An LLM picks likely next words; it does not check facts. So it sometimes invents books, numbers or quotes — very convincingly.
To reduce it: give reliable sources with RAG, instruct it to say “I don’t know”, and have people check important answers.
Example
For legal or medical answers, always show sources and verify.
বাংলায়: হ্যালুসিনেশন — AI যখন আত্মবিশ্বাসের সাথে ভুল বা বানানো তথ্য বলে।