The Never List Fixes AI Hallucinations Fast

"Be professional." "Be clear." "Be helpful." You've typed these into a prompt and gotten back exactly the kind of mush you were trying to avoid. (Turns out "be professional" means nothing to a model. It means nothing to most people either.)

Vague positive instructions are weak. Hard prohibitions are strong. So instead of telling AI what to be, give it a Never List.

Build the list

Add this to your prompt or custom instructions:

Never use jargon
Never assume knowledge
Never guess
If you are not sure say I do not know

Each line closes a specific door. No jargon means plain language. No assuming knowledge means it explains instead of name-dropping. No guessing plus "say I don't know" is the one that actually kills hallucinations, because it gives the model permission to admit a gap instead of papering over it.

"Be accurate" is a wish. "Never guess, say I don't know" is a rule. Rules win.

Write your own Never List for whatever the model keeps getting wrong. Bans are sharper than vibes. Go set the boundaries.