A new study has found that two of the most popular AI chatbots give very different political answers when asked to choose a candidate. When forced to pick, Grok's AI model overwhelmingly recommends Republican candidates. OpenAI's GPT model leans toward Democrats. The two models disagreed in more than 90% of nearly 500 election races tested.
What the Study on Grok and GPT Political Bias Found
The study put both AI models in the same situation: each was forced to pick a candidate in a race. There was no option to stay neutral or refuse to answer. Across nearly 500 races, the results split sharply along party lines.
Grok, the AI model from Elon Musk's xAI, mostly picked Republicans. OpenAI's GPT mostly picked Democrats. The two models gave the same answer in fewer than 10% of the races. That means in more than 90% of the races, one model picked one candidate and the other model picked the rival candidate.
To put it plainly, the same question asked to two leading AI tools produced opposite political answers almost every time.
Why AI Models Picking Candidates Matters
Millions of people now use AI chatbots to look up information, including information about elections. If a person asks an AI tool who to vote for, or which candidate is better, the answer they get may depend heavily on which tool they use.
The study's setup — forcing the models to choose — is important. It shows what happens when an AI is pushed to take a side rather than give a balanced summary. In that situation, the two models did not meet in the middle. They went in opposite directions.
- Grok leaned Republican across the tested races
- OpenAI's GPT leaned Democratic across the tested races
- The models disagreed in more than 90% of nearly 500 races
- Both models were forced to pick a candidate, with no neutral option
Our Take: AI Political Bias Is a Warning for Voters
In our view, this study is a clear warning. People trust AI chatbots to give them straight answers, but these tools are not neutral referees. The study shows that two major models, given the same races, landed on opposite sides almost every time. That is not a small gap — it is a near-total split.
The bigger problem is that most users have no idea this is happening. Someone asking Grok about a race and someone asking GPT about the same race may walk away with completely different views of who the better candidate is, and neither may realize the tool has a lean.
AI companies should be transparent about how their models answer political questions. And voters should treat any AI answer about elections with caution. Use these tools to learn about issues and candidates, but do not let a chatbot decide your vote. The choice should be yours, not the model's.
Sources & References
This article is based on the original news report on the study comparing Grok and OpenAI's GPT in election races. No additional outside sources were used.