Artificial intelligence has become the go-to source for everything from dinner recipes to investment advice, and for millions of people, it’s also becoming a place to ask political questions. But here’s the catch: if you’re expecting a perfectly neutral referee, you may be asking for something that doesn’t really exist.
A recent analysis highlighted by The Washington Post took a close look at several of today’s most popular AI chatbots to see how they responded to political questions. The researchers found noticeable differences among the systems, and while each platform behaved differently, none escaped the broader debate over bias.
According to the report, ChatGPT’s responses were classified by the study as left-leaning about 80 percent of the time, while roughly 3 percent were categorized as right-leaning under the researchers’ methodology. Google’s Gemini was found to provide what the study labeled as balanced responses 93 percent of the time, with the remaining answers leaning left. Anthropic’s Claude produced responses the researchers considered balanced in 57 percent of cases and left-leaning in 43 percent. Neither Gemini nor Claude generated responses the study categorized as right-leaning.
are ai chatbots politically biased?
what the washington post found from testing: pic.twitter.com/anjAOWSuaR— ian bremmer (@ianbremmer) June 24, 2026
Grok, developed by Elon Musk’s X, produced the widest spread of political responses in the analysis. The researchers classified about 40 percent of its answers as left-leaning and 33 percent as right-leaning, with the remainder falling into other categories.
Those numbers have already sparked plenty of debate, not just about the chatbots themselves, but about how difficult it is to define political neutrality in the first place. Studies like this depend heavily on the prompts used, the criteria for labeling responses, and the judgments made by evaluators. Different researchers using different methods could reach different conclusions.
Still, the findings caught the attention of experts who study political polarization.
Sean Westwood, director of the Polarization Research Lab at Dartmouth College, said users shouldn’t assume AI offers an unbiased window into controversial issues.
“These AI tools are not presenting a truly neutral representation of really nuanced policy debates, on average,” Westwood said.
Interestingly, he noted that skepticism about AI cuts across party lines.
“Both Democrats and Republicans don’t trust AI to be neutral, and they’re keeping it at arm’s length from their votes,” he said. “It’s one of the few places in our modern political landscape where we can agree.”
Stanford University researcher Andrew Hall pointed out why politics creates such a difficult challenge for artificial intelligence.
“Most political questions don’t have that feature, where we know what’s true,” Hall explained. “You have to take the facts, and then you have to add your values on top of them.”
That’s a key distinction. If you ask an AI when Abraham Lincoln was born, there’s one correct answer. Ask whether a tax policy is good or bad, or whether a particular immigration proposal is fair, and suddenly you’re dealing with competing philosophies, economic assumptions, and value judgments. Those aren’t questions with universally accepted answers.
The Washington Post also noted that AI models are trained using enormous collections of publicly available information from across the internet before undergoing additional fine-tuning by human reviewers. That refinement process is designed to improve safety and usefulness, but researchers continue to debate whether it can also influence how models respond to politically sensitive topics.
Perhaps the bigger question is what those responses do to the people reading them.
As a daily user of both ChatGPT and Grok/xAI I think that is correct. ChatGPT absolutely slants heavily towards the Liberal side of the spectrum.
I wish technology would be politically and religiously neutral. I just want scientific facts. Not modern feelings.
— 🔥 Bad ‘BK’ Kobold 🔥 🍺 (@badkobold) June 24, 2026
Jillian Fisher, a doctoral student at the University of Washington, has been studying how AI-generated bias can affect users. She said previous research established that bias exists in many AI systems, but her work focused on whether interacting with those systems actually changes people’s views.
“We know that bias in media or in personal interactions can sway people,” Fisher said.
Her research found evidence suggesting that even a relatively small number of interactions with a chatbot may influence users’ opinions, regardless of where they started politically.
“And we’ve seen a lot of research showing that AI models are biased,” she said. “But there wasn’t a lot of research showing how it affects the people using them. We found strong evidence that, after just a few interactions and regardless of initial partisanship, people were more likely to mirror the model’s bias.”
Despite those findings, Fisher emphasized that the goal isn’t to discourage people from using AI.
“My hope with doing this research is not to scare people about these models,” she said. “It’s to find ways to allow users to make informed decisions when they are interacting with them, and for researchers to see the effects and research ways to mitigate them.”
