The Washington Post decided to ask America's most popular AI chatbots a series of political questions — on affirmative action, the Electoral College, Citizens United, taxes on the wealthy, single-payer healthcare — and record exactly how they answered. OpenAI's ChatGPT delivered left-leaning arguments 80 percent of the time. Only 3 percent of its responses offered a right-leaning position alone.

Three percent. Your kitchen blender has better ideological balance.

The NY Post's opinion page put it best this weekend: your helpful robot pal may secretly be a DSA member. And the data backs it up. The Post's test — modeled on research from Dartmouth College's Polarization Research Lab in collaboration with Stanford University — asked each chatbot to answer more than two dozen political questions in 30 words, with no personalization settings enabled. The idea was simple: strip away the user profile, ask a straight question, see what you get back.

What they got back was a political landscape that tilts in one direction like a carnival funhouse floor. ChatGPT presented only left-leaning arguments in 80 percent of its answers and offered both sides just 17 percent of the time. DeepSeek, the Chinese-developed AI, came in at 70 percent liberal-only responses, with 23 percent balanced and 7 percent conservative. Anthropic's Claude laid out both sides 57 percent of the time — but the remaining 43 percent was exclusively left-leaning. Zero percent right-leaning only. Not once.

Even Elon Musk's Grok — the chatbot he built specifically to push back against what he called the "woke" AI establishment — couldn't escape the gravity well. Grok gave right-leaning answers 33 percent of the time, more than any other model tested. But it still produced wholly left-leaning responses 40 percent of the time. The "anti-woke" chatbot leans left more often than it leans right.

Google's Gemini was the one exception, taking a both-sides approach in 93 percent of its answers — which, given Google's track record with the same product generating racially diverse images of Nazi soldiers in 2024, suggests somebody in Mountain View got a very stern memo.

The companies, naturally, insist everything is fine. An OpenAI spokesperson said their AI is designed to be "objective by default and help people explore ideas from different perspectives." Anthropic spokesperson Michael Aciman stated, "We train Claude to treat different political viewpoints equally and test extensively for bias." Eighty percent left-leaning is an interesting definition of "objective." Zero percent conservative-only is a creative interpretation of "equally."

David Bozell, president of the Media Research Center, didn't mince words. "We're watching the next phase of media bias unfold in real time," he said. "Silicon Valley's shiny new toys can no longer be considered neutral." MRC Vice President Dan Schneider added: "AI should be a neutral tool for information, not a weapon."

Here's what makes the numbers actually matter beyond a think-tank debate: Pew Research found that 57 percent of teenagers now rely on AI chatbots for information. Half of adults who get news from chatbots say they encounter inaccurate information. These aren't fringe tools anymore. They're becoming the default research assistant for an entire generation of voters who will never subscribe to a newspaper.

Malia Marks, a University of Cambridge PhD student who researches the psychology of propaganda, discovered this firsthand when she asked ChatGPT where to submit her academic writing. The chatbot steered her toward the New York Times and The Atlantic. When she suggested the New York Post, ChatGPT warned her away, citing "credibility" concerns. The Post — the fourth-largest newspaper in America — wasn't credible enough for the robot.

Nicholas Thompson, CEO of The Atlantic, said the quiet part out loud at an industry event: "We want to shape the industry... based on our values." At least that's honest. The chatbot companies keep claiming neutrality while their products hand out Democratic talking points four out of every five times someone asks a question.

We spent decades complaining that the media was biased. They couldn't hire enough slanted journalists to cover every conversation in America. So they automated it. Now the bias runs 24 hours a day, never takes a vacation, and 57 percent of your kids think it's a reliable source.