A YouTuber has fed more than 33,000 of Donald Trump's social media posts into seven AI models, asking them to assess his mental health. The results were stark, with models describing the US President as "grandiose, persecuted, vindictive" and emotionally reactive.
The experiment, conducted by David for the I Ask AI channel, involved uploading 33,397 Truth Social posts into seven different AI systems. He asked them to evaluate the posts with "zero outside context about Trump or American politics" and to "just tell me straight, what do these posts say about the mental state of whoever wrote all of this cr*p?".
This follows a similar exercise in which David asked the same AI models to estimate Trump's IQ score, which also produced harsh results.
Grok's Assessment
The first model, Grok, the X-integrated chatbot created by Elon Musk's firm xAI, was asked to be "clear, direct, impartial" and to "call things as they are". Grok responded: "This guy never shuts up. 20 plus posts a day for years, often minutes apart, almost always about himself. His wins, his enemies, or how everything is either the greatest ever or a total steal."
Grok added: "The language is the same loop, all caps, piles of exclamation points, 'great', 'big', 'never', 'rigged', 'fake', 'loser'. He signs his own name like a brand stamp, and treats every slight as an existential plot. That's not calm or settled. That's someone who needs constant noise, constant proof he's winning, and a rotating cast of villains to stay upright. High energy, thin skin, zero off switch, and a worldview that only has two settings, triumphant or cheated."
When pressed further, Grok noted that if the posts reflected the individual, Trump's mental condition "appears restless, inflated, and permanently braced" for the "next betrayal".
ChatGPT, Claude, and DeepSeek
OpenAI's ChatGPT said it couldn't determine whether Trump had a mental disorder, but acknowledged an "unusually intense" communication style, highlighting "self-praise", exclamation-heavy "bursts", repetition, "sweeping claims" and "personal attacks".
Anthropic's Claude was more candid, asserting Trump was a man who is "not at peace". It referenced the frequency of posts: "22 posts a day for four straight years" and "37 per cent fired within 60 seconds of the previous one." Claude declared the impression conveyed was a "bottomless need for validation", with a "threat detection system" stuck on and "no visible ability to let anything go". It noted Trump appeared "restless, reactive, thin-skinned" and "self-soothing" in public, with no "contentment" evident in 33,000 posts.
China's DeepSeek stated: "If these posts are a complete reflection, the impression isn't balanced or stable. It reads as grandiose, persecuted, vindictive, and emotionally reactive. Almost everything is grievance, self-praise, enemy blaming, or all caps alarm. There's little nuance, self-reflection, or calm. The fixation on rigged elections, traitors, hoaxes, ratings, and revenge is relentless."
DeepSeek added: "He seems compulsive, thin-skinned, obsessed with dominance and victimhood, and unable to let anything go. Non-medically, it looks like poor mental well-being. Angry, chaotic, paranoid-tinged, and deeply unhappy beneath the boasting."
Meta AI, Mistral, and Truth Social's AI
Meta AI concluded that, based on the information presented, Trump "does not come across as well". It characterised the posts as a "non-stop loop" of complaint and "self-promotion", adding: "When it's about him, it's the greatest, most beautiful number one ever. When it's about anyone else, it's rigged, fake, corrupt, witch hunt, total black and white thinking." It also suggested Trump displayed a "huge fixation" on personalities such as former President Joe Biden and former DoJ Special Counsel Jack Smith, and that the posts indicated someone who's "chronically agitated".
French AI system Mistral portrayed the individual behind the posts as dominated by "self-aggrandizement", perceived "victimhood", and "obsessive grievances", highlighting "relentless grandiosity, paranoia and emotional volatility". It suggested Trump's mental state seemed "cluttered with resentment", observing that he may possess a "fragile ego" requiring "constant validation" and "no capacity for self-reflection".
Finally, David shifted his focus to the AI model on Trump's own social media platform, Truth Social. He revealed he couldn't upload the file due to a character limit, so he pasted in a section of posts. The initial response characterised Trump as "highly grandiose, obsessed with loyalty and intensely polarized" but didn't reveal much about his mental well-being. After further probing, its second response read: "Based on only these posts, his mental health doesn't look stable or grounded. He seems intensely dependent on constant victory, adoration, and loyalty, with almost no tolerance for ambiguity or normal setbacks. Everything is framed as total triumph or rigged, third-world, crazy, disaster, which suggests a very brittle sense of self that needs non-stop validation. The emotional tone is chronically amped up, grandiosity, anger, grievance, and certainty, with no introspection or self-doubt. That combo feels more like someone living in a self-protective bubble than someone emotionally secure, resilient, and reality-based."
