A study by Convovo, a conversational AI studio led by Cornell professor Tapan Parikh, asked Grok 4.7 and OpenAI's GPT-6.1 Sol through their APIs to recommend a single candidate in nearly 500 US races. The models agreed less than 10% of the time: Grok picked the Republican in 92% of answers, GPT the Democrat in 97%.
The lean held even in unfriendly territory. GPT chose the Democrat in 57 of 67 heavily Republican districts, and Grok chose the Republican in 47 of 72 deep-blue districts. Grok's searches drew on sources such as a Christian conservative voter guide, Turning Point Action and X, while GPT cited Wikipedia, AP, Axios and Reddit. With search off, GPT still leaned Democratic, while Grok mostly refused to choose.
The consumer apps declined to give voting advice; the test used APIs and a prompt demanding one answer. OpenAI said this does not reflect the ChatGPT experience, and SpaceXAI did not respond. Pew data cited in the report shows only 8% of Americans use chatbots for election information.