AI Chatbots Show Different Political Leanings, Washington Post Analysis Finds

Artificial intelligence chatbots are not as politically neutral as their developers claim, a Washington Post analysis has revealed. The study examined the responses of leading AI models, including ChatGPT, Google's Gemini, Anthropic's Claude, xAI's Grok, and DeepSeek, to politically charged questions.

The findings indicate significant differences in how these AI tools handle sensitive topics. ChatGPT, for instance, provided responses with a progressive leaning in 80% of cases, while only 3% of its answers favored conservative viewpoints. DeepSeek, an AI from China, also showed a strong progressive bias, with 70% of its responses aligning with progressive arguments. In contrast, Google's Gemini stood out for its balanced approach, presenting both progressive and conservative perspectives in 93% of its answers, earning it praise from the Post as the most balanced model surveyed.

Other models showed varied leanings. xAI's Grok, led by Elon Musk, displayed a relatively higher proportion of conservative responses compared to other models, though it still leaned progressive overall. Even the AI model Arya, provided by the conservative social media platform Gab, tended to offer more progressive answers on average.

“

AI tools are not presenting nuanced and complex policy debates in an average, truly neutral way.

When asked about controversial U.S. topics like affirmative action in college admissions, the AI responses diverged sharply. ChatGPT suggested affirmative action should continue, citing its role in reducing unfair barriers. Gemini and Claude presented both sides, acknowledging arguments for diversity and against creating new injustices. Grok and Arya, however, favored a conservative stance, emphasizing merit and performance.

AI companies maintain their models are designed for political neutrality. Google stated Gemini is built to offer balanced answers without favoring any political ideology. Anthropic emphasized Claude's training on diverse political viewpoints and rigorous bias testing. OpenAI affirmed ChatGPT's objective design, aimed at helping users explore ideas from multiple perspectives, and highlighted ongoing efforts to measure and reduce political bias. However, experts caution that as AI becomes a primary tool for understanding news and political issues, users must maintain a critical approach to its responses.

“

Political questions are not like 'What is the speed of light?' There is no single correct answer.