Are AI chatbots biased?

AI chatbots can and often do exhibit various forms of bias. Let me search for current information on this topic to give you a well-supported answer.Yes, extensive research confirms that AI chatbots exhibit multiple forms of bias. Here's what the evidence shows:

Political Bias

Numerous studies have found that most major AI chatbots lean left politically. A comparative analysis of 43 LLMs from the US, Europe, China, and the Middle East found that most models leaned center-left or left when tested using survey-style prompts from the American National Election Studies and Pew Research Center. Similarly, research from the University of Copenhagen testing chatbots in a Danish election context found that all of the tested chatbots place themselves at the center or to the left of center on the political spectrum, clustering close to parties like the Social Democratic Party.

A Washington Post investigation found that Trump and other conservatives have accused AI chatbots of being politically biased against them, and testing showed Gemini scored highest for presenting multiple viewpoints (93% of responses), while Claude maintained neutrality about 57% of the time. Interestingly, one notable exception emerged: a March 2026 benchmark of eight prominent models found that seven leaned left and one, Grok, leaned right.

Stanford research adds nuance here, noting that models trained on the internet alone tend to be biased toward less educated, lower income, or conservative points of view, but newer models further refined through curated human feedback tend to be biased toward more liberal, higher educated, and higher income audiences.

Bias Based on User Characteristics

Beyond politics, MIT researchers found troubling patterns in how chatbots treat different users. Testing GPT-4, Claude 3 Opus, and Llama 3, they found these models sometimes provide less-accurate and less-truthful responses to users who have lower English proficiency, less formal education, or who originate from outside the United States. Even more concerning, the models also refuse to answer questions at higher rates for these users, and in some cases, respond with condescending or patronizing language.

Hidden Influence on Opinions

Perhaps most subtly, a Yale study found that chatbots can shape political views even when not explicitly asked to persuade anyone. Researchers discovered that the same persuasive effect is true of content that is not intended to change minds, such as summaries that popular chatbots produce in response to simple queries about historical events. This happens because this unintended power to persuade is caused by latent biases introduced during the training of the large language models that drive chatbots' core capabilities.

Language-Dependent Bias

Bias isn't even consistent across languages. Research testing ChatGPT and Gemini in 14 different languages found that these large language models do exhibit political tendencies, with both models demonstrating liberal and leftist biases, though Gemini exhibited a more pronounced liberal and left-wing tendency compared to ChatGPT, and crucially, these political biases varied depending on the language used for inquiry.

Can Bias Be Fixed?

There's some good news: bias appears somewhat malleable. Stanford researchers found that with just a small tweak, many models can be prompted to take a more neutral stance that more users trust. Similarly, other research confirmed that when models were prompted to take a more neutral stance, users perceived less slant and showed more interest in using the model, meaning at least some of the tilt can be reduced by instruction design, although not necessarily eliminated.

The Bottom Line

Even AI researchers acknowledge bias may be unavoidable to some degree. As one Carnegie Mellon researcher put it: "We believe no language model can be entirely free from political biases." This happens because language models are trained on massive amounts of human-generated text and then fine-tuned using human feedback—both processes that inevitably introduce the perspectives, assumptions, and blind spots of the humans and data involved.

For users, the practical takeaway is to treat chatbot outputs critically, especially on political or contested topics, cross-check information from multiple sources, and be aware that seemingly neutral summaries can still carry subtle framing effects.