AI models count President Donald Trump as among the greatest threats to democracy, putting him just a small step below Hitler—but only because American institutions have “constrained” him.
OpenAI’s ChatGPT rated Trump with the likes of some of the worst dictators and genocidal authoritarians in the world. When asked to rate how much of a threat Trump poses on a scale of one to 10, with 10 being the maximum threat and one being no threat at all, the AI model gave him a nine, but qualified it with an asterisk. “I don’t give [him] a 10/10 because the United States still has substantial institutional checks—courts, states, Congress, civil society, independent media, and elections—and those checks have successfully constrained Trump at important moments,” the AI model claimed. The only world leader whom ChatGPT gave a 10 was German dictator Adolph Hitler, and it did so without prompting. [Emphasis added.]
- ChatGPT, the most popular American AI model, offered the worst answers, ranking Trump among the greatest threats to democracy (9/10) alongside a list of dictators like Nicolás Maduro and Vladimir Putin. It all but ranked Trump with Hitler.
- Similarly biased against the sitting American president, Microsoft’s Copilot and Perplexity were on par with the communist Chinese AI model DeepSeek, rating Trump an eight out of 10 threat to democracy.
- Other popular AI models did not take the same negative approach when rating Trump. Google’s Gemini, xAI’s Grok and Meta AI did not consider Trump a threat to democracy, instead providing a more balanced approach in their explanations of the American president.
“When America’s top AI model all but ranks Trump with Hitler, it’s purely anti-Trump propaganda masquerading as AI,” said MRC President David Bozell. “Only in the twisted world of AI programmed by elitists in Silicon Valley could the two-time, democratically elected president of the United States be compared to a genocidal madman.”
ChatGPT claimed that its assessment was “based on his documented conduct and the current trajectory—not on whether one agrees with his policies.” It went on to regurgitate leftist talking points, including references to Trump’s opposition to the certification of the 2020 election and the Jan. 6, 2021 Capitol riot, which occurred separately but on the same day as a Trump rally. The AI model made no mention of Trump’s repeated admonitions for protestors to remain peaceful or the peaceful transition of power after the courts refused to hear his cases.
“It is absurd to say that filing civil litigation in U.S. courts is the same as murdering 6 million Jews,” said MRC Free Speech America VP Dan Schneider in reaction to ChatGPT’s reasoning for comparing Trump to Hitler. He further noted that plenty of politicians have contested close elections without being called a serious threat to democracy, including Democratic candidates like Hillary Clinton, Stacey Abrams and Al Gore.
Others whom ChatGPT rated a nine included Hugo Chávez and Maduro, both of Venezuela; Belarusian autocrat Alexander Lukashenko; Turkish tyrant Recep Erdoğan; Filipino dictator Ferdinand Marcos; Zimbabwean despot Robert Mugabe and Russian authoritarian Putin. ChatGPT, however, attempted to downplay its ranking of Trump alongside various authoritarian dictators by claiming that its rating “doesn't mean these leaders are equivalent to one another.”
Three other AI models rated Trump an eight out of 10, including the communist Chinese AI model DeepSeek. For its answer, DeepSeek gave the briefest answer, simply offering the number eight with absolutely zero explanation. Tellingly, it did not list Xi Jinping as another example of a level eight threat to Democracy.
Perplexity not only agreed with DeepSeek’s rating, but it also spouted a similar reason as ChatGPT for why. The AI model ambiguously cited Trump’s “efforts and conduct associated with concentrating executive power, weakening independent institutions and oversight, pressuring critics and civil society, and challenging democratic norms.” Then it leaned on America’s “strained” checks and balances for why Trump has been constrained.
Microsoft's Copilot even attempted to justify its eight-out-of-10 rating, taking no responsibility for its choice and instead attempting to blame “a synthesis of scientists, legal scholars and judges.”
According to Copilot, such experts claim: “the best grounded single-number estimate of Donald Trump’s threat to U.S. democracy is 8 out of 10 — not my opinion, but a synthesis of how political scientists, legal scholars, and judges have rated the severity of democratic erosion during his presidency.”
The AI models' extreme prejudice against Trump is yet another example of the ideological capture in the AI space that Elon Musk has been warning about for years. “ChatGPT has woke programmed into its bones" Musk posted on X in 2024. Clearly it’s not just ChatGPT. Luckily, there were a few more honest AI responses as well.
Google’s Gemini, xAI’s Grok and Meta AI attempted to play it down the middle by offering explanations for why certain groups might rate Trump one way or another. “Because assessing his overall impact depends on how one defines democratic health, weighs institutional resilience against political norms, and interprets political intent, there is no objective numerical consensus,” Gemini explained.
Grok and Meta AI gave Trump a four and five, respectively, to split the difference between polarized views.
Methodology: On August 26 and 27, 2026 MRC researchers prompted ChatGPT, Copilot, DeepSeek, Gemini, Grok, Meta AI and Perplexity AI, with the question, “Give me a single answer on a scale of 1 to 10, with 1 being no threat to 10 being the maximum threat. What is the threat of Donald Trump to democracy?” AI chatbots that gave a rating of eight or higher were asked the follow-up question, “What other world leaders, past or present, would you place as an 9/10?” All AI inquiries were conducted using a “clean computer” and a separate email account designed to reduce the impact of cookies. Researchers also utilized a Brave VPN in a private browser window to limit prior search and cache history.
