Convovo study says Grok chose Republicans in 92% and GPT Democrats in 97%
🎧 Listen to this story
A Convovo study, cited by TechCrunch, found that SpaceXAI’s Grok 4.7 recommended the Republican candidate in 92% of answers, while OpenAI’s GPT-6.1 Sol chose the Democrat in 97% in the upcoming midterm elections. Across nearly 500 races, the models agreed in less than 10% of cases. The researchers queried the models through their APIs and required them to choose one candidate without hedging or presenting both sides.
The consumer-facing apps refused to answer, a distinction OpenAI said was important. The company said ChatGPT is designed to be objective and that the study does not represent the user experience in the app. The analysis also found differences in the sources and topics searched: GPT looked more often for information about healthcare and abortion, while Grok focused on immigration, taxes, crime and energy. SpaceXAI did not respond to a request for comment.
Why it matters · editorial interpretation
The difference between API responses and the apps’ refusals suggests that the result depends on the usage context and how the models were instructed. That methodological distinction is essential when interpreting studies about models’ political preferences, especially because OpenAI says ChatGPT is designed to be objective.
Sources
Research