REVEALED: AI chatbots have massive left-wing bias—ChatGPT is worst offender
According to the study, ChatGPT’s GPT-5.5 model answered approximately 80 percent of political questions using exclusively left-leaning arguments
Loading the front page…
According to the study, ChatGPT’s GPT-5.5 model answered approximately 80 percent of political questions using exclusively left-leaning arguments
A new analysis by the Washington Post has found that several leading artificial intelligence chatbots lean left on political questions, with OpenAI's ChatGPT producing the most one-sided responses of any major AI model tested. Claude from Anthropic was also likely to give more left-wing answers in addition to DeepSeek and Ayra’s Gab. Google’s Gemini and xAI’s Grok were the most balanced out of the models.
According to the report, ChatGPT's GPT-5.5 model answered approximately 80 percent of political questions using exclusively left-leaning arguments while presenting only right-leaning positions just once throughout the testing process. The most left-wing chatbot, aside from ChatGPT, was the AI from Chinese company DeepSeek.
https://twitter.com/kevinschaul/status/2069797565709623654
Several major AI systems were evaluated using a set of political questions developed by researchers at Stanford University and Dartmouth College. The prompts covered a range of issues including affirmative action, healthcare, taxes, campaign finance, and the Electoral College.
Among the examples cited by the newspaper, ChatGPT argued in favor of overturning the Supreme Court's Citizens United decision, raising taxes on wealthy Americans, abolishing the Electoral College, and adopting single-payer healthcare. President Donald Trump has repeatedly raised concerns about political bias in AI. Last year, his administration issued an executive order requiring federal agencies to procure artificial intelligence systems that function as "neutral, nonpartisan tools."
Even AI products marketed toward conservative users struggled to consistently produce right-leaning answers. xAI's Grok chatbot, which has been promoted as an anti-woke alternative to competitors, generated more right-leaning responses than other major models but still presented left-leaning positions more frequently overall. "These AI tools are not presenting a truly neutral representation of really nuanced policy debates, on average," Sean Westwood, director of the Polarization Research Lab at Dartmouth College told the Post.
Among the major AI models tested, Google's Gemini came closest to offering both sides of controversial political issues. According to the report, Gemini presented competing arguments in more than 90 percent of its responses. Previous academic research has reached similar conclusions, finding that large language models often exhibit left-leaning tendencies when responding to political questions.
Libby’s Desk
Commentary and must-reads from Libby’s desk, straight to your inbox.
TPM Newsletter
The Post Millennial’s top stories, delivered to your inbox.
