What ChatGPT actually is

The one sentence that explains everything

ChatGPT predicts words. That's the whole trick. It was trained on an enormous amount of text — books, articles, conversations, code — and it learned, better than anything before it, what word tends to come next. You give it the start of something, and it continues, one likely word at a time, until it's said its piece.

Hold onto that one idea, because it explains both halves of your experience with it. The magic: it has absorbed more writing than any human could read in a hundred lifetimes, so it can draft, rephrase, summarize, and brainstorm on almost anything, instantly. The mistakes: it is predicting, not looking things up. When it doesn't know something, it doesn't go quiet — it predicts the most plausible-sounding answer and delivers it with total confidence. Understanding this one thing puts you ahead of most people who use it every day.

Think of it as a brilliant intern

The most useful mental model isn't "robot" or "search engine." It's a brilliant, fast, eager intern — one who has read everything but remembers nothing about you or your business unless you tell them, every single time.

• Brilliant: hand it a task and it produces something usable in seconds. Genuinely good at first drafts, rewrites, and getting you off a blank page.
• No memory of your world: it doesn't know your brand, your customer, your prices, or your voice unless you say so. Context isn't optional — it's the whole job (Lesson 2 makes this permanent).
• Occasionally overconfident: a great intern who, rather than admit "I'm not sure," will sometimes hand you a confident, wrong answer with a straight face.

You wouldn't publish an intern's first draft without reading it. Same here. The tool multiplies your effort; it doesn't replace your judgment.

What ChatGPT is NOT

Three misconceptions cause almost every bad experience beginners have:

• It is not (mainly) a search engine. This one has a wrinkle worth getting right. Modern ChatGPT can search the web when a question needs current information, and it'll show links when it does. But that's a feature bolted on top — its native way of answering is still prediction from memory. So it may answer a "what's the latest…" question from stale training data unless it actually searches. When facts and dates matter, ask it to search, and check the links yourself.
• It is not a fact database. It has no internal ledger of guaranteed-true facts. It produces language that sounds like the truth, which is usually close and occasionally invented. Treat every specific claim as "probably right, verify before you rely on it."
• It is not a mind reader. It only knows what's in the conversation. A vague request gets a vague answer — not because it's dumb, but because you handed it almost nothing to work with. Lesson 3 fixes this permanently.

The two rules that keep you safe

You can relax about almost everything in AI except these two. Tape them to your monitor:

1. Never publish a fact it gave you without checking it. Names, numbers, dates, quotes, statistics, anything legal or medical — verify before it goes out under your name. "It sounded confident" is not verification.
2. Never paste in anything you'd be upset to see leaked. Client social-security numbers, unreleased financials, someone else's private messages, passwords — keep them out. Treat the chat box like a postcard, not a vault.

Free vs paid: what actually matters for you

Verified July 2026. Plans, prices, and model names are the fastest-moving part of any AI product — the concepts below will hold, but glance at chatgpt.com/pricing before paying for anything.

A word about model names

Inside ChatGPT you may see a little menu of model names, and they change every few months. Don't overthink it. The default you get on the free plan is genuinely capable and handles everything in this course; paid plans unlock models that think a bit harder on difficult problems. No model choice matters half as much as the quality of your prompt — which is exactly what Lesson 3 is about.