How to Spot and Avoid AI Hallucinations

An AI "hallucination" is when a model produces information that sounds right but is false: a made-up statistic, a fake quote, a citation to a paper that doesn't exist, or an incorrect step in instructions. All current AI assistants can do this.

Why it happens

Language models generate text by predicting likely words based on patterns in their training data. They don't look facts up in a database unless they're using a search or document feature, and even then they can misread or misreport. They're optimised to produce fluent answers, so wrong answers often sound just as confident as right ones.

Warning signs

• Very specific numbers or dates with no source.
• Quotes attributed to real people.
• Citations – especially academic references with authors, titles and page numbers.
• Niche or local information: small businesses, local laws, recent events.
• Recent events after the model's training cutoff.
• Answers that perfectly confirm what you hoped.
• Inconsistency when you ask the same question twice.

A five-minute fact-check routine

1. Identify the claims that matter. You don't need to check everything – just what you'll rely on.
2. Ask for sources, then open them. Does the page exist? Does it say the same thing?
3. Find a primary source: the official website, original study, government page or legislation.
4. Cross-check with a second reliable source.
5. Check dates on any statistics, prices or rules.