Lesson 9 of 9 · AI Basics
Trust and limits
AI states everything in the same confident voice, right or wrong. This lesson is how you stop getting fooled, and what you should never paste into a chatbot.
Hallucination: confidently wrong
Remember lesson 1: AI predicts what a good answer sounds like. Usually that lands on the truth. Sometimes it produces something that merely sounds true: a plausible date, a realistic-looking citation, a statistic that was never real. This is called hallucination, and it arrives in the same confident tone as everything else.
Three confident AI answers. Tap each to fact-check it.
"The Eiffel Tower was completed in 1899 for the Paris World's Fair."
"Water boils at a lower temperature at high altitude, so pasta takes longer to cook in the mountains."
"Utah's Zion National Park was established in 1919, making it Utah's first national park."
Score yourself honestly. Most people can't tell which one is wrong, and that's the point.
The rule of stakes
You don't need to fact-check a brainstorm or a birthday poem. Scale your checking to the stakes:
- Low stakes (ideas, drafts, summaries of text you provided): use it freely.
- Medium stakes (facts you'll repeat, numbers in an email): ask the AI to search the web and show its sources, then click one.
- High stakes (legal, medical, financial, anything you'd sign): AI drafts, a professional verifies. Every time.
One more habit: names, dates, numbers, and citations are where hallucinations concentrate. Double-check those even when everything else reads fine.
What not to paste into a chatbot
Treat a chatbot like a very smart contractor you just met: helpful, but not someone who should hold your passwords. Keep out: Social Security numbers, full account or card numbers, passwords, medical records with names attached, and other people's private information. Business owners: client lists and unreleased financials stay out too, unless you're on a business plan where your company controls the data.
Also check the settings. Consumer AI apps differ on whether chats can be used to improve their models, and most have a toggle to opt out. Flip it once and move on.
You made it
That's the course. You now know what AI is, how it's trained, how to pick a tool, how to ask well, and how to trust carefully. That puts you ahead of most people using these tools daily. Go put it to work, and come share what you build in the Facebook group.
Last updated August 11, 2026