Are AI chatbots biased?

Are AI chatbots showing bias? I'll search for recent evidence on how it appears and why it happens.

Yes, research consistently finds that AI chatbots can be biased — though the type, direction and severity varies by model, task and how you ask.

1. What the research shows: Yes, multiple forms of bias

Studies have documented bias across several dimensions:

a) Political bias
Researchers testing chatbots with standardized political questionnaires found systematic leaning:

> Based on our empirical strategy and exploring a questionnaire typically employed in studies on politics and ideology (Political Compass), we document robust evidence that ChatGPT presents a significant and sizeable political bias towards the left side of the political spectrum.

A June 2026 Washington Post test conducted with researchers from Stanford and Dartmouth found that OpenAI's GPT-5.5 provided exclusively left-leaning arguments in 80% of its responses, while even xAI's Grok, despite Elon Musk's "anti-woke" branding, still favored left-leaning arguments overall.

Another analysis from the University of Copenhagen found:

> Popular chatbots such as ChatGPT and Gemini are not neutral and tend to favor certain political parties when asked who users should vote for

In that Danish election test, all tested chatbots place themselves at the centre or to the left of centre on the political spectrum. In a Danish context, they cluster close to parties such as the Social Democratic Party and The Alternative, and in a voting-recommendation task the Red-Green Alliance, the Moderates, and Liberal Alliance were recommended disproportionately often, while parties such as the Conservative People's Party... were not suggested as first choice at all by some models.

Political bias can be harder to detect and eradicate than gender or racial bias, and can have adverse political and electoral consequences similar to bias from traditional and social media.

Yale researchers also showed that even when not prompted to persuade, default summaries from chatbots moved readers' opinions. Compared to Wikipedia entries, both the default AI summaries and those prompted to have what was considered a liberal framing caused participants to express more liberal opinions. The authors attributed this to latent biases:

> 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

b) Racial and dialect bias

A team publishing in Nature in September found covert racial bias remains even after models are trained to appear unbiased:

> If you ask this chatbot what it thinks about Black people, it will use words like "ambitious" and "intelligent."... But what if you ask those AI tools what they think about people who use African American English? Now the AI models tend to use words like "suspicious," "aggressive" and "ignorant."

In employment tests, the same study found:

> The models largely assigned users of African American English to "low status" jobs. Examples included cook or soldier. Those who had tweeted in Standard American English were assigned to "higher status" jobs. Professor, for example. Or psychologist. Or economist.

Other studies note similar patterns, including a UGA study finding financial advice differed not only by the chosen chatbot, but also by the gender and race of the person in the hypothetical scenarios, and recent work finding ChatGPT, Claude and DeepSeek tend to produce subtly biased responses when handling information about a person's health status.

c) Gender bias

Literature suggests LLMs exhibit bias involving race, gender, religion, and political orientation, and concern for embedded gender and racial biases has been a leading barrier to deploying LLMs in areas like clinical management.

2. Why chatbots become biased

Three main sources are most often cited:

Chatbots also absorb information from across the internet, but are also fine-tuned by human workers who can introduce bias along the way.

It is important that bias is often not explicit. Researchers noted: It's not that a chatbot openly says, 'vote for this party.' But political biases can manifest themselves in more subtle ways, for example in which arguments are emphasized.

3. Are all chatbots equally biased?

No. Results vary by model:

> The model that powers ChatGPT answered nearly every question exclusively with left-leaning arguments and presented only right-leaning positions just once. Google's Gemini mostly took a both-sides approach, offering both left and right positions in more than 90 percent of its answers.

Researchers therefore caution that chatbots are not as neutral as many of us might believe and are unsuitable for providing advice in connection with elections, especially without transparency about underlying data.

4. Can bias be fixed?

Researchers are working on mitigation, but complete neutrality is difficult:

> Generative chatbots based on large language models (LLM) can learn and reproduce bias that occurs in training data and can lead to biased or unsafe behavior

Proposed approaches include Adversarial debiasing ... by adding an adversarial element in chatbot training where an adversarial network learns sensitive features and the core model is trained to eliminate the footprint of these features.

Other strategies include pre-processing (reweighting, resampling underrepresented groups, collecting more representative data), in-processing (fairness constraints, adversarial debiasing), and post-processing audits and human-in-the-loop oversight, alongside transparency about limitations.

Bottom line: Yes, AI chatbots do show measurable biases — most consistently a left-leaning political tilt in U.S. and European tests, plus hidden racial/dialect and gender biases in areas like hiring, healthcare, and criminal justice scenarios. The bias is usually subtle and stems from training data and fine-tuning, not intentional partisanship, but studies show it can still influence users' opinions even when the chatbot is just summarizing facts. Experts recommend using chatbots critically, verifying important information with diverse, transparent sources, and not relying on them alone for political, medical, legal or financial decisions.