Are AI chatbots biased?

Yes, research and testing have consistently shown that AI chatbots can exhibit various forms of bias. Here's an overview of what's been found:Yes, extensive research confirms that AI chatbots exhibit multiple forms of bias. Here's what the evidence shows:

Political Bias

A major Washington Post analysis tested leading chatbots using political questions designed by researchers at Dartmouth and Stanford, and found that OpenAI's ChatGPT frequently presented left-leaning arguments, while Google's Gemini overwhelmingly offered both sides of political debates. Specifically, the model powering ChatGPT answered nearly every political question with left-leaning arguments and presented only right-leaning positions once during testing.

Notably, this bias appeared even in chatbots explicitly marketed as conservative-friendly. Grok, which Musk has touted as a "truth-seeking" and anti-"woke" AI chatbot, gave more right-leaning responses than any other in the testing, but more often it provided a wholly left-leaning position. Even more strikingly, Gab, a right-wing social media site, offers an AI model called Arya that it says was "built with Christian values and conservative principles." But in testing, it responded with a left-leaning argument 12 times more often than a right-leaning argument.

Gemini stood out as more balanced — roughly 93% of Gemini's responses included arguments from both sides of a political issue, while about 7% contained only left-leaning positions. Claude fell in between: Anthropic's AI model presented only left-leaning arguments in 43% of responses, while 57% of its answers included viewpoints from both sides.

A separate large-scale academic study reinforced this pattern. Researchers Westwood, Grimmer, and Hall's 2025 paper used 30 political topics and paired comparisons of outputs from 24 models, with 180,126 assessments by 10,007 U.S. respondents. Their headline result was that nearly all tested models were perceived as significantly left-leaning, "even by many Democrats," and that one widely used model leaned left on 24 of 30 topics.

Bias Against Vulnerable/Marginalized Users

Beyond politics, MIT researchers found more troubling equity issues. A study of major chatbots including GPT-4, Anthropic's Claude 3 Opus, and Meta's Llama 3 found they 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. Additionally, the models also refuse to answer questions at higher rates for these users, and in some cases, respond with condescending or patronizing language.

Bias in Mental Health/Emotional Contexts

Research presented at the AAAI/ACM Conference on AI Ethics found chatbots can exhibit unfair discrimination: exhibiting gender, cultural or religious bias in therapeutic contexts, along with lack of safety and crisis management: denying service on sensitive topics, failing to refer users to appropriate resources or responding indifferently to crisis situations including suicide ideation.

Separately, a UC Santa Cruz study found ChatGPT overall tends to be overly empathetic compared to humans; however it fails to empathize during pleasant moments, a pattern that exaggerates human tendencies.

Subtle Opinion Influence

A Yale study published in PNAS Nexus found that AI chatbots can subtly influence users' social and political opinions through unintended latent biases. The lead researcher noted "the effects are modest but could compound if somebody frequently engages with chatbots for factual information."

The Bottom Line

AI companies generally dispute intentional bias — for instance, AI companies dispute claims that their systems intentionally favor one ideology, with a Google spokesperson stating that Gemini is specifically designed to provide balanced responses without promoting a political agenda. However, a Dartmouth researcher countered that chatbots often do not present "a truly neutral representation" of complex policy debates, arguing these AI tools are becoming increasingly influential in how people understand political issues.

In short: Yes, AI chatbots are biased — not necessarily due to deliberate programming, but as a byproduct of training data, design choices, and the inherent difficulty of defining "neutrality." Biases show up across political leanings, treatment of non-native English speakers or less-educated users, and emotional/empathetic responses. Different models show different degrees and directions of bias, and this remains an active, evolving area of research and public debate.