Use AI safely

How to Spot AI Mistakes: 7 Rules Before You Trust an Answer

By the AI Made Plain team · Updated November 2026 · 7 min read

AI chatbots write smoothly and sound sure of themselves, even when they are wrong. Experts call these wrong-but-confident answers "hallucinations". The good news: a few habits catch most of them.

Why AI makes mistakes

A chatbot predicts likely words based on patterns in huge amounts of text. That makes it great at explaining and writing. But it does not always know whether a fact is true, recent, or about your exact situation. So it can mix things up, use old information, or fill a gap with something that only sounds right.

The 7 rules

1. The more it matters, the more you check

A dinner idea does not need checking. A medicine dose, a legal deadline, a tax rule or a money decision always does. Use AI to understand the topic, then confirm with an official source or a professional.

2. Be suspicious of exact details

Names, dates, prices, statistics, quotes and links are where mistakes hide. If an answer gives a precise number, find that number somewhere else before you repeat it.

3. Ask for sources, then open them

We tested this. We asked ChatGPT for three scientific studies about electric scooters in Meknes, Morocco, a topic with almost no research. To its credit, it said it could not verify any study on exactly that topic, and offered related studies instead.

Then actually click them. Some chatbots can search the web and show real links. Others may describe sources that do not exist. A source only counts if you can open it and it says the same thing.

4. Ask it to check itself

This does not make the answer perfect, but it often reveals weak spots.

5. Watch the date

Prices, laws, product features and "the latest" anything change fast. Ask: "Is this current? When was this true?" and check the official website.

6. Ask a second time, differently

Ask the same question in a new chat, or in another chatbot. If the answers disagree, that is your signal to check carefully.

7. Use your common sense

If something sounds too good, too strange or too convenient, stop. You know your life better than any chatbot.

Quick checklist

Type of answer	Risk	What to do
Ideas, drafts, rewording	Low	Use freely, edit to taste
Explanations of a topic	Medium	Good for learning; confirm key facts
Numbers, dates, prices, quotes	High	Always check a second source
Health, legal, money decisions	Very high	Use only to prepare questions for a professional