The Washington Post conducted a test to assess the political bias of AI chatbots, including OpenAI’s ChatGPT and Google’s Gemini, amidst accusations from President Donald Trump and conservatives that these tools are politically biased. The testing revealed that OpenAI’s model predominantly provided left-leaning responses, with 80 percent of its answers favoring this perspective, while Gemini offered a more balanced approach, presenting both sides in over 90 percent of its responses. This comes in the context of an executive order signed by Trump mandating chatbots to be neutral, a concept that users prefer according to research from Dartmouth and Stanford, yet one that many AI chatbots struggle to achieve in practice, as previously highlighted by multiple academic studies.

Gab: Gab operates as a right-wing social media platform and provides the Arya AI model. The company markets Arya as built with Christian values and conservative principles. Washington Post testing found the model responded with left-leaning arguments far more frequently than right-leaning ones.
Grok: Grok is the AI chatbot offered by Elon Musk through SpaceX. It was evaluated in Washington Post tests and produced more right-leaning responses than any other model tested. Despite this, it still frequently provided wholly left-leaning positions across the questions.
Claude: Claude is the chatbot developed by Anthropic. It was tested by the Washington Post alongside other models using researcher-designed political questions. Responses from Claude often presented opposing perspectives on issues such as the Citizens United decision.
Google: Google develops the Gemini AI model used across its chatbot products. The model was included in Washington Post testing that examined political leanings across hot-button issues. A company spokesperson stated that Gemini is designed to deliver balanced responses without favoring any political ideology.
OpenAI: OpenAI develops advanced AI language models that power the ChatGPT chatbot. Its GPT-5 model was the focus of Washington Post testing on responses to political questions. The company has publicly aimed for neutrality in its models while acknowledging that personalization may be needed to address differing views of what neutral means.
SpaceX: SpaceX is the company Elon Musk has linked to offering the Grok AI chatbot. The Washington Post testing evaluated Grok responses on political topics as part of broader scrutiny of chatbot bias. The entity appears in the news specifically through its association with Grok.
DeepSeek: DeepSeek is a Chinese AI company whose model was evaluated in the Washington Post political bias tests. Its responses aligned closely with left-leaning positions in a manner similar to OpenAI’s model. The company did not respond to requests for comment on the test results.
Anthropic: Anthropic develops the Claude AI models and chatbot. Its Claude Opus 4 model participated in Washington Post testing on political question responses. A company spokesperson emphasized training Claude to treat different political viewpoints equally and extensive pre-launch bias testing.
Elon Musk: Elon Musk founded and leads multiple technology companies and offers the Grok chatbot. He has marketed Grok as a truth-seeking and anti-woke alternative to other AI systems. The Washington Post included Grok in its bias testing and found it cited left-leaning arguments more often than right-leaning ones on average.

Executive Order: President Donald Trump signed an executive order requiring AI chatbots to function as neutral, nonpartisan tools.
User Preference: Research from Dartmouth and Stanford showed that people prefer neutral AI answers even over responses that match their own party preference.
Academic Studies: Multiple prior academic studies have found that AI models powering chatbots tend to favor left-leaning positions on political topics.