Skip to content
Safety

Hallucination

When an AI model generates information that sounds plausible but is factually incorrect or entirely fabricated. Common with statistics, citations, and historical claims.

Why it matters

This is the single biggest reason not to trust AI blindly. A model can invent a court case, a research paper, a quote, or a statistic and present it in perfectly confident, professional language. People have gotten into real trouble by pasting fabricated citations into legal filings or reports. Knowing hallucinations happen turns you into a careful user who verifies anything that actually matters.

A concrete example

Ask an AI for "three studies proving X" and it may hand you three official-sounding titles, authors, and journal names that don't exist. The wording looks flawless, which is exactly the trap. Before you rely on any fact, name, number, or link an AI gives you, check it against a real source. Treat AI output as a confident first draft, never as a verified reference.

How to use it

Sort your requests by what a wrong answer would cost. Brainstorming, rewriting and summarising something you can see are low risk, because you would notice a mistake. Facts, figures, citations, quotations, names, dates, legal rules and anything you will repeat to someone else are high risk and need checking against a real source before you use them. That single distinction prevents most of the damage people come to regret.

The common mistake

Asking the model whether it is sure. It will often apologise and produce a different answer with equal confidence, which feels like verification and is not. Confidence is generated text like everything else. Check against a source outside the model.

Related terms

Put Hallucination into practice

Access 750+ AI models and 60+ tools through Vincony — start free with 100 credits.