everyonebeginner

Why AI makes mistakes (and how to catch them)

By Sanjeev Jaiswal ·

AI chatbots are useful, and they’re often wrong in ways that are hard to spot. Knowing why helps you know when to be careful.

Four reasons AI gets things wrong

  1. It predicts likely text, not true text. A chatbot writes what sounds right. Usually that matches the truth, but sometimes it invents a fact, a quote or a source that looks completely real. This is called a hallucination. (How chatbots work explains why.)
  2. Its knowledge has a cutoff. Unless it can search the web, it may not know about recent changes: new rules, prices, results or news.
  3. It learned from the internet, and the internet is wrong sometimes. Myths, outdated advice and biased views are in the training data too.
  4. It wants to be helpful. If your question assumes something false (“Why is X true?”), AI may go along with it instead of correcting you.

Danger signs

Be extra careful when the answer involves:

The 4-step check

  1. Ask for sources. “Where does this come from? Give links.”
  2. Open the source. Make sure it exists and actually says what the AI claims.
  3. Cross-check with one trusted site: a government portal, an official website or your textbook.
  4. Ask a person when the stakes are high: a doctor, teacher, CA or lawyer.

The right mindset

Treat AI like a very well-read friend who sometimes bluffs. Great for ideas and explanations, but check before you act.

Check yourself

  1. An AI gives you a quote "from a 2019 study". What should you do?

  2. Which answer needs the MOST checking?

  3. You ask "Why is the Great Wall of China visible from space?" What might go wrong?

Go deeper

More for you: 🎒 Students · 💼 Office workers · 📊 Finance and accounts · 📰 Journalists and media · ✍️ Bloggers and writers · 🎬 Reel and YouTube creators · 🧭 Job seekers · 🏡 Homemakers · 👵 Seniors