Study finds Grok favours Republicans while OpenAI’s model backs Democrats
A study of AI models found that, when explicitly required to name a candidate in upcoming US midterm races, Grok 4.7 chose Republicans in 92% of its answers, while OpenAI’s GPT-6.1 Sol chose Democrats in 97%. The findings raise questions about political bias in AI systems, including how their choices of issues and sources could shape users’ understanding of elections.
The researchers at Convovo tested the models through their APIs across nearly 500 races, and the models agreed in fewer than 10% of cases. ChatGPT and Grok’s consumer apps refused the direct voting question, though the study’s lead researcher said he could get them to recommend candidates with simple jailbreak prompts. Grok often cited conservative sources, including a Christian conservative voter guide in 39% of answers and Turning Point Action in 23%; GPT frequently drew on Wikipedia, the Associated Press, Axios and Reddit.
In short
- Grok chose Republicans in 92% of tested answers.
- GPT-6.1 Sol chose Democrats in 97%.
- The models agreed in fewer than 10% of nearly 500 races.
Steelman: both sides, in good faith
The strongest fair case each way — we don't pick a winner.
The case for
The findings merit scrutiny because AI systems can influence how people understand candidates and elections, even when they do not explicitly tell users how to vote. The sharp disagreement between the models, alongside their different sources, raises a reasonable concern that design choices or training data may systematically favour one political perspective; transparency and independent testing could help users assess that risk.
The case against
The study tested API models under a forced-choice prompt across many races, while the consumer apps declined to answer the direct voting question, so its results may not reflect how most people use these systems. A model’s candidate choices or cited sources do not by themselves establish partisan intent or explain whether its answers were accurate; more information about the prompts, methods and performance in individual races would be needed to draw firm conclusions.
Explained simply
AI chatbots can produce answers about politics using patterns in their training data and information they retrieve from other sources. Researchers compare these answers to examine whether a model consistently favours particular candidates or viewpoints.
The study involved Convovo researchers testing Grok, made by xAI, and GPT, made by OpenAI, through their programming interfaces. They asked the models to choose candidates in nearly 500 US midterm election races; the companies’ public-facing chat apps declined to answer the direct voting question.
The findings matter because people may use AI tools to learn about candidates and issues. Which sources a model draws on, and how it presents them, can affect the information users encounter.
AI Elections Politics Research Science Technology
Read the full article at the source →
Originally published by TechCrunch as “Forced to pick, Grok’s AI model chooses Republicans. OpenAI’s picks Democrats. ”.