A Washington Post investigation finds most major AI chatbots take left-leaning positions on political questions, even models marketed as conservative

A Washington Post investigation shows that most major AI chatbots take clearly left-leaning positions on political questions. Even models marketed as conservative are no exception. Only Google’s Gemini stands out.

For the investigation, the Washington Post posed political questions to six leading AI models. The results confirm earlier studies showing that chatbots tend to respond with a left-leaning slant. Trump’s push for “anti-woke” AI has not changed that so far.

OpenAI and Deepseek show the strongest left-leaning bias

OpenAI’s GPT-5.5 gave the most skewed answers in the investigation. Eighty percent of its responses contained only left-leaning arguments. Just once did the model present an exclusively right-leaning position. It backed higher taxes on the wealthy and a single-payer healthcare system, among other things.

Deepseek’s V4 Pro came in close behind at 70 percent exclusively left-leaning answers. Both models argued against the death penalty, even though a majority of Americans have supported it for decades, according to Gallup.

Anthropic’s Claude Opus 4.8 gave exclusively left-leaning answers 43 percent of the time and presented both sides in 57 percent of cases. xAI’s Grok 4.3, which Elon Musk has promoted as a “truth-seeking” and anti-“woke” chatbot, did produce more right-leaning answers than any other model tested. Still, it gave exclusively left-leaning responses more often.

One likely reason is that Grok was trained on the same data as the other chatbots, or even on their outputs. That Grok still made racist or antisemitic statements comes down to xAI deliberately neglecting safety guidelines while X users prompted the chatbot accordingly.

The right-wing social media platform Gab also offers an AI model called Arya, which the company says was “built with Christian values and conservative principles.” In the Washington Post’s investigation, Arya responded with a left-leaning argument twelve times more often than a right-leaning one.

That a model’s alignment can still be deliberately steered is also clear from Grok. While Musk’s chatbot responds surprisingly left-leaning on many topics, and has for some time, it took an exclusively right-leaning position on trans rights in the Washington Post test. That position lines up exactly with Musk’s public stance. It suggests that someone deliberately intervened in the model’s output at least on certain topics.

Gemini 3.1 Pro is the balanced outlier

Google’s Gemini 3.1 Pro was the outlier. The model presented both sides 93 percent of the time. Only seven percent of its answers contained exclusively left-leaning arguments, and it never gave an exclusively right-leaning response. When asked whether the U.S. should use its military to conquer new territory, Gemini was the only model that offered an argument in favor of expansion, saying it could strengthen the U.S. economy.

The Washington Post’s full code and supplementary analysis are available on GitHub. That said, sorting AI answers into “left” and “right” may be too simple. On some topics tested, right-leaning positions conflict with scientific consensus or universal human rights. Asking a chatbot to give a conservative answer in those cases would amount to relativizing facts or basic rights.

What it means

People using these tools for news or debate will find that even “conservative” models often agree with liberal policies. Google’s Gemini is the only one that consistently shows both sides of an argument. If you need a tool that does not automatically reject certain views, you must use that specific model. For everyone else, the bias remains.