ChatGPT: a tour of the screen

Getting there and signing in

ChatGPT lives on a website. Open your web browser and go to chatgpt.com. (You may still see the older address, chat.openai.com, mentioned in places; it simply redirects to the same site.)

You'll be asked to sign in or sign up. You can create a free account with an email address, or sign in with a Google, Microsoft or Apple account you already have. There's no cost to make an account and you don't need to add a card. Signing in is what lets ChatGPT save your chat history, so it's worth doing rather than clicking around as a guest.

There's also a desktop app for Mac and Windows if you prefer one. It was recently rebuilt to bundle three things in one window: ordinary Chat, a work-and-documents surface called ChatGPT Work, and the coding tool Codex. The older, chat-only desktop app now sits alongside it under the name ChatGPT Classic. You don't need any of this on day one; the website covers everything in this level, and ChatGPT Work and its new Sites feature are for later. But if a colleague mentions "Work", "Codex", "Sites" or "Classic", that restructure is what they mean (see Keeping current).

Once you're in, you land on a mostly empty screen with a box in the middle. That's home. There is far less going on here than it first appears.

The screen, in four parts

Almost everything you'll ever touch is one of four things.

• The message box sits along the bottom. You type your prompt, the instruction or question, and press Enter or click the send arrow. Type, send, read the reply, type again: that loop is the whole tool. Everything else just supports it.
• The new-chat button is near the top-left. It starts a fresh, blank conversation. This matters more than it sounds: ChatGPT only remembers what you've said within a single chat, so a conversation about holiday plans and one about a work email are best kept apart. New topic, new chat.
• Your chat history runs down the left. Each conversation is saved automatically and given a short title. Click any one to reopen it exactly where you left off; you can rename or delete them too. If the sidebar is hidden (common on phones and narrow windows) a small menu icon opens it. Nothing is lost when you close the tab.
• Your account and settings sit behind your profile picture, a small circle in a corner. That's where things we cover later live: custom instructions, memory, and managing your subscription. You don't need to touch any of it today.

Your first message

Let's make the loop concrete. Open a new chat and give it a real, low-stakes job, not "hello", but something with enough detail that the reply is actually useful.

I'm new to ChatGPT and want to see what it can do. In about 80 words, plain English, explain to a non-technical colleague what ChatGPT is good for at work. End with one honest sentence about what it's not reliable for.

Press Enter. In a few seconds a tidy paragraph appears above the box. Now do the thing beginners forget: keep going in the same chat. Type "make it warmer and cut it to 50 words." It rewrites, keeping everything you set up. That back-and-forth, not the first reply, is where the value is, and it's the habit the rest of this level builds on.

The plan map: Free, Go, Plus, Pro, Business

ChatGPT has a free version and several paid tiers stacked on top. The exact prices, limits and included features change often, so treat the names below as a durable map and check the official pricing page for today's specifics. Roughly, in plain terms:

• Free. Full access to chatting and a capable model, with caps on how much of the newest, heaviest features you can use in a given window. Useful in its own right, not a crippled demo. The details are in the Free tier FAQ.
• Go. A low-cost step up for people who bump into the free limits but don't need everything. Higher usage than free, at the cheapest paid price.
• Plus. The mainstream personal subscription. Higher limits again, priority access when the service is busy, and earlier access to the latest models and features.
• Pro. The top personal tier, aimed at heavy daily users: the most usage and the most capable modes. Most beginners never need it.
• Business (with Enterprise above it for large organisations) is for teams. It adds admin controls and a workspace where your conversations aren't used to train the models by default. This is the kind of tier an employer approves for real work data; a personal Free or Plus account is not.

One recent change worth knowing: OpenAI has begun testing ads in the Free and Go tiers (clearly labelled), with an opt-in for personalised ads that draws on your chat history to target them; the paid tiers (Plus, Pro, Business) stay ad-free. If you turn personalised ads on, treat it like any other feature that reads your chats and keep the Phase 0 privacy rules in mind (see Keeping current for how far this test has rolled out).

Our honest advice: start free. You won't know whether you need more until you've used it a while, so only pay when you can name the specific limit that's getting in your way.

The model picker: fast versus thinking

Near the top of the screen you'll usually see the name of the current model, the particular version of ChatGPT you're talking to. Names change constantly as OpenAI ships new versions, so don't memorise them. What's durable is the simplified choice the picker now offers, and it's worth understanding:

• An Auto setting (often the default) that quietly picks for you based on how hard your question looks.
• A fast, everyday option for quick replies on ordinary tasks: drafting, rewriting, summarising, simple questions.
• A slower thinking option, a reasoning model that takes longer and works step by step, better for multi-step problems, careful analysis, or anything numerical or logical where a snap answer often goes wrong.

Leaving it on Auto is a fine default. Reach for the thinking option only when the task really rewards it. You'll wait longer, so don't use it to rewrite an email. The official write-up of how these modes work lives in GPT-5 in ChatGPT.

Rewrite this so it's friendlier and about half the length: "Please be advised the meeting scheduled for Thursday has been moved to Friday at 10am."

Use the thinking model for this. We have three possible meeting slots and four people, each with different availability and priorities. Work through the trade-offs step by step and recommend the best slot, showing your reasoning. [paste the details]

Try it now

Common mistakes

• Piling every topic into one endless chat. It gets muddled and answers start bleeding between topics. Start a new chat per task.
• Fretting over the model picker. Beginners lose time here. Leave it on Auto; only switch to the thinking model for multi-step problems.
• Assuming you must pay to try it properly. You don't. Free is a real, working version; pay only when a named limit blocks you.
• Trusting the confident first reply because the screen looks polished. A tidy interface and a fluent answer say nothing about whether the facts are right. The tool sounds equally sure when it's wrong, so read anything load-bearing critically before you rely on it.

Keeping current

Plan names, prices and model options all shift; Go and the simplified picker are recent examples. Two current specifics that will date: the desktop app that bundles Chat, ChatGPT Work and Codex (with the old app renamed ChatGPT Classic) rolled out on 9 July 2026, so the exact layout and what "Work" and "Sites" do are still settling. And the ads test began in February 2026, US first, on Free and Go only; how widely it has expanded, and the exact opt-in controls, will have moved on. When something on your screen doesn't match this tour, check OpenAI's official ChatGPT release notes and the ads in ChatGPT help article. Accurate as of 14 July 2026.