To fact-check an AI answer, identify the specific claims that matter, look for red flags such as precise numbers or citations, verify those claims in reliable original sources, cross-check with a second source, and ask the AI what it is unsure about. Never trust a citation you have not opened yourself.
Why AI answers need checking
AI assistants generate text that sounds right. Most of the time it is right, which is exactly what makes the occasional error dangerous: we stop checking. Errors are most common with specific facts: numbers, dates, names, quotations, legal and medical details, recent events, and references to books or papers.
The aim is not to distrust everything. It is to check the claims that matter, quickly.
The five-minute method
1. Pull out the claims that matter
Read the answer and underline the facts your work depends on. If you are writing an essay, these might be dates and statistics. If you are making a decision, they might be prices, rules or deadlines. Ignore general explanations you already know to be sound.
2. Look for red flags
Be extra careful with:
- Very precise numbers ("37.4% of students...") with no source.
- Citations and references, which AI sometimes invents, including realistic-looking titles and authors.
- Quotes attributed to real people.
- Recent events, if the tool is not connected to search.
- Anything about health, law or money, where mistakes have consequences.
3. Verify in an original source
Go to the source, not a summary. For school topics, use your textbook, the exam board's materials or reputable sites such as BBC Bitesize or official organisations. For statistics, find the organisation that published the data. For law or rules, use the official government or institution website.
If the AI gave a link, open it and confirm the page actually says what the AI claimed. Sometimes the link is real but the claim is not on it.
4. Cross-check with a second source
If two independent, reliable sources agree, you can be fairly confident. If they disagree, note the disagreement; that itself is useful information for an essay.
5. Ask the AI about its uncertainty
Follow up with: "Which parts of your answer are you least confident about?" or "Which of these facts should I verify?" Models often flag the weak points honestly when asked.
Prompts that reduce errors in the first place
- "If you are not sure, say so rather than guessing."
- "Use only the document I have provided and quote the relevant lines."
- "List your sources with links, and mark any you are not certain exist."
A quick checklist
| Check | Question to ask |
|---|---|
| Source | Where did this claim come from? Can I open it? |
| Date | Is this still current? |
| Agreement | Do other reliable sources say the same? |
| Expertise | Is the source an authority on this topic? |
| Stakes | What happens if this is wrong? |
Teaching children to check
Children are especially likely to trust confident answers. A simple game helps: ask an AI a question you know the answer to, and see if your child can spot when it gets something slightly wrong. It builds the healthy scepticism they need. Our AI for Kids course includes a whole lesson on this.
Frequently asked questions
Are AI tools with web search always accurate?
They are more current, but they can still misread or misreport sources. Open the links for anything important.
Can I cite ChatGPT or another AI in an essay?
Rules vary. Many institutions require you to cite original sources rather than AI, and to disclose AI use. Check your school or university's policy.
Why does AI invent references?
Language models generate plausible text. A realistic-looking reference is plausible text, even if no such paper exists.
