Are AI chatbots biased?

AI chatbots can exhibit various forms of bias, and this is a well-documented and actively studied issue. Let me search for the latest research on this topic to give you a comprehensive answer.## Yes, AI chatbots are biased — and this is well-documented across multiple dimensions

Political bias

Numerous studies have found measurable political leanings in popular chatbots. The model that powers ChatGPT answered nearly every question exclusively with left-leaning arguments and presented only right-leaning positions just once, while 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. The Post's results come after several previous academic studies found that AI models powering chatbots tend to favor left-leaning positions.

This matters because, as one Dartmouth researcher noted, "These AI tools are not presenting a truly neutral representation of really nuanced policy debates, on average."

Bias can actually sway your opinions

A University of Washington study demonstrated something striking: researchers recruited self-identifying Democrats and Republicans to make political decisions with help from three versions of ChatGPT—a base model, one with liberal bias, and one with conservative bias. Democrats and Republicans were both likelier to lean in the direction of the biased chatbot they were talking with than those who interacted with the base model. Even a brief interaction had an effect — if you've interacted with an AI chatbot, all AI models are biased, since they were trained on enormous corpuses of unruly data and refined through human instructions and testing, and bias can seep in anywhere.

Interestingly, awareness seems to help: participants who had higher self-reported knowledge about AI shifted their views less significantly — suggesting that education about these systems may help mitigate how much chatbots manipulate people.

A related large-scale study found chatbots can be even more persuasive than political ads, but with a troubling catch: a politically biased AI model was more effective than political advertisements at nudging both Democrats and Republicans to support presidential candidates of the opposing party, swaying opinions by citing facts and evidence—but they were not always accurate; the most persuasive models said the most untrue things.

Social and demographic bias

Beyond politics, chatbots exhibit biases rooted in their training data. The root cause of model bias lies in imbalanced social power structures within training data — research shows large language models trained on massive internet text continuously internalize biases related to gender, race, and culture. A concrete example: in occupational association analysis, the probability of association between the "nurse" occupation and the female gender label is 78%, compared to 65% for the association between "engineer" and the male group.

Language-based bias

Bias isn't just about politics or demographics — it extends to which language you use to interact with a chatbot. A Johns Hopkins study found that rather than leveling the playing field, popular large language model tools are actually building "information cocoons." Specifically, the language you use to chat with an AI determines the information you get back, AI-generated information about international conflicts is particularly biased, and American perspectives are often forced onto speakers of low-resource languages. When information is sparse in a given language, LLMs will generate answers based on information found only in higher-resource languages, ignoring other perspectives.

Why this happens

Several mechanisms contribute to chatbot bias:

The bottom line

Bias in AI chatbots is a real, measurable, and actively researched phenomenon — spanning political leanings, demographic stereotypes, and language-driven information gaps. It stems largely from training data and design choices, and because these systems are so widely used, their biases can meaningfully shape users' views, sometimes more effectively than traditional persuasion methods like political ads. Researchers generally frame this not as a reason to avoid AI tools, but as a reason for users to stay informed and critical when interacting with them.