The very first thing to do after signing up for ChatGPT is to open a new chat and paste in one real task you are already facing — not a test question, not "tell me about AI" — then ask for a short, checkable output you can verify yourself.
So if you have a confusing email from your landlord, paste the whole thing and type: "Summarize this in three bullet points and tell me what they are asking me to do."
That single move teaches you more in five minutes than a week of reading guides, because it shows you what the tool is actually good at and where it quietly gets things wrong. Everything else — prompting tricks, paid tiers, which model to pick — only matters after you have felt that first real result.
## The first thing to do, and why it works
The reason a real task beats a test question is that ChatGPT has no way to know what "good" looks like for you. It predicts plausible text based on your prompt, so a vague prompt gets a vague answer. When you paste actual material — an email, a job description, a messy paragraph of your own writing — you give it context to work from, and you instantly get a reference point for judging the output.
According to the AI-Mind AI Tool Database, ChatGPT is developed by OpenAI and sits in the chat category, with a free tier running GPT-4o mini and paid tiers at $20 per month for Plus and $200 per month for Pro. That free tier is genuinely enough for your first session. You do not need to pay anything to learn whether this tool fits how you work.
Keep the first task small enough that you can check the answer in under a minute. If you paste a ten-page contract and ask for a summary, you cannot tell whether the summary is right. If you paste a three-sentence message, you can spot a mistake immediately. Small and checkable beats big and impressive.
## A worked example you can copy
Say you are applying for a job and the posting says "we need someone comfortable with data reporting and cross-team communication." You are not sure how to phrase your experience. Open a new chat and type:
"Here is a job posting: [paste it]. Here is my current resume summary: [paste two or three lines]. Rewrite my summary so it speaks to this posting. Keep it under 60 words and do not invent any experience I did not mention."
What you get back is a draft you can edit, not a finished document. That distinction matters. The model will often tighten your wording in ways you would not have thought of, and it will sometimes add a skill you never claimed — which is exactly why the instruction "do not invent any experience" is worth including. Read the output line by line against your original. Anything you cannot defend in an interview, cut.
This is also where you learn the tool's personality. Some models lean formal, some lean chatty. According to the AI-Mind AI Tool Database, Claude is Anthropic's safety-first assistant with a free tier and a Pro plan at $17 per month billed annually or $20 monthly, while Google Gemini offers a free tier on Gemini 2.5 Flash with Advanced at $19.99 per month. If ChatGPT's tone does not suit you, trying one of those on the same task is a fair comparison — same prompt, different output, and you decide which one you would actually send.
## What AI is bad at, so you are not surprised
AI chat tools are weak at anything requiring verified facts about the world unless you supply them. They are weak at math you cannot check. They are weak at knowing your context — your boss's preferences, your company's tone, what you already tried last week. They are also confidently wrong in a way that reads exactly like being right, which is the single most important thing for a beginner to internalize.
The practical rule: never send, publish, or act on AI output without reading it fully. That is not a legal disclaimer, it is the actual workflow. Treat every response as a first draft from a fast but careless assistant.
There is also a limit on how much a single chat can hold. Long conversations get unwieldy, and the model can lose track of instructions you gave twenty messages ago. That is your signal to start fresh.
## How to phrase your first prompt, and when to start over
Use a simple three-part shape: what you have, what you want, and any constraint. "Here is X. Give me Y. Keep it under Z words and do not make anything up." That structure works for almost every beginner task, from rewriting an email to planning a study schedule.
For your first session, two decision rules are worth committing to. First, start a new chat whenever you switch projects — one thread per task keeps the context clean and makes it easy to find that good answer later. Second, continue the same thread when you are refining one thing, because the model remembers what you already rejected. If you told it "too formal" three messages ago, it will still know that in the same thread.
The signal that the free tier is no longer enough is specific, not vague: you are hitting usage limits in the middle of real work, or you need a feature the free plan does not include, such as access to the strongest model or file uploads for larger documents. Until you can name the wall you are hitting, stay free. Upgrading before you have a named reason just means paying for a tool you have not learned to use yet.