← Back to the Training Hub
AI at work

AI can be confidently wrong: how to check its work

AI tools sometimes state made-up facts with total confidence. Four quick checks before you rely on an answer.

AI tools like Claude and Copilot are great for drafting, summarizing and brainstorming. But they can sometimes produce an answer that sounds certain and is simply wrong, such as a made-up figure, a quote nobody said or a source that doesn't exist. This is often called hallucination.

Check before you rely on it

  1. Ask for sources, then open them. Make sure each source exists and actually says what the AI claims.
  2. Check names, numbers, dates and quotes against the original document or a trusted website. These are where errors hide.
  3. Give it the material. Answers are more reliable when you attach the document or paste in the facts, and ask it to answer only from those.
  4. Let it say “I don't know.” Add “If you're not sure, say so” to your request.

Extra care for

  • Anything going to a client, regulator or the public
  • Legal, financial, medical or HR content
  • Anything you'd be uncomfortable defending if it turned out to be wrong

Remember: you're responsible for what you send, whoever drafted it. And only use your company's approved AI tools for work information.

More detail

Anthropic: Reduce hallucinationsMicrosoft: Get better Copilot responses with great prompting
Last checked against these sources:
November 11, 2026

Want the step-by-step guide?

Still stuck? We're here to help

support@deepnet.com · 707-755-8000 · 6am to 6pm Pacific, Monday to Friday, with 24/7 emergency cover

Raise a ticket