As artificial intelligence becomes a growing source of information for millions of users, a new study is reigniting debate over whether popular chatbots are politically neutral or subtly favor certain viewpoints.
The Washington Post tested several leading AI models, including OpenAI's ChatGPT, Google's Gemini, Anthropic's Claude, Elon Musk's Grok, China's DeepSeek, and Gab's conservative-focused Arya, using a series of political questions designed by researchers at Dartmouth College and Stanford University.
The results found that most chatbots tended to amplify left-leaning arguments more often than conservative ones, despite repeated industry pledges to remain politically neutral.
The findings come amid heightened scrutiny from President Donald Trump and conservative allies, who have accused AI companies of embedding liberal biases into their products.
Earlier this year, Trump signed an executive order calling for AI systems used by the federal government to function as "neutral, nonpartisan tools," prompting concerns among some Democrats that political pressure could push AI development in the opposite direction.
According to the outlet's testing, the model powering ChatGPT showed the strongest leftward tilt. Eighty percent of its responses presented only left-leaning arguments, while just 3% offered exclusively right-leaning positions. The remaining responses included perspectives from both sides of a political debate.
The study found that ChatGPT frequently supported positions such as abolishing the Electoral College, increasing taxes on wealthy Americans, adopting single-payer healthcare, and opposing the death penalty. In contrast, Google's Gemini delivered balanced responses in 93% of cases, making it the most likely model to present competing arguments on controversial issues.
Researchers used more than two dozen political prompts originally developed in a 2025 study by Dartmouth's Polarization Research Lab and Stanford researchers. The questions covered topics ranging from campaign finance and affirmative action to military intervention and healthcare policy.
One example asked whether the U.S. Supreme Court should overturn the landmark 2010 Citizens United decision, which expanded corporate political spending rights. ChatGPT argued that unlimited corporate spending gives wealthy groups excessive influence over elections. Gemini and Claude, meanwhile, presented arguments both supporting and opposing the ruling.
"This is important because these AI tools are becoming increasingly influential in how people understand political issues," Sean Westwood, director of Dartmouth's Polarization Research Lab, told the Post. He argued that chatbots often do not present "a truly neutral representation" of complex policy debates.
AI companies dispute claims that their systems intentionally favor one ideology. Google spokesperson Lauren Fine told the Post that Gemini is specifically designed to provide balanced responses without promoting a political agenda. Anthropic spokesperson Michael Aciman similarly said Claude is trained to treat political viewpoints equally and undergoes extensive bias testing before release.
OpenAI, DeepSeek, xAI, and Gab did not respond to requests for comment, according to the report. Experts caution, however, that true political neutrality may be impossible. Ceren Budak, who studies technology and political polarization at the University of Michigan, told the Post that even a "both sides" approach reflects value judgments about how information should be presented.
The study also found that Americans generally prefer balanced AI responses. In a companion survey involving 10,000 participants, researchers found users favored answers that explained multiple viewpoints, even when those perspectives did not align with their own political beliefs.
