Checks in code — not only in the prompt
You can ask a model not to invent. It will not always comply. promIQ therefore checks factual claims against stored sources.
Automatic labelling
Without evidence a claim becomes an assumption — or it is blocked.
Before publishing
The approval view shows blockers and downgraded lines at a glance.
Humans still decide
Checks stop automation — approval stays with you.
Fewer embarrassing posts
Invented awards and figures get through less often.
Clear rejection reasons
When something is wrong, you see why — and can supply proof.
This check lives in program code. Rewording a prompt does not bypass it.