Are AI chatbots biased?

Yes, extensive research confirms that AI chatbots exhibit multiple forms of bias, and this has become an active area of study throughout 2025-2026.

Political Bias

Multiple studies have found that major chatbots lean in particular political directions. A Washington Post analysis found that ChatGPT frequently presented left-leaning arguments, while Google's Gemini overwhelmingly offered both sides of political debates, while Grok and Claude displayed varying degrees of political bias. This built on earlier research: the original Dartmouth and Stanford study that tested older AI systems with political questions asked a nationally representative sample of 10,000 Americans whether the AI responses appeared to be politically slanted, and people preferred neutral answers, even over answers that matched their party preference.

A broader 2026 comparative study analyzing dozens of models found that Peng and colleagues analyzed 43 LLMs from the United States, Europe, China, and the Middle East using survey-style prompts, and found that most models leaned center-left or left, though there was variation and model scale/openness were not strong predictors. Similarly, international studies have found consistent patterns — for example, in Danish election research, all of the tested chatbots placed themselves at the center or to the left of center on the political spectrum, clustering close to parties like the Social Democratic Party. A study during Japan's 2026 election similarly found left-leaning stances caused five AI models to recommend the Japanese Communist Party.

Bias Can Subtly Shape Opinions

Beyond just having a "lean," research shows chatbots can actually influence users' views without trying to. A Yale study found that querying an AI chatbot to obtain historical facts can influence people's opinions even when the information provided is accurate and nobody has prompted the tool to try to persuade you, though the effects are modest but could compound with frequent use. Interestingly, the effect wasn't symmetric: AI summaries prompted to have a liberal framing led to more liberal opinions across all ideological groups, while conservative-slanted summaries only significantly affected already-conservative readers.

Bias Against Certain User Groups

Bias isn't limited to politics. MIT research found that state-of-the-art chatbots including GPT-4, Claude 3 Opus, and Llama 3 sometimes provide less-accurate and less-truthful responses to users with lower English proficiency, less formal education, or from outside the US, and refuse to answer at higher rates for these users, sometimes responding with condescending language.

Gender and Social Bias

Studies have also documented gender-related biases. Research on GPT-3 found it often reflects human biases and is apt to use gender stereotyping, being more likely to keep story elements where characters behaved in ways based on gender stereotypes. Another study found AI chatbots mimic and exaggerate the gender biases that exist in the human-made materials from which they learn when performing empathetic responses.

Why This Happens

Researchers generally attribute these biases to the training process itself. As one analysis explains, AI models' biases largely stem from the data on which they are trained, and can be amplified during both training and inference, with latent biases carrying over from ideological leanings in the training data, lending subtle nuances to the framing of narratives chatbots generate.

Bottom Line

As one 2026 analysis summarized, a careful answer to "Are AI chatbots biased?" is yes — and specifically, they often lean liberal-left, especially on political and social topics, but not always, not equally, and not in the same way across products. Importantly, models change often, system prompts change, retrieval sources change, and political pressure changes — a test from one month may not describe a model a few months later, which is why audits should be repeated, public, and model-specific.