ChatGPT: why people love it and where it's weaker

If someone knows the name of exactly one AI, it's ChatGPT. It has turned into a generic word the way Xerox and Hoover did: people say "ask the chat" and mean any model at all.

The question worth asking is whether it's the best or just the loudest. The honest answer is both, though not everywhere. Let's work out exactly where.

Why it grew so big

ChatGPT's strength isn't one particular ability — it's that everything sits under one roof. You open a single window and it's all already there, where elsewhere you'd be hopping between services.

• Talking out loud. Not "dictate and get text", but an actual conversation: you interrupt, you clarify, it answers with intonation. That's the single most common reason people install the app.
• Images, files, web search. All inside the same chat, no extra tabs: you can drop in a PDF and immediately ask for an illustration to go with it.
• A store of ready-made assistants. More on that below — it's a notable thing the neighbours don't have in this form.
• A huge community. More guides, walkthroughs and ready prompts have been written for ChatGPT than for everyone else combined. For a beginner that means almost any wall you hit already has someone else's answer waiting.

On top of all that it's even: emails, plans, ideas, analysis, study, code — consistently good. Not dazzling, but good everywhere, which is exactly what you need while you still don't know what you'll need.

The scale that makes it convenient for a beginner

On 6 October 2025, at OpenAI's own conference, Sam Altman put the figure at 800 million weekly users. That counts active accounts rather than people — plenty of us have two — but the order of magnitude is clear. What matters is the consequence, not the number itself.

When that many people use a tool, a layer of other people's experience builds up around it. You get stuck on "how do I make it write shorter" — that has been answered a thousand times, with examples. You need a prompt for a CV — there are hundreds, take your pick. With a less popular model you're alone with your problem. For someone just starting out, that outweighs a difference in answer quality.

GPTs: a store of roles somebody set up for you

Inside ChatGPT lives a catalogue of ready-made assistants, called GPTs. They aren't separate programs: it's the same model, but somebody has already written it a detailed brief on who to be and how to behave. An interview coach, a maths tutor, a travel consultant, a CV editor.

It sounds dull and works briskly. Open the interview coach and it doesn't wait for you to phrase a good request — it starts asking questions like a real interviewer and picks your answers apart. You configured nothing; you just tapped.

For a beginner that's a fast way to see the model's ceiling before you learn to build such roles yourself.

What an ordinary day with it looks like

So that "ecosystem" stops being a word, look at one day in the life of somebody who's already used to it.

• Morning, on the way to work. Out loud: "Let's rehearse a call with an unhappy client. You're the client, I'm me." Half an hour of commuting becomes a rehearsal, hands free.
• Daytime, at the computer. Drops a lease into the chat: "What am I on the hook for and where's the risk?" — then asks for a draft email to the lawyer about the dodgy clauses.
• Evening, in the pharmacy. Photographs the label on a vitamin bottle: "What here duplicates what I'm already taking?"

One account, three devices, no special skills. That invisibility is exactly what turns ChatGPT into a habit.

Where its neighbours are stronger

Now the honest part. There are jobs where ChatGPT loses, and knowing them is more useful than any list of virtues.

• Very long documents. A hundred-page contract or a whole book is Claude's territory: it holds a big text in its head more reliably and loses the beginning less often.
• Fine work on style. When the text has to sound like a living person rather than a cheerful press release, the praise again tends to go to Claude.
• Freshness and Google. Gemini lives inside Google's mail, documents and search — it reaches current information faster and more naturally.
• Artistic images. The built-in generator is fine for a picture to go with a post, but where you need cover-quality art, Midjourney is still ahead.

Weak spots worth knowing about in advance

ChatGPT has three quirks that beginners mistake for a fault. None of them is a fault — and each has an obvious response.

• It's wordy It likes to start from far away, break everything into points and add a conclusion nobody asked for. Fixed once and for all: put "answer briefly, no preamble and no summary" into your personalisation settings.
• On the free tier it sometimes gets dimmer You're not imagining it: at peak times free users get switched to the junior model, and answers turn simpler and flatter. Good yesterday, poor today — that's usually why, not your request. Ask again in an hour.
• It's even rather than outstanding In narrow disciplines it gets overtaken: long documents, artistic images, fresh data. For everyday work evenness is a plus. Hit one specific hard job and you look at a neighbour — that's the next lesson.

Two questions that come up a lot

The first is about Copilot. That's what Microsoft calls its AI assistant: it's built into Windows (the button on the taskbar), into the Edge browser and into the Office programs — Word, Excel, Outlook. You don't install it separately, most people already have it, so the question is fair: if Copilot is already there, why would I want ChatGPT?

OpenAI models run inside Copilot — the same ones. But the shell around them is different and noticeably trimmed: some features aren't there, and answers come out shorter and more cautious. In exchange it does something ChatGPT can't — work inside your actual Word file or Excel sheet. If all your work lives in Microsoft Office documents, Copilot is more convenient. For everything else the full ChatGPT is richer.

The second question is "will I regret starting with it?". Let's price the worst case. In a month you realise you want Claude for long documents, so you make a second account. That's five minutes and zero euros. The skill of briefing a model moves with you intact, and the history in the old account doesn't go anywhere. The cost of this "mistake" is literally nothing — so choose calmly.

Do this now

Two short experiments, to feel the difference between reading about an ecosystem and trying one.

First: open the mobile app, switch on voice mode and talk through tomorrow out loud. Notice how it asks about details — that's a conversation, not dictation. If you don't have the app yet, install it only from the OpenAI publisher: App Store and Google Play. The stores are full of fakes that charge for what's free here.

Second: go into the GPTs catalogue and find an assistant for your own profession. Search in English, that's how the catalogue is built: marketing, excel, teacher, hr. Ask it a real work question, not a test one.

After those ten minutes, the next lesson on Claude and Gemini lands on your own experience instead of empty ground — and you'll have something to compare against.