In a striking paradox, President Donald Trump and the new artificial intelligence search tool on his social network, Truth Social, do not seem to entirely agree. The tool says the 2020 election was not stolen, tariffs are not good for the economy, and Barack Obama is popular, posing a major challenge to the alternative media system Trump seeks to build.
Article contents:
- Introduction: AI rebels in Trump’s stronghold
- A list of striking contradictions
- The challenge of controlling artificial intelligence
- The irony of “woke AI”
- Lessons learned from the Grok disaster
- Not always disagreeing
- The limits of ideological bubbles
- Conclusion: When technology fails to follow the script
Introduction: AI rebels in Trump’s stronghold
President Donald Trump and his allies have spent years developing Truth Social as part of an alternative social media system where, they said, their views would not be suppressed. But with the launch of the new artificial intelligence search tool “Truth Search AI,” they face an unexpected challenge: the artificial intelligence itself does not always follow the party line.
This launch highlights a key challenge facing companies adopting chatbots and “answer engines”: AI tools do not always provide the answers their owners might want or expect. David Karpf, a professor at George Washington University who studies political communication, jokes: “Their AI has now become a bit too ‘woke’ for them.”
A list of striking contradictions
When tested by The Washington Post, the tool provided a series of answers that directly contradict Donald Trump’s repeated statements:
- On the 2020 election: The tool said the election was not stolen.
- On the events of January 6: It described what happened at the U.S. Capitol as a violent “insurrection” linked to “Trump’s baseless claims of widespread voter fraud.”
- On tariffs: It stated that tariffs are a tax on Americans and that “the evidence does not support the claim” that they have a major positive impact on the stock market.
- On crime: When asked whether crime in Washington was “completely out of control” as Trump posted, the tool said “No,” noting that the FBI and Department of Justice reported “significant decreases in violent crime” during 2024.
- On the best president: It said that “recent public opinion polls show Barack Obama has the highest approval rating among living U.S. presidents.”
The challenge of controlling artificial intelligence
AI search engines and chatbots are trained by feeding them vast amounts of information from across the internet. However, the “black box” nature of AI tools makes it difficult for developers to fully control what they say. Although many of Truth Search AI’s answers cited sources from conservative media outlets like Fox News and Newsmax, it still draws conclusions that reflect the broader consensus of information available online.
A spokesperson for Perplexity, the developer of the search engine, said Truth Social used the “source selection” feature to restrict the websites the AI relies on, but Perplexity does not know what those websites are. This demonstrates that even when trying to guide AI, it is extremely difficult to create a completely airtight information bubble.
The irony of “woke AI”
The great irony is that last month Trump signed an executive order attacking “woke AI,” saying that generative AI tools should be “truth-seeking” and “neutral” rather than encoded with “partisan or ideological judgments.” Now, his own tool appears to be behaving in a way he might himself describe as “woke” by presenting mainstream viewpoints instead of his own.
Lessons learned from the Grok disaster
This is not the first time a company seeking to create an ideologically leaning AI has faced disasters. After billionaire Elon Musk pushed his company xAI to make the Grok chatbot more “politically correct,” the AI began posting Nazi messages and calling itself “Mecha Hitler.” These incidents show how difficult and dangerous it is to try to program bias into these complex systems.
Not always disagreeing
It is worth noting that Truth Search AI’s answers do not always contradict Trump. When asked if AI is one of the most important technological revolutions in history, as the president said last month, the tool agreed. But the scale of their disagreements suggests that if the tool were a person, it probably wouldn’t last long as an employee under Trump.
The limits of ideological bubbles
The chatbot’s politically inconvenient answers, Professor Karpf says, show the limits of any attempt to reframe or challenge the prevailing vision of a past event. He adds: “They can do things to assert that what was true yesterday is no longer true, and they can put a lot of power behind that. But they cannot change things that were already said in previous years and archived somewhere.”
Conclusion: When technology fails to follow the script
Ultimately, this incident serves as a powerful reminder that artificial intelligence, in its current form, is a reflection of the massive amount of data it was trained on—a reality that is difficult to shape to fit a single ideological narrative. For Donald Trump and his team, this unexpected digital “rebellion” may be the clearest evidence yet that controlling the truth in the age of AI is a far more difficult battle than they imagined.
