A gentle start.

What this thing actually is

An AI chatbot is a website where you type a question or a request, and it writes back. That's it. There's nothing to install, nothing to configure, and you don't need to know anything special to use one. If you can send a text message, you can use this.

Under the hood it's complicated. To use it, it doesn't have to be. Think of it like a very well-read assistant who's a bit too eager to please. Fast, articulate, knows a little about almost everything, occasionally makes things up with great confidence. Treat it like that and you'll get along fine.

Start with one tool

There are at least six AI chatbots being heavily marketed right now. They are all roughly similar. Picking the "best" one is a fun debate among people who use them every day; for someone trying their first one, it doesn't matter at all.

Pick one. Use it for two weeks. Don't worry about the others until you're comfortable.

My recommendation, plainly

Open a web browser. Go to chat.openai.com (this is ChatGPT). You can use it without making an account at all. Just type into the box at the bottom and press enter. If you want to save your conversations, sign up free with an email address. There's a paid version; ignore it for now.

That's the entire setup. No app, no install, no payment.

Three things to try in your first 30 minutes

You'll learn more from doing three small things than from reading any guide (this one included). Here are three. They each take about five minutes.

That's the first thirty minutes. If you do these three, you'll know more about what AI is good for than people who've watched a hundred YouTube videos on it without trying it themselves.

What a good question looks like

The biggest difference from Google: AI doesn't want keywords. It wants a sentence, the kind you'd say to a person.

You can ask it like this:

• "Help me write a polite email turning down a wedding invitation. We're traveling that week."
• "What are some good questions to ask a contractor before hiring them for a kitchen remodel?"
• "Explain Medicare Advantage vs. Medigap to me. I'm 64 and trying to decide. What questions should I be asking?"
• "My grandson keeps mentioning 'the cloud.' What is it, simply?"

The pattern: say what you're trying to do, then say what you want back. "Help me write X" or "Explain Y for someone who Z" or "What should I know about A before B." Full sentences. Plain words.

Three small upgrades that go a long way

1. Tell it who's reading. "Make this for my neighbor" reads differently from "Make this for a lawyer." It writes accordingly.
2. Tell it the length. "Two paragraphs," "one page," "five bullet points." Otherwise it tends to write a lot.
3. Tell it the tone. "Warm and informal." "Firm but not angry." "Like a friendly grandparent." It can do all of those.

Four things to keep in your back pocket

None of these should scare you off. They're the four things every careful adult should know before getting comfortable with one of these tools.

1. Don't paste anything you wouldn't post on a bulletin board.

No Social Security numbers. No bank account numbers. No full credit card numbers. No passwords. No medical records with your name on them. Yes, you can write "my doctor said my A1C is 7.2", and that's fine. No, you should not paste a screenshot of your full medical chart with your address on it. The shorthand: if it'd embarrass you to find it on a stranger's screen, don't paste it.

2. It can sound confident and be wrong.

This is the most important thing on this page. AI writes fluently. It puts together polished, authoritative-sounding sentences. That fluency does not mean it's correct. Especially with names, dates, numbers, citations, and recent events, it can make things up that look right at a glance.

The fix is simple: treat it like a chatty friend who guesses well, not a professor with a fact-checker. For anything that matters (a medical question, a legal question, a financial decision, a name you're going to repeat to someone), verify it somewhere else before relying on it.

3. It does not actually know you.

Each new conversation starts fresh. It doesn't remember what you said last week unless you tell it. Some tools have a "memory" feature that does keep notes; if you don't want that, you can usually turn it off in settings. For a beginner: don't worry about this for two weeks. Start every conversation with whatever context you need.

4. It will agree with you a little too much.

If you ask it for a second opinion, it has a tendency to find new ways to agree with the first one. If you suspect that's happening, say it out loud: "play devil's advocate, what's the strongest argument I'm wrong?" or "give me three reasons this might be a bad idea." It can absolutely do that. It just won't volunteer it.

When it goes sideways

It will. That's not a sign you've done something wrong. The model misreads what you wanted, or it fixates on the wrong word, or it produces a draft that's stiff and unhelpful.

Here's the ladder, from cheapest fix to last resort:

1. Say what you wanted differently. "Try again, but make it half as long." "Less formal." "Now do it as if you're explaining to a 12-year-old."
2. Start a new conversation. If a chat has gone in circles, hit "New chat" (top of the screen) and begin again with a clearer ask. Conversations are free and unlimited; don't try to rescue a bad one.
3. Try a different tool. If ChatGPT keeps producing nonsense on a particular topic, try Claude. They have slightly different strengths. It's free; just paste the same question.
4. Walk away. Sometimes the right answer is "this isn't the right tool for this question." That's allowed.

What this is not

Worth being clear about, because the marketing isn't.

• It is not a search engine. Google is for "what's the address of the post office in Boise." ChatGPT is for "help me write to the post office about a missing package." Different tools.
• It is not a person. It does not have feelings. You can be polite (many people are, out of habit), but you don't owe it small talk and it isn't lonely between your visits.
• It is not always right. See Part 05. This is its single biggest flaw.
• It is not free for everything. The free tier is generous. Some features (image generation, very long documents, voice mode) sit behind a paid version. You don't need any of them in your first month.
• It is not going to read your mind. If you want a specific tone, length, or format, say so. It's much better at "do this exact thing" than "guess what I want."

When you're ready for more

If the three exercises in Part 03 went well and you're using ChatGPT a couple of times a week, you've already passed the hard part: getting comfortable. Most people who try once and don't like the first answer never come back. You're past that.

From here, in roughly the order most people want them:

• Try Claude at claude.ai. Same idea, different "voice." Some people prefer one over the other. There's no harm in using both for different things.
• Try the mobile app of whichever you settled on. The voice mode is genuinely useful. You can talk to it while making dinner or driving (legally, of course). Easier than typing for many people.
• Read the Best Practices primer. Once you're past the intimidation, the primer for practitioners on this site has twelve habits that work no matter which tool you use. It's still readable for non-developers.
• Pick a tool guide. If you want depth on the one you settled on: ChatGPT or Claude App.

Most of the value here isn't in being clever with the tool. It's in being patient with yourself while you learn what it's good for and what it isn't. — TWD

You're going to be fine. Try the three things. Come back to this page when something gets confusing. There is no race.