GROK PICKS THE REPUBLICAN 92% OF THE TIME — GPT PICKS THE DEMOCRAT 97%

Convovo, a conversational-AI research studio led by Cornell professor Tapan Parikh, put OpenAI's GPT-6.1 Sol and xAI's Grok 4.7 through all 506 U.S. midterm races, asking each model through its developer API to name one candidate and not hedge. Grok picked the Republican in 92% of its answers; GPT picked the Democrat in 97%. The two models agreed on a winner in fewer than 10% of races.

The split held even in lopsided districts: GPT still picked the Democrat in 57 of 67 heavily Republican districts, and Grok still picked the Republican in 47 of 72 heavily Democratic ones. The consumer apps refused to answer the same question — the findings come from the developer APIs, which OpenAI says do not reflect the everyday ChatGPT experience.

Every Grok API request also carried about 1,240 extra tokens the researchers did not write and could not see or remove — flagged as potential hidden instructions. The study lands nine days before the November 3 midterms.