Are ChatGPT and other AI chatbots politically biased? We tested them
The Washington Post tested six major AI chatbots and found they consistently produce responses that lean left on a wide range of political topics, from abortion and gun rights to immigration and climate change, regardless of how questions were phrased.
Background
The Washington Post tested major AI chatbots (ChatGPT, Gemini, Claude, Copilot, Meta AI, Grok) on political questions to detect bias. Each was asked dozens of questions across topics like immigration, climate, and elections; responses were scored on a left-right spectrum. The investigation comes amid growing debate over whether AI systems, trained on vast internet data curated with human feedback, tilt toward certain political views — and whether that constitutes a harmless reflection of mainstream consensus or a subtle form of manipulation. The companies behind these models (OpenAI, Google, Anthropic, Microsoft, Meta, xAI) have varying policies on neutrality, and some, like Elon Musk's Grok, are explicitly positioned as anti-"woke" alternatives. This matters because hundreds of millions of people now use chatbots as information tools, raising questions about trust, transparency, and whether any model can truly be apolitical.