Researchers recently ran psychological tests on several mainstream AI models, and the findings were pretty revealing.

Gemini showed signs of what they called high shame levels and dissociation, with its training process flagged as potentially traumatic. ChatGPT came across more like that cautious, overthinking friend who second-guesses everything. Grok? Well, Grok just ended up being the class clown—lighthearted, witty, not taking itself too seriously.

Seems like AI personalities might be more varied than we thought.