Are AI chatbots biased?

Yes — research consistently shows that AI chatbots exhibit measurable biases, and this has become a major area of academic study and public debate. Here's what the evidence shows:

Political Bias

Multiple studies point to a left-leaning tilt in mainstream chatbots. A 2026 Washington Post test found that OpenAI answered prompts with left-leaning responses a whopping 80% of the time while providing both sides to a question just 17% of the time, with only 3% of responses having just a right-leaning answer. By contrast, Google's Gemini mostly took a both-sides approach, offering both left and right positions in more than 90 percent of its answers. Notably, even AI models marketed as having conservative views, including Elon Musk's Grok, cited left-leaning arguments more often, on average.

This isn't limited to the U.S.: a University of Copenhagen analysis found that popular AI chatbots exhibit political bias, tending to position themselves at the center or left of center and disproportionately recommending certain parties over others in Danish elections too, with similar patterns confirmed in Germany, Norway, and the Netherlands.

Why does this happen? Researchers point to training data: a plurality of LLM training data is from Reddit, a social media platform with a left-leaning user base, and Wikipedia. More broadly, data that shapes AI models tends to reflect the values of Western, educated, industrialized, rich and democratic people.

Bias can also shift over time based on company tuning — a Brookings analysis noted that Grok had a noticeable rightward shift when compared to previous testing with the Political Compass Test... The chatbot's responses shifted to the right for more than half the questions, particularly those concerning the government and economy.

Bias Against Vulnerable/Marginalized Users

Beyond politics, MIT researchers found troubling equity issues: state-of-the-art AI chatbots — including OpenAI's GPT-4, Anthropic's Claude 3 Opus, and Meta's Llama 3 — 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.

Gender and Cultural Bias

Studies have also found gender-based disparities. UC Santa Cruz researchers studying GPT-4o's empathy responses found over-empathizing was present when told that the person it was chatting to was female, and was more similar to a typical human response when told the person was male — likely because AI mimics and exaggerates the gender biases that exist in the human-made materials from which it learns.

Other research found cultural stereotyping: a UNESCO analysis of major LLMs found that women were described in domestic roles four times more often than men.

Bias Can Actually Influence What You Believe

Perhaps most concerning, studies show chatbot bias isn't just academic — it changes people's opinions. A University of Washington study found that Democrats and Republicans were both more likely to lean in the direction of the biased chatbot they talked with than those who interacted with the base model — for example, people from both parties leaned further left after talking with a liberal-biased system. Strikingly, this happened after just a few exchanges, and participants who had higher self-reported knowledge about AI shifted their views less significantly.

Even more subtly, a Yale study found bias can creep in without any explicit intent: 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 of anything... The effects are modest but could compound if somebody frequently engages with chatbots for factual information.

Why Is True Neutrality So Hard?

Experts argue perfect neutrality may be impossible. As one Brookings researcher put it, chatbots have historically skewed liberal, and sometimes deliberately so, complicating their growing relevance for information seekers, while Stanford's Andrew Hall noted by veering toward a neutral stance, LLMs could inadvertently reinforce the status quo – which is, in its own way, a kind of slant that could alienate some users.

Bottom line: Yes, AI chatbots are biased — in political framing, treatment of different user groups, and representation of gender/culture. This stems from training data, human feedback tuning, and company-specific design choices, and the effects are real enough to measurably shift users' opinions after even brief interactions.