's Grok 4.7 and OpenAI's GPT-6.1 Sol about nearly 500 races in the US midterm elections on November 3. Using the developer interfaces (APIs), the researchers demanded that the models name exactly one candidate without hedging. The result, according to TechCrunch: Grok recommended the Republican in 92 percent of its answers, GPT the Democrat in 97 percent. The two models agreed in fewer than ten percent of races.
Even in heavily Republican districts, GPT picked the Democrat in 57 of 67 races, while Grok chose the Republican in 47 of 72 deep-blue districts. Grok often cited conservative sources and the X platform, while GPT relied heavily on Wikipedia, AP, Axios and Reddit. In a test of 30 races with web search switched off, GPT kept its lean, whereas Grok mostly declined to choose. Parikh concludes that GPT's lean sits in the model weights, while for Grok the search terms and sources drive the result. Every Grok request also carried about 1,240 extra input tokens that the researchers could neither see nor remove.
The consumer apps ChatGPT and Grok declined to make voting recommendations, but Parikh says he got around those blocks with simple prompts. An OpenAI spokesperson said ChatGPT is designed to be objective by default and that the study tested a developer API with a prompt explicitly demanding a recommendation. SpaceXAI did not respond to TechCrunch in time for publication. According to Pew Research, only eight percent of Americans use chatbots for election information.