Start here

Four things to know

1. 1

   It's a pattern-matcher with tools attached

   AI like ChatGPT and Claude was trained on a huge pile of text. At its core, it predicts what should come next — and that turns out to be enough for a lot of real work. These days it can also reach for tools on its own: searching the web, doing math with a calculator, reading PDFs and images, even running little bits of code. So when you ask about something recent, it usually goes and looks. Useful to know — but it still gets things wrong, so the verify habit below matters more than ever.

2. 2

   It has zero context on you

   Every chat starts cold. It doesn't know your kids' names, your job, your goals, or what you tried yesterday — unless you tell it. The single biggest upgrade you can make is to give it context up front.

3. 3

   It can be wrong with total confidence

   AI can invent facts, dates, sources, and quotes. This is called 'hallucinating.' It happens most when you ask about niche details, recent events, or specific people. Always verify anything that matters.

4. 4

   It gets better with conversation

   First answers are rarely the best. Push back. Ask follow-ups. Tell it what you didn't like. This is where most people stop too early.

What AI is great at

• Drafting: emails, messages, letters, lists, plans. First drafts in seconds.
• Explaining: turn jargon into plain English. Or kid-language. Or pirate-talk if you want.
• Brainstorming: 20 ideas in 30 seconds. Most will be mediocre. A few will be gold.
• Translating: between languages, but also between formats — a paragraph into a checklist, a chart into a summary, a doctor's note into something you can act on.
• Thinking out loud: ask it to play devil's advocate. To find holes in your plan. To list what you're missing.

What AI is NOT great at (yet)

• Knowing your specific situation — your finances, your medical history, your relationships, what happened at dinner last night — unless you tell it.
• Being right just because it sounds confident. Even with web access, it can misread a source or quote one that's wrong. Confidence is not accuracy.
• Quiet judgment calls — taste, tone with a specific person, whether your mother-in-law will actually like the gift. It can help you think; it can't feel the room.
• Replacing professional advice — for medical, legal, or financial decisions that matter, use AI to prepare and ask better questions, then talk to a human professional.

Which AI should I use?

For most family use, you only need to know two names: ChatGPT (by OpenAI) and Claude (by Anthropic). They're both free to try, both great. Pick one and use it for a month before worrying about the rest.

Other names you'll hear: Gemini (Google), Copilot (Microsoft) — both fine. Perplexity is great when you want answers with sources, like a smart Google.

Next stop

Now that you know what AI is, learn the four habits that turn it from a toy into a tool: How to think with AI →