AI for Everyone · Part Three · Free · No signup

Trust, but verify.

These tools can be wrong, and they are wrong in a particular way that catches careful people off guard. They do not hedge. Here are the three habits that make that harmless — and the four things that never go in an AI chat.

Part 3 · 3:32 · captions available · Dr. Pesach Walter, Ed.D.

Three habits

Thirty seconds each. Do them three times and they become automatic.

Ask for the source

“What sources support that? Give me links.” If it cannot produce a real one, treat the claim as unconfirmed.

Click the link

Actually click it. Does the page exist? Does it say what was claimed? Most errors die at this step.

Check what could cost you

Money, medicine, legal deadlines, your identity. For those four, AI prepares your questions. A human being answers them.

Practice · Part 3

Do this today.

Read each step, then actually do it. Ten to twenty minutes. The point is not to be perfect — it is to finish having done it yourself.

Step One

Ask something where being wrong would matter

Pick a real question with real consequences — a benefit rule, a medication interaction, an enrollment deadline, a tax threshold. Then verify it in front of yourself.

What are the current rules for [your question]? Then tell me how confident you are, and what sources support that. Give me links to the original sources, not to summaries.

Step Two

Open one link and read it yourself

Confirm three things before you believe anything:

  • The page actually exists.
  • It says what the summary claimed it said.
  • The publisher is credible — a university, a government agency, a medical society — not a website selling something.

This is called hallucination in the field. I prefer a plainer name: confident guessing. It happens because the tool predicts what a good answer looks like, and a good-looking answer is usually a correct one. Usually.

Step Three

Run the privacy check on your own prompt

Reread what you typed. Remove anything that identifies a real person — including you. You can still get help with sensitive matters; just describe them without identifiers. “A seventy-four-year-old family member with early kidney disease” gets you the same quality of help as a name and a chart number, and it costs you nothing.

Step Four

Practice on a claim that looks perfect

Ask for something specific enough to check — a statistic with a year attached, or a study with an author. Then follow the trail all the way to the original. Sooner or later you will find one that does not hold up, and that is the day this lesson becomes permanent.

Give me three published statistics about adults over sixty enrolling in graduate programs in the United States, each with the year, the publishing organization, and a direct link to the original report.

One more caution

Advertisements that target people our age

If you see an ad promising a fast doctorate, a degree designed just for seniors, or a program that will waive your requirements — be skeptical. Real programs have standards, and standards are a compliment.

If you are unsure about a program you have been shown, that is exactly the kind of question worth bringing to a conversation before you give anyone your information.

Print this and tape it up

Use your browser's print command to put this page on paper. The video and buttons will not print — only what you need.

  1. Never goes in an AI chat: Social Security number · bank or credit card numbers · passwords · other people's private information.
  2. Three questions that verify any answer: What sources support that? · Does the link actually say it? · Would being wrong here cost me money, health, or time?
  3. For money, medicine, law, and identity: AI prepares the questions. A human professional answers them.

“Verifying is not fear. It is the ordinary professional caution you have practiced your whole life.”