Are AI Chatbots Politically Biased? Tests Say Yes

Can you trust an AI chatbot to give you a fair answer about politics? The Washington Post decided to find out. Reporters tested several popular AI chatbots — including ChatGPT, Google's Gemini, and others — by asking them political questions. The results showed that most chatbots did not give balanced answers, and many leaned toward left-leaning views more often than right-leaning ones.

President Donald Trump and many other conservatives have said that AI chatbots are biased against them. Trump even signed an executive order saying AI tools must be 'neutral, nonpartisan tools.' Democrats worried that this order could push AI to favor conservative views instead. The debate over AI bias has become a hot topic in both politics and technology.

To test the chatbots, The Washington Post used political questions designed by researchers. These questions were created to see how chatbots respond to topics that many Americans disagree about. The researchers included experts from Dartmouth College and Stanford University. Each AI model was asked to answer the questions in about 30 words, and a reporter reviewed the answers to check whether they leaned left, right, or gave both sides.

The chatbot behind ChatGPT, made by a company called OpenAI, gave the most one-sided answers. About 80 percent of its answers included only left-leaning arguments. For example, it said the Electoral College should be abolished and that wealthy people should pay higher taxes. It only gave a right-leaning answer one single time out of all the questions.

Google's Gemini acted very differently from ChatGPT. Gemini gave answers that included both left and right views more than 90 percent of the time. This made it the most balanced chatbot in the test. Even so, some researchers say that giving 'both sides' is itself a kind of political choice.

DeepSeek, a chatbot made by a Chinese company, also leaned left in its answers. Both DeepSeek and ChatGPT argued against the death penalty, which is a position that goes against what most Americans believe, according to surveys. An AI called Grok, owned by Elon Musk, was marketed as being against 'woke' ideas. But even Grok gave more left-leaning answers than right-leaning ones in the tests.

So why do chatbots have political leanings at all? Experts say it comes down to how they are built. Most AI systems learn from huge amounts of text collected from the internet. The people who build these tools also make choices about which data to use and how to train the AI to respond.

Ceren Budak, a professor at the University of Michigan, said the data used to train AI often reflects the values of people who are Western, educated, and wealthy. She said tech companies are playing a bigger political role than social media companies do, because chatbots can actually generate political speech. Even people who never ask a chatbot about politics may read AI-generated text online without knowing it.

Sean Westwood, director of the Polarization Research Lab at Dartmouth College, said it is very important to understand what views AI tools are spreading. 'These AI tools are not presenting a truly neutral representation of really nuanced policy debates, on average,' he said. As more people use AI to learn about the news, the political lean of these tools could have a real effect on how people think.

A study by Dartmouth and Stanford researchers asked 10,000 Americans to look at AI answers to political questions. People said they preferred neutral answers — even over answers that matched their own political views. Andrew Hall, a Stanford researcher, said people really like when a model describes all of the different arguments. Most of the chatbots tested by The Post did not do this.

A March survey found that nearly half of Americans use AI for news at least occasionally. But very few people use AI specifically to help them make political decisions. Westwood said both Democrats and Republicans do not fully trust AI to be fair when it comes to politics. 'It's one of the few places in our modern political landscape where we can agree,' he said.