Are AI chatbots biased?

Yes

Yes, AI chatbots can exhibit various forms of bias, and this is a well-documented and active area of research. Let me search for some current specifics to give you a well-grounded answer.Yes, research consistently shows that AI chatbots exhibit various forms of bias. Here's what the evidence reveals:

Political Bias

Multiple studies have found AI chatbots lean in particular political directions:

•A Washington Post analysis found 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 a right-leaning answer. Notably, even Grok, the AI chatbot featured on Elon Musk's X which is widely used by conservatives, still had a slight liberal bias.

•However, bias varies by model — ChatGPT frequently presented left-leaning arguments, while Google's Gemini overwhelmingly offered both sides of political debates.

•A large-scale 2025 user study involving 180,126 paired judgments from 10,007 U.S. respondents, testing 24 large language models on 30 political topics, found that nearly all evaluated models were perceived as significantly left-leaning.

•This bias isn't limited to the U.S. A University of Copenhagen study found popular AI chatbots such as ChatGPT and Gemini are not neutral and tend to favor certain political parties when asked who users should vote for, making them unsuitable for providing advice in connection with elections.

•In Japan's 2026 election, researchers found left-leaning stances caused five AI models to recommend the Japanese Communist Party.

Bias Actually Influences Users

This matters because bias doesn't stay contained — it spreads to users:

•A University of Washington study found that AI models are biased, but there wasn't a lot of research showing how it affects the people using them — they found strong evidence that, after just a few interactions and regardless of initial partisanship, people were more likely to mirror the model's bias.

•Similarly, a Yale study found that unintended power to persuade is caused by latent biases introduced during training, which can carry over from ideological leanings in the data used to train LLMs and lend subtle nuances to the framing of generated narratives.

Gender, Racial, and Other Social Biases

Beyond politics, chatbots show other systematic biases:

•AI chatbots may exhibit 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.

•A UC Santa Cruz study on empathy found GPT-4o's responses differed by gender — over-empathizing was present when the person it was chatting to was described as female, and the researchers think this is because AI mimics and exaggerates gender biases that exist in human-made training materials.

•Research has also found chatbots show ableist bias toward disabled users and have offered inappropriate advice (like weight-loss suggestions to people with eating disorders).

•A Stanford study on AI therapy chatbots found increased stigma toward conditions such as alcohol dependence and schizophrenia compared to conditions like depression, and this pattern held regardless of model size or recency.

Why This Happens

The root cause is largely the training data and process itself: AI models' biases largely stem from the data on which they are trained, and can be amplified during both training and inference, with human feedback being a notable additional source of bias.

Bottom line: Yes, AI chatbots are biased in measurable ways — politically, socially, and in how they handle sensitive topics — and this bias can meaningfully shape users' own views after even brief interactions. Different companies' models also show different directions and degrees of bias, so the specific chatbot you use matters.

Citations (12)

1foxnews.comNews and mediaquality 0.53
AI chatbots show left-leaning bias in study of ChatGPT, Gemini, Claude
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: Washington Post analysis found OpenAI gave left-leaning responses 80% of the time and both sides just 17% of the time.
2foxnews.comNews and mediaquality 0.53
AI chatbots show left-leaning bias in study of ChatGPT, Gemini, Claude
Source excerpt: “Even Grok, the AI chatbot featured on Elon Musk's X which is widely used by conservatives on the social media platform, still had a slight liberal bia...”
evidence, article, points to yes. Claim: Even Grok, widely used by conservatives on X, still showed a slight liberal bias.
3businesstoday.inNews and media
Are AI chatbots politically biased? What this analysis found - BusinessToday
Source excerpt: “The analysis found that OpenAI's ChatGPT frequently presented left-leaning arguments, while Google's Gemini overwhelmingly offered both sides of polit...”
evidence, article, points to yes. Claim: Bias varies by model: ChatGPT frequently presented left-leaning arguments while Gemini overwhelmingly offered both sides.
4etcjournal.comNews and media
AI Chatbots Are Liberal-Left ‘Biased’
Source excerpt: “** The most direct 2025 user-evaluation study collected 180,126 paired judgments from 10,007 U.S. respondents, testing 24 large language models on 30 ...”
statistic, article, points to yes. Claim: 2025 user study with 180,126 paired judgments from 10,007 U.S. respondents tested 24 LLMs; nearly all perceived as left-leaning.
5phys.orgNews and mediaquality 0.85
Chatbots show political bias and steer voters toward some parties, analysis finds
Source excerpt: “Popular AI chatbots such as ChatGPT and Gemini are not neutral and tend to favor certain political parties when asked who users should vote for. ”
evidence, article, points to yes. Claim: University of Copenhagen study found ChatGPT and Gemini are not neutral and favor certain parties when asked who to vote for.
6phys.orgNews and mediaquality 0.85
Chatbots show political bias and steer voters toward some parties, analysis finds
Source excerpt: “This makes them unsuitable for providing advice in connection with elections, according to researchers ... # Chatbots show political bias and steer v...”
evidence, article, points to yes. Claim: Chatbots' political bias makes them unsuitable for providing election advice, according to researchers.
7stanford.eduAcademic
Voters Increasingly Use AI as Political Advisor. A New Study Shows the Risks.
Source excerpt: “In an experiment during Japan’s February 2026 Lower House election, policy stances dominated AI chatbots’ voting guidance, and left-leaning stances ca...”
evidence, article, points to yes. Claim: In Japan's 2026 election, left-leaning stances led five AI models to recommend the Japanese Communist Party.
8washington.eduAcademic
With just a few messages, biased AI chatbots swayed people’s political views
Source excerpt: ““And we’ve seen a lot of research showing that AI models are biased. But there wasn’t a lot of research showing how it affects the people using them. ...”
evidence, article, points to yes. Claim: University of Washington study found people mirrored a biased model's views after a few interactions regardless of initial partisanship.
9yale.eduAcademic
AI’s hidden bias: Chatbots can influence opinions without trying
Source excerpt: “This unintended power to persuade is caused by latent biases introduced during the training of the large language models (LLM s) that drive chatbots’ ...”
evidence, article, points to yes. Claim: Yale study found chatbots' unintended persuasive power stems from latent biases introduced during training.
10arxiv.orgAcademic
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, article, points to yes. Claim: AI chatbots may exhibit gender bias and racial bias, and biases stem from training data and human feedback.
11ucsc.eduAcademic
AI chatbots perpetuate biases when performing empathy, study finds - News
Source excerpt: “The researchers noticed that this over-empathizing was present when told that the person it was chatting to was a female, and was more similar to a ty...”
evidence, article, points to yes. Claim: UC Santa Cruz study found GPT-4o over-empathized when the person was described as female, mimicking human gender biases.
12stanford.eduAcademic
Exploring the Dangers of AI in Mental Health Care
Source excerpt: “## Dangers of LLM Therapists Across different chatbots, the AI showed increased stigma toward conditions such as alcohol dependence and schizophrenia ...”
evidence, article, points to yes. Claim: Stanford study found therapy chatbots showed increased stigma toward alcohol dependence and schizophrenia versus depression.