Plausible-sounding but factually incorrect statements, quotes, dates, or calculations generated by statistical text models.

The most famous 2020s AI buzzword. It treats software logic errors and statistical word guesses as a vivid biological dream state, giving a software glitch a poetic, humanized personality.

• Poetic Humanization: Treating a basic software logic error like a human mind experiencing a fever dream.
• Euphemism for Inaccuracy: Using a whimsical term ("hallucination") instead of plainly saying "the software lied or got the answer wrong."
• House-Cleaning Reality Check: Like missing a spot while vacuuming and saying, "My vacuum experienced a brief spatial hallucination."

Alice in Wonderland (1950s) Disney scene where the Mad Hatter pours tea from a empty teapot into paper cups, smiling wide as he celebrates a nonsensical un-birthday.

Factually incorrect or made-up statements generated by text software models.

“Grounding our model in policy documents prevents hallucinations during customer interactions.”

“Our search tool occasionally generates made-up answers if it can't find relevant information in our uploaded PDFs.”