Checking AI's work
Fact-checking, source-checking and spotting hallucinations: the habits that turn a fluent draft into something you can sign off.
6 lessons · about 1 hour
What’s inside
- 01Why checking is your job
AI writes fluently whether or not it is right, so the responsibility for accuracy stays with the person who uses the output. · 9 min
- 02Spotting hallucinations
The tell-tale signs that an AI answer contains invented material, and quick tests that expose it. · 10 min
- 03Checking citations and sources
How to confirm a cited source exists, says what the AI claims, and is worth trusting, with the places to look in the UK. · 11 min
- 04Checking numbers and calculations
How to catch wrong totals, misleading percentages and unit mix-ups before a figure goes anywhere important. · 10 min
- 05Second opinions: lateral reading and self-critique
Check claims the way professional fact-checkers do, and use the model's own self-review for what it is good at, without relying on it. · 11 min
- 06A sign-off checklist by risk level
Match how hard you check to what's at stake, with a practical checklist for low, medium and high-stakes work. · 10 min
