← Back to articlesAI basics

How do you spot an AI hallucination?

An AI hallucination is plausible content without adequate grounding.

The short answer

Check names, papers, links and numbers especially carefully. Agreement from another model is not a substitute for original evidence.

Try it in three steps

  • Select three checkable factual claims from an answer.
  • Ask for a supporting source passage for each.
  • Remove unsupported claims or label them unverified.

A practice prompt

Before you start

Check three answer statements against the supplied practice source.

Practice material (fictional)

Source: The trial used 10 anonymous records and did not measure time saved. Answer: It processed 10 records; covered every order; saved 20 hours monthly.

Replace placeholders in the prompt with this material, or append the material if there are no placeholders. Reading time excludes doing the exercise.

Separate facts, inferences and advice. Attach a working supporting source to each fact or mark it unverified.

What to look for (not a recorded test result)

Only the first statement is supported. Remove or flag the others rather than accepting plausible-looking numbers.

If it goes wrong

A page about a related topic does not count as evidence for the specific claim.

Check your result

Facts have locatable evidence and inferences are clearly identified.

References and scope

The links below support the concepts or product navigation. The steps and exercises are our teaching suggestions, not official certification or measured customer outcomes. Product interfaces and availability may change. References checked on 2026-10-17.

Start with one task

Describe the current process and its slowest step. The initial conversation is free.

Let’s talk ↗
Need help? See small-scale implementation services ↗

Keep reading