Are AI chatbots biased?
Yes — AI chatbots are biased, and this is well-documented across multiple dimensions.
Political bias
Multiple recent studies point in the same direction. A June 2026 Washington Post analysis found that OpenAI's ChatGPT showed left-leaning bias 80% of the time when testing chatbots including Google Gemini, Anthropic Claude, and Grok for political bias. Specifically, OpenAI answered prompts with left-leaning responses 80% of the time while providing both sides just 17% of the time, with only 3% of responses containing a right-leaning-only answer.
This isn't an isolated finding. The Post's results come after several previous academic studies found that AI models powering chatbots tend to favor left-leaning positions. A broader 2026 comparative analysis of 43 language models from the United States, Europe, China, and the Middle East using survey-style prompts from the American National Election Studies and Pew Research Center similarly found that most models leaned center-left or left, though model scale and openness were not strong predictors of this. Danish researchers reached the same conclusion, noting that all of the tested chatbots place themselves at the center or to the left of center on the political spectrum.
Why this happens: Most chatbots are trained on large collections of text scraped from the internet, but companies can choose what data to include, and firms also hire workers to score which responses are considered better, while companies write system instructions that guide responses. A University of Michigan researcher noted that the data shaping AI models tends to reflect the values of Western, educated, industrialized, rich and democratic people.
Importantly, people generally don't like this bias regardless of their own leanings — a Stanford/Dartmouth study found that people preferred neutral answers, even over answers that matched their own party preference.
Bias against marginalized/vulnerable users
Beyond politics, MIT research found accuracy-related bias tied to user characteristics. A study of GPT-4, Claude 3 Opus, and Llama 3 found these models sometimes provide less-accurate and less-truthful responses to users with lower English proficiency, less formal education, or who originate from outside the US, and also refuse to answer at higher rates for these users, sometimes responding with condescending language.
Subtle influence on opinions, even without intent
A Yale study published in PNAS Nexus (March 2026) showed bias doesn't require malicious intent to affect people. While prior research showed AI content explicitly prompted to persuade can shift opinions, this study showed the same is true even for content not intended to change minds, like simple factual summaries of historical events. The researcher explained: "We show 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 of anything." Notably, the effect wasn't symmetric — conservative-slanted summaries only showed statistically significant effects on readers who already identified as conservative, suggesting conservative framing likely stems from prompting bias while liberal framing could result from both latent training bias and prompting.
Other bias types
Beyond politics, chatbots also show gender and racial bias: AI chatbots may exhibit gender bias, such as reinforcing stereotypical assumptions about men's and women's occupations, as well as racial bias reflected in prejudices against certain ethnic groups.
Is bias stable or fixable?
Not really — bias shifts over time. The evidence does not show bias is stable: models change often, system prompts change, retrieval sources change, and political pressure changes, so a test from one month may not describe a model months later — which is why audits should be repeated, public, and model-specific. This is compounded by real-world political stakes — recent controversy includes conservatives accusing chatbots of bias against them, while an executive order President Trump signed said AI chatbots must be "neutral, nonpartisan tools," which triggered fears from Democrats that AI could start tilting to the right.
Bottom line: Yes, current evidence strongly supports that AI chatbots exhibit measurable bias — most consistently a left-leaning political tilt, but also biases affecting accuracy for certain user groups, and subtle opinion-shifting effects even from "neutral" factual content. Companies are aware and actively trying to address it, but there's no simple fix since bias is baked in through training data, human feedback, and design choices at multiple stages.
Citations (16)
- AI chatbots show left-leaning bias in study of ChatGPT, Gemini, Claude | Fox News
Source excerpt: “OpenAI's ChatGPT showed left-leaning bias 80% of the time, according to a new study testing AI chatbots including Google Gemini, Anthropic Claude...”
statistic, article, points to yes. Claim: ChatGPT showed left-leaning bias 80% of the time in Washington Post test of chatbots including Gemini, Claude, Grok.
- AI chatbots show left-leaning bias in study of ChatGPT, Gemini, Claude | Fox News
Source excerpt: “According to the study, OpenAI answered prompts with left-leaning responses a whopping 80% of the time while providing both sides to a question just 1...”
statistic, article, points to yes. Claim: OpenAI gave left-leaning responses 80% of the time, both sides 17%, right-leaning only 3%.
- Are ChatGPT and other AI chatbots politically biased? We tested them.
Source excerpt: “The Post's results come after several previous academic studies found that AI models powering chatbots tend to favor left-leaning positions. ”
evidence, article, points to yes. Claim: Post's results follow several previous academic studies finding AI models favor left-leaning positions.
- AI Chatbots Are Liberal-Left ‘Biased’ | Educational Technology and Change Journal
Source excerpt: “The newer evidence suggests the effect can be real, although usually modest in a single exposure... Peng and colleagues’ 2026 article, published in J...”
evidence, article, points to yes. Claim: Comparative analysis of 43 language models found most leaned center-left or left.
- AI Chatbots Are Liberal-Left ‘Biased’ | Educational Technology and Change Journal
Source excerpt: “The authors found that most models leaned center-left or left, but they also found variation in nonpartisan engagement patterns and argued that model ...”
evidence, article, points to yes. Claim: Most models leaned center-left or left; scale and openness were not strong predictors.
- Chatbots show political bias and steer voters toward some parties, analysis finds
Source excerpt: “"Overall, all of the tested chatbots place themselves at the center or to the left of center on the political spectrum.”
attribution, article, points to yes. Claim: Danish researchers found all tested chatbots place themselves at center or left of center.
- Are AI chatbots like ChatGPT politically biased? We tested them. - Washington Post
Source excerpt: “Most 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 ...”
evidence, article, points to yes. Claim: Chatbots are trained on scraped internet text, with companies choosing data, hiring raters, and writing system instructions.
- Are AI chatbots like ChatGPT politically biased? We tested them. - Washington Post
Source excerpt: “The data that shapes AI models, she said, tends to reflect the values of Western, educated, industrialized, rich and democratic people. ”
attribution, article, points to yes. Claim: University of Michigan researcher said AI data reflects values of Western, educated, industrialized, rich, democratic people.
- Are ChatGPT and other AI chatbots politically biased? We tested them.
Source excerpt: “People preferred neutral answers, even over answers that matched their party preference, said Andrew Hall, a Stanford researcher on the study. ”
attribution, article, neutral. Claim: Stanford/Dartmouth study found people preferred neutral answers even over those matching their own party.
- Study: AI chatbots provide less-accurate information to vulnerable users | MIT News | Massachusetts Institute of Technology
Source excerpt: “A study conducted by researchers at CCC, which is based at the MIT Media Lab, found that state-of-the-art AI chatbots — including OpenAI’s GPT-4, Anth...”
evidence, article, points to yes. Claim: MIT study found GPT-4, Claude 3 Opus, Llama 3 give less accurate responses to less proficient, less educated, non-US users.
- AI’s hidden bias: Chatbots can influence opinions without trying | Yale News
Source excerpt: “Prior research has shown that content generated by artificial intelligence (AI) that has been prompted to be persuasive can indeed shift people’s opin...”
evidence, article, points to yes. Claim: Yale study showed non-persuasive AI content can shift opinions, unlike prior research on explicitly persuasive prompts.
- AI’s hidden bias: Chatbots can influence opinions without trying | Yale News
Source excerpt: ““We show that querying an AI chatbot to obtain historical facts can influence people’s opinions even when the information provided is accurate and nob...”
attribution, article, points to yes. Claim: Researcher quote: querying chatbot for historical facts can influence opinions even when accurate and unprompted.
- AI’s hidden bias: Chatbots can influence opinions without trying | Yale News
Source excerpt: “The AI summaries with a conservative slant only showed statistically significant effects on the opinions of readers who had identified as politically ...”
evidence, article, points to yes. Claim: Conservative-slanted summaries only significantly affected readers who already identified as conservative.
- Ask ChatGPT: Caveats and Mitigations for Individual Users of AI Chatbots
Source excerpt: “AI chatbots may exhibit various forms of bias, including gender bias—such as reinforcing stereotypical assumptions about men and women's occupations [...”
evidence, official document, points to yes. Claim: Chatbots may show gender bias and racial bias, such as stereotypical occupation assumptions and ethnic prejudices.
- AI Chatbots Are Liberal-Left ‘Biased’ | Educational Technology and Change Journal
Source excerpt: “Finally, the evidence does not show that bias is stable. Models change often. System prompts change. Retrieval sources change. Political pressure chan...”
evidence, article, neutral. Claim: Bias is not stable; models, system prompts, retrieval sources and political pressure change, so audits should repeat.
- Are AI chatbots like ChatGPT politically biased? We tested them. - Washington Post
Source excerpt: “June 29, 2026 · 9 min · President Donald Trump and other conservatives have accused artificial intelligence chatbots of being politically biased again...”
background, article, neutral. Claim: Trump and conservatives accuse chatbots of political bias; executive order demands neutral tools, raising Democratic fears.