ChatGPT launched in November 2022 and reached one million users in five days. Instagram took two and a half months to do the same. Something had clearly shifted — for the first time, genuinely powerful AI was accessible to anyone with a phone or laptop, no technical knowledge required.
But despite how widely it's used, most people still don't really understand what ChatGPT is, how it works, or why it sometimes gets things spectacularly wrong. This guide covers all of that in plain English.
What ChatGPT actually is
ChatGPT is a large language model (LLM) — a type of AI trained on an enormous amount of text. We're talking billions of documents: books, websites, articles, code, conversations. Through this training, it learned patterns in how language works — what words tend to follow other words, how ideas connect, how different types of writing are structured.
When you type something into ChatGPT, it doesn't search the internet for an answer (unless you're using the web browsing feature). It generates a response word by word, with each word chosen because it's the most probable useful continuation of what came before.
It's a bit like autocomplete on your phone — but trained on so much text that it can complete entire essays, debug code, write poetry, explain quantum physics, or help you plan a holiday. Same underlying idea, vastly more capable.
This is why ChatGPT can do so many different things well. It's not switching between different specialised programs — it's applying the same underlying capability to wildly different tasks, because language underlies all of them.
Who makes ChatGPT?
ChatGPT is made by OpenAI, an American AI company founded in 2015. The GPT in the name stands for Generative Pre-trained Transformer — a technical description of how the model works, which you don't need to understand to use it effectively.
OpenAI also makes DALL-E (an image generator), Sora (a video generator), and Codex (a code-focused model). The most capable version of ChatGPT is called GPT-4o.
ChatGPT vs other AI tools
ChatGPT is the most widely used AI assistant in the world, but it's not the only one worth knowing. The main alternatives:
- Claude (by Anthropic) — Often described as feeling more like talking to a thoughtful person. Particularly strong at working with long documents, careful reasoning, and accurate responses. Many professional writers and researchers prefer it.
- Gemini (by Google) — Google's AI assistant, integrated with Google's services. Strong at tasks involving Google Docs, Gmail, and web search.
- Microsoft Copilot — Built on GPT-4, deeply integrated into Microsoft Office products. Useful if your work is primarily in Word, Excel, or Outlook.
All of these are genuinely capable for everyday tasks. The best one is whichever you'll actually use.
What can you use ChatGPT for?
The honest answer is: a lot. Here are some of the most useful everyday applications:
- Writing help — drafting emails, improving your own writing, creating reports, writing social media posts
- Research and summarising — getting a quick overview of a topic, summarising long documents, explaining complex concepts
- Planning — brainstorming ideas, thinking through decisions, creating schedules and outlines
- Learning — having concepts explained at whatever level you need, getting feedback on your work
- Creative tasks — writing, generating ideas, getting past creative blocks
- Everyday admin — drafting complaints, writing cover letters, preparing for interviews
What ChatGPT cannot do
Understanding the limitations is just as important as understanding the capabilities.
It makes things up
This is the most important limitation to understand. ChatGPT can confidently tell you things that are completely wrong. This is called a hallucination — and it happens because the model generates plausible-sounding responses, not necessarily accurate ones. It has no way to check whether what it's saying is true.
Never trust ChatGPT for facts that matter — medical decisions, legal questions, financial advice, or anything where being wrong has real consequences — without verifying against a reliable source. It's a brilliant assistant. It's a poor authority.
Its knowledge has a cutoff
ChatGPT was trained on data up to a certain date, so it doesn't know about events after that point. The free version has a knowledge cutoff; the paid version with web browsing enabled can search for current information.
It doesn't remember previous conversations
By default, each conversation starts fresh. ChatGPT doesn't remember what you told it last week. Some features let you save memory across conversations, but it's limited.
Is ChatGPT free?
Yes — the free tier of ChatGPT is genuinely capable and covers most everyday needs. ChatGPT Plus costs around £20/month and gives you access to the most powerful model (GPT-4o), faster responses, image generation via DALL-E, and web browsing.
Our honest recommendation: start with the free tier, build the habit, and only upgrade when you find yourself hitting the limits.
- ChatGPT is an AI trained on vast amounts of text to generate useful responses
- It predicts the most likely useful next word — it doesn't look up facts
- It can hallucinate — confidently saying things that are wrong
- The free tier is good enough for most people to start with
- It's one of several good options — Claude and Gemini are worth knowing too
Want to go from curious to confident?
Our AI Essentials course covers exactly this — what AI is, how to use it safely, and how to get genuinely useful results. 18 lessons, lifetime access, verified certificate.
View AI Essentials →