lesson
25 minutes

When NOT to Use AI

AI is powerful, but it's not always right. Learn to spot when AI fails, what hallucinations are, and why knowing AI's limits is your biggest superpower.

You've been learning how to use AI. Now let's talk about something just as important: when to stop using it.

This might sound strange in a course about AI. But here's the thing -- the people who get the most value from AI are the ones who know exactly where it breaks down. That's what separates someone who uses AI well from someone who gets burned by it.

AI Gets Things Wrong

AI models like ChatGPT, Gemini, and Claude don't actually "know" anything. They predict what words should come next based on patterns in their training data. Most of the time, those predictions are impressive. But sometimes, they're completely wrong.

And here's the dangerous part: AI sounds confident even when it's wrong.

It won't say "I'm not sure about this." It'll give you a detailed, well-written answer that happens to be totally made up.

What Are Hallucinations?

When AI generates information that sounds real but isn't, we call that a hallucination. Some real examples:

  • A lawyer used ChatGPT to write a legal brief. The AI cited six court cases that didn't exist. The lawyer filed the brief without checking. He was sanctioned by the judge.
  • AI has generated fake medical advice, including drug dosages that could be dangerous.
  • Students have turned in AI-written essays with citations to books and articles that were never published.

These aren't rare edge cases. Hallucinations happen regularly, especially when you ask AI about specific facts, dates, statistics, or specialized knowledge.

When You Should NOT Rely on AI

Here are situations where you need to be extra careful -- or skip AI entirely:

Medical decisions. AI can help you understand general health topics, but it should never replace a doctor. If someone asks AI "What's this rash?" and follows the answer without seeing a professional, that's risky.

Legal advice. Laws vary by state, city, and situation. AI doesn't know your specific circumstances, and it might cite laws that don't exist or don't apply to you.

Financial decisions. AI can explain what a 401(k) is, but it shouldn't be making investment decisions for you. It doesn't know your full financial picture.

Anything with real consequences for real people. If you're writing a report that affects someone's job, benefits, or housing, double-check every AI-generated fact.

Emotional support in crisis. AI can simulate empathy, but it's not a therapist, counselor, or crisis hotline. If someone is in danger, call 988 (Suicide & Crisis Lifeline) or 911.

How to Spot AI Mistakes

Here are practical ways to catch AI errors before they cause problems:

  1. Ask for sources. If AI gives you a fact, ask where it came from. Then go check that source yourself.
  2. Cross-reference. Don't rely on one AI tool. Check important information against trusted websites, books, or experts.
  3. Watch for too-perfect answers. If something reads too smoothly and covers every angle perfectly, be suspicious. Real information is often messy.
  4. Test with questions you know the answers to. This helps you calibrate how reliable the AI is for your specific topic.

Why This Matters for Our Community

In San Antonio's underserved communities, bad AI advice can hit harder. If someone gets wrong information about SNAP benefits, immigration processes, or job applications, the consequences are real. People might miss deadlines, lose benefits, or make decisions based on fiction.

That's why at Community Incubator, we say: knowing when NOT to use AI is an AI skill.


Try This

Ask ChatGPT or another free AI tool: "Who was the first principal of [your local school]?" or "What year was [a local San Antonio landmark] built?" Then try to verify the answer with a Google search. Did the AI get it right?


Talk About It

Have you ever gotten wrong information from AI? What happened? How would you handle it differently now? Share your experience with the group.

If you or someone you know is in crisis, do not turn to an AI chat tool. Call or text 988, the Suicide and Crisis Lifeline, any time, day or night.

Stop signs: get a person, not a chatbot

0 of 5 done

Quiz: AI or a person?

  1. 01A friend asks AI whether two medicines are safe to take together. What is the better choice?

    Choose one.

  2. 02What is a hallucination in AI?

    Choose one.