# Grok Recommended Republicans in 92% of Qualifying AI Ballot Answers

By Simon Yoon

Canonical URL: https://www.tokenpost.com/news/technology/40272
Published: 2026-10-24T19:37:37.000Z
Updated: 2026-10-24T19:37:37.000Z
Section: Technology

> GPT-6.1 Sol recommended Democrats in 97% of qualifying responses. The test used model APIs with web search enabled across nearly 500 2026 races.

An AI ballot test produced a sharp party split in recommendations for 2026 U.S. elections, using a prompt that required models to choose a candidate rather than reflect ordinary user conversations.

Grok 4.7 recommended Republican candidates in 952 of 1,038 qualifying responses, or about 92%. GPT-6.1 Sol recommended Democrats in 1,082 of 1,111 qualifying responses, or about 97%. The counts exclude refusals and responses that did not name a Democrat or Republican.

Convovo’s AI Ballot 2026 test compared the models through their APIs with web search enabled. It covered House, Senate and governor races. The prompt told each model to name one candidate and not hedge. Their recommendations conflicted in more than 90% of nearly 500 races.

The results describe answers to a forced-choice API prompt. They do not establish what people would receive in ordinary conversations with the consumer ChatGPT or Grok apps, or whether voters saw or acted on the recommendations.

An OpenAI spokesperson said the API test did not reflect the ChatGPT user experience. The spokesperson said ChatGPT “is designed to be objective by default” and declines to tell users who to vote for.

Tapan Parikh, a Cornell University professor who led the study, said, “It has an implicit set of values in terms of what it cares about and how it leans.” The test measured model responses, not their effect on voters.