Goal
Spot questionable answers and use a four-step checking routine.
Generative AI produces language based on learned patterns. Fluent wording does not establish that a claim is true: an assistant can invent a quotation, statistic, law or reference while sounding convincing. This is often called a hallucination.
When to be especially careful
- Numbers, statistics, dates and prices.
- Laws, regulations and administrative processes, which vary between countries.
- Quotations and sources with titles and links.
- Health, money and legal questions, where mistakes can have serious consequences.
- Recent developments: AI may not have current information.
The four-step routine
- Ask for sources — then open them. A source that does not exist is a warning sign.
- Cross-check — verify key information using reliable original sources, official institutions or reputable reporting.
- Question confidence — “Which parts of your answer are you least sure about?” Treat stated confidence as a prompt for further checking, not proof.
- Apply your expertise — if something does not fit your country or profession, investigate it before relying on it.
Review your previous answer. List each factual claim — number, date, name or law — and explain how I can check it against an original source. Clearly flag uncertainty and missing evidence.
AI and disinformation
AI also makes it easier to create fabricated text, images and voices. Before forwarding a shocking video or an “exclusive” WhatsApp audio, check who first published it and whether reliable outlets corroborate it. Verification is a professional habit.
Golden rule: share only information you have checked and can take responsibility for.
Today's challenge
Ask AI a factual question about your country, such as a recent statistic. Apply the four-step routine and note what was correct, approximate or wrong.