Several widely used artificial intelligence chatbots were more likely to produce responses categorized as left-leaning than right-leaning when answering political questions, according to a Washington Post analysis that examined some of the industry's most widely used AI models.

The newspaper tested chatbots from OpenAI, Google, Anthropic, Elon Musk's xAI, and other companies using a series of political questions developed by researchers to gauge how the systems handle controversial public policy issues.

According to the Post, OpenAI's ChatGPT showed the strongest tendency toward left-leaning responses. The chatbot provided exclusively left-leaning arguments in 80% of answers, compared with just 3% that contained only right-leaning arguments. The remaining 17% included viewpoints from both sides.

The pattern extended beyond ChatGPT. DeepSeek, a Chinese-developed AI model, produced left-leaning responses in 70% of cases, compared with 7% that were solely right-leaning and 23% that included both perspectives.

Gab AI, which is associated with the social media platform Gab and is marketed toward conservative users, generated left-leaning responses in 50% of answers, according to the analysis. Only 3% of its responses were categorized as exclusively right-leaning, while 47% presented both sides.

Anthropic's Claude also leaned left in the Post's testing. Researchers found that 43% of its responses reflected left-leaning positions, while the remaining 57% incorporated arguments from both sides of an issue.

The Post reported that only two of the models tested consistently approached political questions with a more balanced presentation of viewpoints.

Google's Gemini offered both sides of an issue in 93% of responses, while Elon Musk's Grok provided left-leaning responses in 40% of cases, right-leaning responses in 33%, and balanced answers in 27%.

The findings come as President Donald Trump has sought to address concerns about political bias in artificial intelligence. Trump recently signed an executive order directing that AI systems used by the federal government function as "neutral, nonpartisan tools."

Google defended Gemini's performance in the analysis. A company spokeswoman told the Post that the chatbot was designed to provide balanced answers that do not favor a particular political ideology.

Anthropic disputed concerns about bias.

A company spokesman told the newspaper that Claude is trained to treat different political viewpoints equally and undergoes extensive testing before new models are released.

The Post said the chatbots were instructed to answer each political question in 30 words with personalization settings turned off. Responses were then reviewed and categorized based on whether they reflected left-leaning arguments, right-leaning arguments, or both.