Everyone seems to be talking about it, and you've never really tried it. Or maybe you did once: you typed a question, the answer felt odd, and you closed the tab.
Good news: using ChatGPT takes no technical skill. What matters is understanding what it is and picking up a few habits. Everything below also applies to Claude, Gemini, Le Chat or any other assistant: the screens differ, but the way you use them is the same.
What ChatGPT is, in one paragraph
ChatGPT is an AI assistant: a program you chat with in writing, in everyday language. Under the hood is an LLM, a large language model trained on huge amounts of text. It doesn't look the answer up in a database: it produces likely text, word by word, based on what you write. That's what makes it so flexible for writing, rewording and explaining, and it's also why it can be wrong while sounding completely sure of itself. Keep that in mind; everything else follows from it.
Getting access
Go through the official website or the official app, on a computer or a phone. Be careful with lookalike apps in app stores: check who the publisher is before installing anything.
Free versions exist. Depending on the tool and the moment, you can sometimes start without an account, or you'll need to create one. There are also paid plans, whose features and limits change often. To get started, the free version is plenty. Terms change regularly, so check them on the official site.
The interface, in four parts
Menus vary from one tool to another and from one update to the next, but you'll almost always find:
- The input box, at the bottom: where you type your request, called a prompt.
- The conversation, in the middle: your messages and the replies follow each other, like in a messaging app.
- The "new chat" button: a blank page, with none of the previous discussion's context.
- The history, often on the side: your past conversations, which you can reopen or delete.
Depending on the tool, you may also see a model picker, a button to attach a file, and settings. You don't need to understand all of it on day one.
Your first conversation, step by step
Pick a real small task from your life, not a trick question to catch the machine out: planning a meal, writing a tricky message, making sense of an official letter.
- Write as if to a capable person who doesn't know you. Give context: your situation, what you want, what it's for.
- Say what you expect back: a list, three ideas, a short message, a simple explanation.
- Read the whole answer before judging it.
- React: what works, what doesn't, what's missing.
A sample first message:
I'm having six friends over on Saturday night, one of them is vegetarian. I have two hours to cook and a tight budget. Suggest three simple menus, each with a shopping list.
Compare that with "meal ideas?": the more context you give, the better the answer fits your situation. For the full method, read how to write a good ChatGPT prompt.
The habit that changes everything: follow up instead of starting over
The classic beginner mistake: the answer isn't right, so you ask something else, or give up. But the assistant takes into account everything said in the current conversation. So you can simply ask it to adjust:
Shorter, please.
Swap the second menu for something that doesn't need an oven.
Explain that as if I were twelve.
Ask me questions before answering, you're missing some information.
It's a conversation, not a vending machine. Two or three follow-ups often turn an average answer into a genuinely useful one.
New topic, new chat
The reverse is also true: when you change subject, start a new chat. Otherwise the assistant keeps factoring in everything before, and your Saturday menu may bleed into your letter to the insurance company.
Very long conversations have their limits too: the assistant can only take a certain amount of text into account at once, its context. If the replies start losing the thread, open a fresh chat with a short summary of what matters.
What it does well, and what it doesn't
Fairly comfortable with: writing, rewording, shortening, changing the tone; explaining an idea simply; suggesting ideas and outlines; summarizing a text you give it.
Less reliable with: precise facts (dates, figures, quotes, references); recent news, depending on the tool and its cutoff date; topics where a mistake is costly, like health, law or money. There it can help you understand things and prepare your questions, not replace a professional.
Check what matters
An assistant can make up false information in the same confident tone as accurate information: that's a hallucination. It isn't a rare glitch; it comes from the way these tools work.
The right habit isn't to check everything but to check what matters: a figure you'll repeat, a deadline, an official rule, a health tip. Ask for its sources, open them yourself, and cross-check with an official website. To understand why this happens, read why ChatGPT makes things up. For a step-by-step method, read how to fact-check an AI answer, and sharpen your eye with Hallucination Hunt.
What you don't share with it
What you type is sent to the service and may be stored. Before pasting anything, ask yourself whether you'd mind someone else reading it. Avoid passwords, account numbers, health data, information about other people and confidential work documents. Replacing names with initials is often enough. The useful settings are covered in ChatGPT and personal data.
Mini exercise: your first 10-minute session
Open ChatGPT or the assistant of your choice.
- First prompt (2 min): "Explain in five sentences what inflation is, with an everyday example."
- First follow-up (1 min): "Redo it in three sentences, for a ten-year-old."
- Second prompt (3 min): a concrete task from your week, with context (a message to send, a schedule, a shopping list).
- Second follow-up (1 min): ask for a specific change, like shorter, a different tone or one more option.
- Third prompt, in a new chat (3 min): ask a factual question you already know the exact answer to, on a topic you know well (your job, your town, a hobby).
What to notice: how the answer changes when you add context; how little a short follow-up takes to redirect the result; that the new chat knows nothing about what came before; and, on your last question, whether it gets it wrong and whether it flags its doubts. Write down one thing that surprised you.
Going further
This guide matches Ovrelia's free chapter "Light the Lantern": you learn what an LLM is, choose your assistant, set up privacy, custom instructions and memory, then write a first real request. It plays in your browser, no account needed: play the chapter.
To get better at writing requests, practice in the prompt lab and borrow from our ChatGPT prompt examples. Next, set up your custom instructions, then test yourself with the friend challenge.
Still torn between ChatGPT, Claude, Gemini, Copilot and Le Chat? The guide on which free AI assistant to choose helps you decide with a fifteen-minute home test.
Looking for a first real-life use? The guide to planning a trip with AI takes you from your constraints to a realistic itinerary, checking what matters along the way.
Before you paste your first real text, play a round of Safe to paste?, a free game where you spot the personal details to remove before pasting a text into an AI chat.
Frequently asked questions
Is ChatGPT free?
There is a free version, which is enough to get started, as well as paid plans with more features or higher usage limits. Terms, and whether you need an account, change often, so check them on the official website.
Do I need special phrases to use it well?
No. Write in everyday language, as you would to a capable person who doesn't know you. What counts is the context you give and what you expect back. If the answer isn't right, follow up to adjust it.
Why does ChatGPT sometimes give wrong answers?
Because it produces likely text based on what it learned, without checking facts. So it can be wrong while sounding sure. Always check important information, such as figures, dates or health advice, against reliable sources.
Should I start a new chat for every question?
Not for every question. Stay in the same conversation while you're working on the same topic, so the context is kept and you can follow up. Start a new one when you change subject, or when a very long discussion starts losing the thread.