AI models write convincing text, and that is exactly why their mistakes are easy to miss. An answer can be well organized, confident and still wrong. This is often called a “hallucination”. A short checking routine protects you from passing an error along.
Why AI gets things wrong
A language model produces the words that are most likely to follow, based on patterns in the text it learned from. It is not looking facts up in a database. Its knowledge can be out of date, and when it lacks information it may fill the gap with something that merely sounds right.
Five checks worth making
- Ask for sources, then open them. References can be invented. A citation only counts once you have seen that it exists and says what the answer claims.
- Verify numbers, names, dates and quotes. These are the details most likely to be wrong and most damaging when they are.
- Ask for the reasoning. Have the model explain step by step how it reached the answer. Gaps are easier to spot in the steps than in the conclusion.
- Ask again in a different way. Rephrase the question, or ask what might be wrong with the first answer. Inconsistent answers are a warning sign.
- Match the effort to the stakes. A brainstorming list needs a glance. A figure going into a contract or a report needs proper verification.
Signs you should slow down
- Very specific figures with no source.
- Direct quotations attributed to a person or a document.
- Legal, medical, tax or financial questions.
- Recent events, prices or anything that changes quickly.
None of this makes AI less useful. It simply puts it in the right role: a fast, capable assistant whose work you review, not an authority whose word you take.