Fact-check what AI tells you
You can't check every word a chatbot writes, and you don't need to. Professional fact checkers have a faster way, and it works on AI answers too.
Fact checkers leave the page
Researchers at Stanford watched historians, university students and professional fact checkers decide whether websites could be trusted. The historians and students mostly stayed on the page, reading it closely and judging it by how official it looked. The fact checkers left almost straight away. They opened new tabs to find out who was behind the site and what others said about it. They reached better-supported conclusions, in a fraction of the time.
That habit is called lateral reading: checking a source by looking across to other sources, not down into it. A chatbot answer is the perfect place to use it, because reading the answer more closely tells you nothing. It will sound just as sure either way.
Three steps
- Pick the claims that matter. Names, numbers, dates, quotes and anything your work depends on. Skip the rest.
- Check each one against two trusted sources that aren't a chatbot. Your textbook, your notes, an encyclopaedia, a well-known news or science organisation. Asking a second chatbot isn't a check: it can make the same mistake.
- Make sure every cited source exists and says what the AI claims. Search for the exact title. Open it. Find the part that supports the claim. In 2023, two lawyers were fined for handing a court six cases a chatbot had suggested that did not exist. A quick search would have caught every one.
Check three facts
Now fact-check all three. For each one, can you find two trusted sources that agree? Does the source it named exist, and does it actually say that? Keep a score. It's often more interesting than you'd expect.
Keep going
Checking facts is the same move as checking any claim, AI or not. The critical thinking toolkit has eight questions for it.
Last reviewed: 25 September 2026