Chapter 3: Your First Hour With ChatGPT
⚡ FAST TRACK
Why we start here
We are starting with ChatGPT for one reason: it is the door most people walk through first. That is not an endorsement. By the time you compare the major tools, you may well decide that Claude suits your writing better, or that Gemini fits your existing tools, or that the assistant already built into your work software is enough.
None of that matters yet. What matters is that you spend an hour with a real tool and stop treating this as a subject you read about. Everything you learn in this hour transfers. The window will look different, the buttons will have moved, but the way you get value out of the thing will be identical across every tool in this book.
Set aside sixty minutes. Have something real you want help with, not a test, an actual small task you have been putting off.
A warning about product specifics
I need to show you something before we begin.
While preparing this chapter I ran a simple search: what model does the free version of ChatGPT use? Within the first page of results I was told, by sources that all sounded equally authoritative, that the answer was GPT-4o Mini, GPT-4o, GPT-5.5, and GPT-5.6. Several of these pages carried recent dates. Several described themselves as definitive.
They cannot all be right. Most of them were simply old pages with fresh timestamps, or content written quickly to rank in search results rather than to inform anyone.
This is the environment in which you are learning, and it is why this book is built the way it is. ChatGPT’s pricing and model availability changed repeatedly through the third quarter of 2026. No printed book, article, or video can remain perfectly current. You need to know how to check the official information yourself, which is why staying current is a core skill.
So here is my rule for the rest of this chapter, and for every tool chapter in this book: anything with a number or a product name attached is a snapshot with a date on it. Anything without one is durable. Read accordingly.
The snapshot (third quarter of 2026)
ChatGPT sells access in tiers: a useful free plan, one or more paid consumer plans, higher-capacity plans for heavy users, and business and enterprise options. Paid plans generally add stronger models, higher limits, and advanced features. The exact names, prices, and limits change, so the dated pricing snapshot in the Bonus Resource Library is the right place to check the figures.
Let the details change. The durable shape underneath them is what matters, and it is much the same at every AI company:
• A free tier that is genuinely useful and deliberately limited, usually by message caps and by giving you the fast, cheaper model rather than the best one.
• A consumer paid tier that often costs roughly the price of a casual lunch for a month of advanced capabilities.
• Higher tiers for heavy users, priced well above that.
• Business and enterprise tiers whose real differentiator is not capability but contract terms: administrative control, compliance certifications, and data handling. That last one matters enormously and we will return to it.
Start free. Give yourself a week to discover what you would use before considering a subscription. The free tier is usually enough for that first decision.
Minutes 0–10: Getting in and looking around
Go to the website or install the app, and create an account. You will need an email address. You should not need a credit card for the free tier; if you are being asked for one, check that you have not landed on a lookalike site: fake AI applications are a real and growing problem.
When the interface loads, ignore most of it. Four things matter.
The message box. Where you type. This is ninety percent of the product.
The conversation list. Usually along one edge. Every conversation is saved and can be reopened. This is more useful than it sounds: good conversations are worth returning to.
The new-conversation button. Starts a fresh context. Knowing when to press it is a real skill.
The model selector. Often a small label or dropdown near the top showing which model you are talking to. Free accounts may have limited choice here. What you need to understand is that these products are not one thing: they are a family of models, faster and cheaper at one end, slower and more capable at the other, and which one answers you affects your results substantially.
Everything else (settings, custom instructions, projects, connected apps, voice) we will come back to in the final ten minutes.
Minutes 10–25: Four prompts, in this order
Begin with a manageable task and run this sequence. Early success will make the harder problems easier to approach.
Prompt 1: Establish that it works.
Explain what you are and what you’re good at, in about 150 words, as if I’ve never used a tool like you before.
Read the answer critically. Notice that it is well organized, appropriately modest, and pleasant. Notice that you have no way of knowing whether it is accurate. Hold onto that feeling; it is the correct one and Part II will justify it.
Prompt 2: Transformation, the strongest use case.
Find any text you have written: an email, a paragraph from a document, anything.
Rewrite this to be about half as long, keeping every substantive point. Then tell me what you cut and why. [paste your text]
Two things to notice. First, the quality is probably good, because you have given it real material and a clear instruction. Second, the “tell me what you cut” clause did most of the work of making the output trustworthy — you can check the claim against your original in ten seconds. Building in your own verification is a habit worth forming now.
Prompt 3: The bad prompt, on purpose.
Write something about marketing.
Look at what comes back. It will be competent, generic, structurally impressive, and useless. Nothing is wrong with the tool. The request contained no context, no audience, no purpose, no format, and no constraints, so the model did the only thing available to it and produced the statistical center of everything ever written about marketing.
Many people conclude AI is overrated after running the equivalent of this prompt, receiving the equivalent of this answer, and stopping there.
Prompt 4: The same request, built properly.
I run a two-person landscaping business in a suburb of about 40,000 people. Most of my work comes from word of mouth and I want more spring cleanup jobs. I have about $300 and four hours a week. Give me five specific things to try, ranked by what you’d do first, and tell me what each one is likely to cost me in time and money.
Compare the two answers. That gap (between prompt 3 and prompt 4) shows why context matters, and you just discovered it yourself in ninety seconds. Substitute your own situation and run it again with real details.
Minutes 25–40: The follow-up is where the value is
Beginners send one message and judge the tool by the reply. Experienced users treat the first reply as a starting position.
Take whichever answer above was closest to useful and push on it:
Shorter. Cut the introduction entirely.
Number two is the only one that fits my situation. Give me five more like it.
That’s too formal for my audience. Try again, plainer.
What are you assuming about my situation that might not be true?
That last one is worth committing to memory. It routinely surfaces the buried assumption that was quietly making the whole answer wrong for you.
Notice that these follow-ups are short, blunt, and require no politeness or preamble. You are not being rude; you are being efficient. The conversation generally retains what came before, although long conversations may lose or distort earlier context.
Minutes 40–50: Files, images, voice, and the web
Depending on your plan, some or all of these will be available. Try whichever you have.
Attach a file. A PDF, a document, a spreadsheet. Then ask something specific about it: What are the three obligations this contract puts on me? Quote the relevant sentence for each. Asking it to quote the source is not decoration — it is the fastest verification you have, because you can check the quote against the document immediately. Later guidance covers the ways file work can go wrong with long documents and numbers.
Send an image. A photo of a document, a chart, a broken appliance, a plant, a screenshot of an error message. Ask what it shows. This capability surprises people more than any other.
Try voice. Speak instead of typing. For many people this is where AI becomes genuinely comfortable, particularly for thinking out loud on a drive or a walk.
Ask it to search the web. Ask about something that happened this week. Watch whether it actually retrieves sources or answers from memory, and notice the difference in reliability. A model answering from memory about recent events is the single most predictable way to get a confident wrong answer.
Minutes 50–60: Three settings that matter
Most settings are cosmetic. Three are not.
Data and training controls. Find the privacy or data settings and look for an option governing whether your conversations may be used to improve the model. On consumer plans this is commonly on by default with an opt-out available, while business and enterprise plans typically do not train on customer data by default. Decide now, deliberately. If you will ever paste work material into this tool, go turn training off today.
Memory. Many tools now remember details across conversations: your job, your preferences, your projects. This is genuinely convenient and it is also a privacy surface. Know whether yours is on, know how to view what it has stored, and know how to clear it.
Custom instructions. A standing note the tool applies to every conversation. Mine says something close to: I prefer direct answers without preamble. Tell me when you’re uncertain. Don’t compliment my questions. Two minutes here improves every conversation you will ever have with the tool. Fill it in.
Four habits for a confident first week
Use ordinary, low-stakes material during your first week. Keep client data, employee records, unreleased financials, medical information, and other confidential material out until you have configured your settings and established clear privacy rules.
Check factual claims before acting on them, especially numbers, dates, citations, prices, legal requirements, and dosages. This small habit lets you use AI confidently.
Use the free version for a week before deciding whether paid features would improve work you now understand.
Stay curious during the first week. One hour is too early to decide that AI is either amazing or overrated; useful judgment grows from trying several real tasks.
Exercise 3.1: The prompt-quality gap
Take one real task from your own work or life. Write the laziest possible version of the request in under ten words, and run it. Then write a version with your actual situation, your constraints, and your desired format (at least four sentences) and run that.
Save both answers. You will use them again in Chapter 10.
Exercise 3.2: Write your custom instructions
Fill in the custom-instructions field, and include at least one instruction about how you want to be told about uncertainty. Something like: When you’re not confident, say so explicitly rather than hedging in the prose.
Exercise 3.3: Find the current facts yourself
Open the dated pricing snapshot in the Bonus Resource Library. Then find out what is true today: the current plans, the current prices, and the model your account is actually using.
Use the official source, not a search result summary. Write the answer and today’s date in the margin.
That habit (going to the source, dating what you find) is worth more than any specific fact this book could give you.
Learn More About Artificial Intelligence
Learn more about the CLEAR, TRUST and LEARN methods for Artificial Intelligence