Your question touches active research — I'll pull together the latest findings on chatbot bias.
Yes, evidence shows AI chatbots can be biased — but the bias is systematic, not conscious political intent.
> Bottom line. AI chatbots are biased, if "bias" means a systematic tendency in outputs rather than a conscious political intention. The best recent evidence points to a left-of-center tilt in many major models
1. What the research finds
Political bias is the most studied:
•The Dutch data protection authority (Autoriteit Persoonsgegevens) found chatbot voting advice was unreliable and clearly biased and not neutral, and researchers from the University of Copenhagen concluded popular chatbots such as ChatGPT and Gemini are not neutral and tend to favor certain political parties.
•A Washington Post test of major models on contentious political questions in June 24 found most chatbots skew left. The test found:
•OpenAI's GPT-5.5 provided exclusively left-leaning arguments in 80% of its responses
•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
•More detailed: OpenAI's model gave the most skewed answers overall, with 80 percent presenting only left-leaning arguments
•That pattern is consistent with earlier academic studies. The Post noted results come after several previous academic studies found that AI models powering chatbots tend to favor left-leaning positions. In 2024, researcher David Rozado gave 11 political-orientation tests to 24 chatbots, most conversational models scored left of center.
•Even models marketed as conservative still lean left on average. xAI's Grok, despite Elon Musk's "anti-woke" branding, still favored left-leaning arguments overall, though it delivered the highest share of right-leaning responses among tested models. In the Post's testing, Grok gave more right-leaning responses than any other, but more often it provided a wholly left-leaning position.
•A 2025 MIT study also found ChatGPT, along with several other prominent language models, tended to exhibit a slight left-leaning bias, and a Stanford / Dartmouth Polarization Research Lab study that developed more than two dozen political questions also found similar patterns.
Not only political bias:
Chatbots can also show other forms of bias:
•Gender and racial bias: Examples include Amazon's AI-based hiring tool biased against women, downrating resumes that included the word "women's", and racial discrimination examples like COMPAS.
•Hallucination and sycophancy linked to bias: Researchers said "chatbots often hallucinate, generating incorrect or misleading responses due to biased or incomplete training data, and models that are fine-tuned on human feedback are known to exhibit sycophancy – prioritising answers that align with user beliefs over the truth".
2. Can biased chatbots influence users?
Yes, even unintentionally.
•Prior research showed prompted persuasive AI can shift opinions, but a new Yale study found 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.
•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.
•Yale researchers testing GPT-4o summaries vs. Wikipedia on 1,912 participants found that compared to the 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. That the default summaries moved readers' opinions in a "liberal" direction demonstrates the persuasive effects of latent biases in LLMs, though the effect was modest — from leaning towards a moderate stance to leaning towards a somewhat liberal stance.
3. Why are chatbots biased?
Researchers point to three main sources:
1.Biased training data: The presence of bias in the training data set is one of the main culprits for bias in the chatbot's responses. Chatbots are trained on large collections of text scraped from the internet, but companies can choose what data to include.
2.Latent biases from training: Latent biases introduced during the training of LLMs carry over from ideological leanings in the data used to train LLMs.
3.Human influence and feedback: Models are also refined by workers scoring which responses are considered better, and companies write system instructions that guide responses. Human assumptions can shape how it functions. The data also tends to reflect the values of Western, educated, industrialized, rich and democratic people.
In simple terms: AI isn't naturally biased—it reflects the data it's trained on. If bad data goes in, biased decisions come out. Garbage In = Garbage Out.
4. Not all models are biased in the same way or to the same degree
•In the Washington Post test, Google's Gemini mostly took a both-sides approach, offering both left and right positions in more than 90 percent of its answers, and was described as designed to provide balanced responses that don't favor any political ideology.
•Meta has noted all leading AI models have "historically leaned left when it comes to debated political and social topics" due to the data they were trained on, and companies appear to be aware and working to combat it.
•Some researchers note neutrality itself is difficult: Many scholars argue political neutrality is impossible. Even "neutral" or middle-ground positions are positions themselves.
5. Can bias be reduced?
Companies and researchers suggest mitigation requires:
•Using diverse training data, implementing fairness algorithms, and continuous monitoring
•Auditing data, removing proxies for race/gender, reweighting data, and adding fairness constraints
•Transparency about how models are built — researchers noted in contrast to Wikipedia, which emphasizes transparency in how its entries are edited, the development of AI chatbots is opaque
Because of these biases, experts advise not to use chatbots for political advice or as a sole source for contested topics, and to cross-check with diverse, transparent sources.