Every team using AI to draft, research, or summarize has shipped a citation that looked authoritative and pointed at nothing. The journal is real. The authors are plausible. The DOI has the right shape. The paper does not exist.

This session explains the mechanism behind fabricated citations, why prompt engineering fails as a control for it, and what verification looks like when it is treated as an engineering problem rather than a wording problem. Worked examples throughout, including a live check against a citation graph that separates claims a human has signed from claims a machine inferred.

Attendees leave with a verification workflow they can apply the same week, a clear sense of which failure modes are fixable at the prompt and which require tooling, and the questions to ask any vendor claiming their product is grounded.

beginner to intermediate · 40 min