[fragmede]

The OpenAI paper on hallucinations gives actual technical reasons for them, if you're interested.

https://openai.com/index/why-language-models-hallucinate/

https://arxiv.org/abs/2509.04664