In AI terms, a hallucination is when a model confidently produces information that sounds plausible but is actually false, fabricated, or unsupported by any real source. It's one of the most important limitations to understand before relying on AI tools for factual content.
Why It Happens
Language models like ChatGPT generate text by predicting the most statistically likely next word, based on patterns learned during training — they aren't actually looking anything up or checking facts. When a model doesn't genuinely "know" an answer, it can still generate something fluent and convincing that simply isn't true.
Common Examples
- Inventing a book, study, or article that doesn't actually exist
- Fabricating statistics or dates that sound believable but are wrong
- Confidently attributing a quote to the wrong person
- Making up plausible-sounding URLs or citations
Protecting Against It
- Fact-check any specific claims, statistics, or dates before publishing AI-generated content
- Ask the AI to cite its sources, then independently verify those sources actually exist
- Be especially cautious with anything historical, statistical, or highly technical
- Use AI-generated drafts as a starting point, not a finished, publish-ready product
Hallucination is a real limitation of the technology, not a rare glitch — treating every factual claim from an AI tool with a healthy amount of scepticism is simply good practice.
« Back to Index