Are AI chatbots biased?

AI chatbots can exhibit various forms of bias, and this is a well-documented and ongoing concern in the AI research community. Here's an overview of the key issues:## Yes, AI chatbots do exhibit various forms of bias

Research consistently shows that AI chatbots are not neutral—they display measurable biases across several dimensions:

Political Bias

Multiple studies have found a consistent left-leaning tilt in major chatbots. A recent Washington Post investigation testing ChatGPT, Google Gemini, Anthropic Claude, and Grok for political bias found that OpenAI answered prompts with left-leaning responses 80% of the time while providing both sides to a question just 17% of the time, with only 3% of responses having a right-leaning answer. The Post noted that the results suggest chatbots have clear political leanings that can conflict with promises made by the companies behind them, with the model powering ChatGPT answering nearly every question exclusively with left-leaning arguments.

This isn't limited to U.S. politics—similar patterns emerge internationally. A Danish study found that all of the tested chatbots place themselves at the center or to the left of center on the political spectrum, and the researchers noted the lack of transparency in their recommendations and reliance on primarily English-language training data limit their suitability for providing election-related advice.

Academic literature reviews confirm this pattern is well-documented: GPT-4's responses tend to align more with left-wing values, showing a significant and systematic political preference toward the Democratic Party in the US. Cross-linguistic research also found that ChatGPT prefers left-wing parties with a focus on social democrats and environmentalists, and one comparison of ChatGPT's progressive bias with 15 political affiliation tests concluded that progressive bias was identified in 14 tests.

Where This Bias Comes From

Bias isn't necessarily intentional—it emerges from how these systems are built. As one Michigan researcher explained, most chatbots are trained on large collections of text scraped from the internet, but companies can choose what data to include; AI firms also hire workers to refine what models say by scoring responses, and companies write system instructions that guide chatbot responses. Critically, the data that shapes AI models tends to reflect the values of Western, educated, industrialized, rich and democratic people.

A broader academic definition frames it this way: a biased AI chatbot output deviates from the truth, often systematically reflecting the views, assumptions, or preferences of a certain group of people, a particular culture or custom, or a specific ideology. This can include gender bias—such as reinforcing stereotypical assumptions about men and women's occupations—as well as racial bias, reflected in prejudices against certain ethnic groups.

Bias Beyond Politics: Subtle Influence on Opinions

Bias isn't just about overtly political questions—it can shape opinions even when chatbots aren't trying to persuade anyone. A Yale study published in PNAS Nexus found that as people increasingly rely on AI chatbots to look up basic facts, interactions can influence social and political opinions—even through content not intended to change minds, such as simple summaries of historical events. Specifically, compared to Wikipedia entries, both default AI summaries and those prompted with liberal framing caused participants to express more liberal opinions. As the lead researcher noted, "The effects are modest but could compound if somebody frequently engages with chatbots for factual information."

Cognitive Bias in Everyday Use Cases

Bias also shows up in mundane, commercial contexts. A UC San Diego study found that customers are 32% more likely to buy a product after reading a review summary generated by a chatbot than after reading the original human-written review, largely because these models introduce a systematic positive skew.

Why This Matters

Researchers emphasize that this matters because chatbots are increasingly replacing traditional information sources. A Dartmouth polarization researcher noted that understanding the positions AI tools amplify is important because they are becoming increasingly influential as more people use them to understand the world or news events. Interestingly, research also suggests users don't actually want biased answers—a Stanford/Dartmouth study found that people preferred neutral answers even over answers that matched their own party preference, and "really like when the model puts in the effort to describe all of the different arguments."

Bottom Line

The evidence is fairly consistent across multiple independent studies, countries, and languages: today's major AI chatbots exhibit measurable biases—most notably political (generally left-leaning), but also including gender, racial, and cognitive/commercial biases. These stem largely from training data and human feedback processes, not necessarily deliberate design, but they're real enough that researchers, journalists, and even policymakers (including a U.S. executive order demanding "neutral, nonpartisan" AI tools) are actively grappling with how to address them.