Are AI chatbots biased?

I'll search recent research to answer whether AI chatbots show bias.

Yes, research consistently shows AI chatbots can be biased — though the type and degree of bias varies by model, topic, and language.

1. What the research finds

Yes, but not necessarily intentionally. Studies have found systematic biases in many leading large language models (LLMs) that power chatbots:

•Political bias: Multiple academic and journalistic tests have found a consistent left-leaning bias. For example, researchers found that pretrained language models "do have political leanings that reinforce the polarization present in pretraining corpora, propagating social biases into hate speech predictions and misinformation detectors". The scientists determined that OpenAI's ChatGPT and its GPT-4 model to be the most left-leaning libertarian chatbot, while Meta's LLaMA leaned the furthest right.

•A Washington Post evaluation found that AI chatbots exhibit a left-leaning political bias, with OpenAI's GPT-5.5 answering nearly every hot-button question with left-leaning arguments and Google's Gemini 3.1 Pro offering both left and right positions in over 90% of its answers. Specifically, OpenAI's GPT-5.5 was found to provide exclusively left-leaning arguments in 80% of its responses, while Google's Gemini offered both perspectives in over 90%. Google's model stood out for its balanced approach.

The bias was also found to be language-dependent, with both ChatGPT and Gemini demonstrating liberal and leftist biases that varied depending on the language used.

Gender, Racial and Cultural biases:
Common types of bias include:

> - Cultural Bias: Responses may not resonate with users from diverse cultural backgrounds. ... - Gender Bias: The AI might reinforce traditional gender roles
> - Racial Bias: Historical prejudices can be reflected in the AI's output, potentially alienating users from marginalized communities.

Other studies have quantified this:

•In hiring tests, all models showed a strong bias for white-sounding names (85%), while Black male names were never chosen over white ones.

•Research has shown that biased chatbot responses occur in approximately 30% of interactions involving sensitive topics like race, gender, or socioeconomic status.

•A UN Women analysis released in June 2026 reviewed 133 AI systems. Forty-four percent showed clear gender bias. Twenty-six percent combined gender and racial prejudice.

Specific harms have been documented in advice-giving:

> Researchers at MIT and Stanford examined how leading large language models respond when prompted as male or female users seeking investment guidance. Chatbots consistently offered women more cautious portfolios. Men received recommendations with higher equity exposure.
> Over a simulated lifetime, that difference added up. Women ended up with about $60,000 less by age 60.

2. Why does it happen?

Chatbots are not explicitly programmed to be biased. Bias typically comes from training data:

•Bias Amplification: I am trained on massive datasets of text and code, which reflect the biases present in society.

•This unintended power to persuade is caused by latent biases introduced during the training of the large language models (LLMs) that drive chatbots' core capabilities. Those latent biases — which can carry over from ideological leanings in the data used to train LLMs — lend subtle nuances to the framing of the narratives the chatbots generate.

3. Does it influence users?

Yes, even when not intended to persuade. Prior research has shown persuading content can shift opinions, but a 2026 Yale study found:

> this study provides evidence that the same is also true of content that is not intended to change minds, such as the summaries that popular chatbots produce in response to simple queries about historical events.

The researchers noted: "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". Compared to Wikipedia entries, both the default AI summaries and those prompted to have liberal framing caused participants to express more liberal opinions, and that the default summaries moved readers' opinions in a "liberal" direction demonstrates the persuasive effects of latent biases in LLMs.

Researchers also warn that effects are modest per interaction but could compound if somebody frequently engages with chatbots for factual information.

4. Is it the same for all chatbots?

No. Studies show significant variation:

•As noted, the Washington Post and Stanford/Dartmouth research found users perceived OpenAI models as having a left-leaning slant four times greater than Google models.

•Even models marketed as conservative alternatives showed left-leaning tendencies despite branding as more neutral or conservative.

Developers are aware and trying to mitigate bias through more diverse training data, safety protocols, demographic parity testing, and offering balanced viewpoints, but prior research notes the problem is particularly acute in chatbots that haven't been trained on diverse conversation types or lack robust safety protocols.

> Bottom line: AI chatbots reflect and can amplify biases present in human society and their training data — including political, gender, racial, and cultural biases. The bias is often subtle and unintentional, but measurable and potentially influential. Cross-checking important information, asking for multiple perspectives, and using human oversight for high-stakes decisions (finance, hiring, medical, political) remains important.

Citations (3)

1ainvest.comNot yet classified
AI Chatbots Show Left-Leaning Bias, Finds Report
Backs: “The scientists determined that OpenAI's ChatGPT and its GPT-4 model to be the most left-leaning libertarian chatbot, while Meta's LLaMA leaned the furthest right. A Washington Post evaluation found that AI chatbots exhibit a left-leaning political bias, with OpenAI's GPT-5.5 answering nearly every hot-button question…”
2webpronews.comNews and media
AI Chatbots Quietly Shortchange Women in Finance and Hiring
Backs: “A UN Women analysis released in June 2026 reviewed 133 AI systems. Forty-four percent showed clear gender bias.”
3yale.eduAcademic
AI’s hidden bias: Chatbots can influence opinions without trying
Backs: “Bias Amplification: I am trained on massive datasets of text and code, which reflect the biases present in society. This unintended power to persuade is caused by latent biases introduced during the training of the large language models (LLMs) that drive chatbots' core capabilities.”