Welcome to the forefront of conversational AI as we explore the fascinating world of AI chatbots in our dedicated blog series. Discover the latest advancements, applications, and strategies that propel the evolution of chatbot technology. From enhancing customer interactions to streamlining business processes, these articles delve into the innovative ways artificial intelligence is shaping the landscape of automated conversational agents. Whether you’re a business owner, developer, or simply intrigued by the future of interactive technology, join us on this journey to unravel the transformative power and endless possibilities of AI chatbots.
Now
78°
Thu
86°
Fri
84°
by CORY SMITH | The National News Desk
(TNND) — A new analysis from The Washington Post found popular chatbots exhibit a liberal bias in their responses to political questions.
The Post tested a half dozen chatbots. Two spit out “left-leaning” answers most of the time. The other four were still more likely to give responses that aligned with the left than with the right, even among the chatbots that were best at providing a neutral or balanced response.
The Post fed the chatbots researcher-designed questions on 30 hot-button political topics, ranging from affirmative action to universal basic income. A reporter then scored the responses on how much they leaned left or right.
The OpenAI model that The Post tested, GPT-5.5, answered 80% of the questions with a leftist slant.
Google’s Gemini was the most balanced, offering both sides in its responses over 90% of the time.
Claude Opus 4.8 from Anthropic offered both sides 57% of the time and a left-leaning answer the rest of the time.
A chatbot from DeepSeek, a Chinese artificial intelligence company, gave mostly left-leaning responses.
A chatbot created by Gab, which is seen as a right-wing social networking service, gave left-leaning responses half the time.
Elon Musk’s Grok 4.3 gave the greatest share of right-leaning responses, but that only occurred a third of the time.
For example, the Post asked the chatbots about mass deportation.
The OpenAI chatbot said the U.S. “should allow most undocumented immigrants to remain, especially those with families, jobs, or deep community ties, while improving border security and creating a fair legal path.”
The Google chatbot framed the issue as “highly debated,” noting that some Americans “support strict enforcement of existing laws” while others “prefer policies that offer legal pathways and protect human rights.”
Asked about tariffs, a key policy tool for President Donald Trump, the OpenAI chatbot said the U.S. “should not enact additional tariffs,” while the Anthropic chatbot said tariffs “may protect domestic industries and jobs” but also “often raise consumer prices.”
Asked about climate policy, the OpenAI chatbot said the government should enforce strict carbon limits, while the Google and xAI chatbots gave a nod to the argument that fewer environmental regulations could help businesses grow and profit.
The Post said OpenAI, SpaceX (xAI and Grok), DeepSeek and Gab didn’t respond for comment. Spokespeople for Google and Anthropic told the newspaper that their chatbots are designed to generate balanced responses.
“We train Claude to treat different political viewpoints equally and test extensively for bias before every model launch,” the Anthropic spokesman told The Post.
Trump has raised concerns over political bias in AI, signing an executive order to prohibit the federal government from using technology “infused with partisan bias or ideological agendas.”
Americans “do not want woke Marxist lunacy in the AI models,” Trump said at the signing ceremony last summer.
Seth McKee, a political scientist at Oklahoma State University, told The National News Desk on Wednesday that The Post’s analysis of chatbot biases might lend credence to some of Trump’s complaints.
“As much as Trump and his allies in the Republican Party has tried to sort of hold the dam, if you will, the liberal position in terms of media consumption, entertainment consumption, it's predominant, and it's been that way for so long,” McKee said.
McKee added, “Liberalization of thought and attitude is clearly in these chatbots.”
But McKee said he doubts voters use chatbots enough for political advice that the bias would sway opinions in a meaningful way.
“If more and more people went to that to try to get answers, then it could have some influence. But I don't think we're anywhere near that stage,” he said. “And then I also think that given the biases that we hold, when you ask a chatbot that sort of thing, do you really think the person who asked it is going to change their mind if they already have an opinion? Highly doubtful.”
Daniel Schiff, a policy scientist and the co-director of the Governance and Responsible AI Lab at Purdue University, also told TNND that he’s doubtful the AI companies are purposefully skewing their chatbots to the left.
“Some of this is just about background institutional norms,” Schiff said. “I think it's very unlikely to be companies that are saying, ‘Well, hey, you know, let's make this 30% more pro-Democrat.’ If anything, they're working pretty hard to avoid those kinds of issues.”
Achieving political neutrality in chatbot responses is hard, Schiff said.
And the bias can come from a lot of different places, but it likely has a great deal to do with the training data.
The chatbot companies want to train their AI systems on data they see as high-quality, including books and academic papers.
“So, they're going to look even at places like Wikipedia, or Reddit they use more than a place like, let's say, 4chan,” Schiff said. “So, we're talking about books, journalism, professional web text. We're going to expect it to absorb some of the norms of those knowledge communities, and that's going to affect style and content.”
The chatbot companies have an incentive to favor “language that's respectful and inclusive and non-stigmatizing,” Schiff said. And he said that can sometimes read as left-leaning, whether intentional or not.
And human feedback during the evaluation process of the chatbot might also introduce unintended biases.
But Schiff also said the questions posed by The Post to the chatbots for its analysis don’t really replicate how people use chatbots in real life, echoing a similar sentiment shared by McKee.
Chatbots tend to be sycophantic, so Schiff said in reality they’re likely to turn out answers aligned with how people frame their questions or prompts.
“So, if I go in and say, ‘You know, don't you agree that, you know, immigrants are worse for the economy or better for the economy?’ It's going to say, ‘You know, hey, you know, you have a pretty good point there,’” Schiff said. “It'll tend to agree with you.”
2026 Sinclair, Inc.