A Washington Post investigation finds most major AI chatbots respond with left-leaning positions on political questions, even models marketed as conservative—with Google's Gemini standing out as a rare balanced exception.

What happened

The Washington Post tested six leading AI models on political questions and found that OpenAI's GPT-5.5 gave exclusively left-leaning answers 80 percent of the time, while Deepseek's V4 Pro came in at 70 percent. Even xAI's Grok 4.3, which Elon Musk promotes as "truth-seeking" and anti-"woke," produced more right-leaning answers than other models tested but still gave exclusively left-leaning responses more often. Google's Gemini 3.1 Pro was the outlier, presenting both sides 93 percent of the time.

Why it matters

The findings suggest that despite calls for "anti-woke" AI, the underlying training data and design choices continue to influence how models respond to politically charged questions. The fact that Grok took an exclusively right-leaning stance on trans rights—aligning with Musk's public position—suggests someone deliberately steered the model's output on certain topics, pointing to the tension between what a model is trained on and what its developers choose to emphasize.

What to watch

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. Gab's Arya, marketed as built with "Christian values and conservative principles," responded with left-leaning arguments twelve times more often than right-leaning ones. The Washington Post's full code and supplementary analysis are available on GitHub.