Vendor-neutral · Beginner · 6 lessons · about 35 minutes · updated 2026-09-12
How to Tell When AI Is Wrong: Fact-Check Any AI Answer
A fluent answer is not a checked answer. Six short lessons that teach you why a chat model can be confidently wrong, how to spot it, and the one thirty-second check that works every time.
For anyone who uses an AI chat tool and has had the uneasy feeling of not knowing whether to believe it. No technical background needed. If you have ever pasted an AI answer somewhere it mattered, this course is for you.
Course outline
- 1Hallucinations in plain words: why a fluent answer can be falseA chat model does not look things up — it predicts the next word. That one fact explains every confident mistake it makes, and once you have it, the rest of the course is just practice.6 min · 3 objectives
- 2Confidence versus accuracy: the signals that actually matterTone tells you nothing. Learn the four things that do — a named source, a stated limit, a date, and a specific you can check — and practise telling a confident answer from a reliable one.5 min · 3 objectives
- 3Checking sources and asking for evidenceMake the model do the first pass of its own fact-check: ask it to list what it is unsure about and where you should look. Then check exactly those things, and nothing else.6 min · 3 objectives
- 4Verifying important claims: date, number, name, quoteFour kinds of claim carry almost all the risk in an AI answer. Learn to find each one, check it against an official page, and stop when the four are done — that is the whole method.7 min · 3 objectives
- 5Brainstorming versus factual research: when AI is the right toolAI is a superb tool for generating, drafting, structuring and explaining. It is the final authority for nothing. Learn to sort your tasks into the two piles before you start, and the checking takes care of itself.5 min · 3 objectives
- 6Repetition with variation: the same check, four different jobsA plant photo, a product spec, a news summary, a price comparison. Four very different questions, one identical check. Practise it until it is a reflex, and finish by choosing between two answers about AI pricing.6 min · 3 objectives
Frequently asked
Why does AI give wrong answers?
A chat model predicts the most likely next word rather than looking facts up. When it knows the subject well, the likely words are the true ones. When it does not, it produces likely words anyway, and they read exactly like an answer. That is a hallucination, and every chat model does it.
Which AI is most accurate?
The newer and larger models on the major paid plans make fewer errors, and browsing features reduce them further. None is accurate enough to skip the check. This course teaches a check that works the same on ChatGPT, Claude, Gemini and anything else.
How long does the check take?
About a minute for a typical answer: find the dates, numbers, names and quotes, and open one official page per item. The extraction step can be done by the AI itself; only the comparison with the source is yours.
Do I need a paid plan for this course?
No. Every exercise runs on a free account, and each lesson says honestly where a paid plan would help — mostly through browsing, which attaches sources you would otherwise search for. Paying reduces errors; it never removes the check.