ChatGPT is a conversational AI assistant from OpenAI. You type a question or an instruction, it replies in ordinary language, and the conversation continues with it remembering what was said earlier in the thread.
You can start using it in about a minute at chatgpt.com. This guide covers what it is good and bad at, what the plans cost, and the mistakes that produce disappointing results.
Checked August 2026. This field moves quickly — plans, limits and model names change every few months. Check OpenAI’s own pricing page for what applies today.
Getting started
- Go to chatgpt.com, or install the app for iOS, Android, Windows or macOS.
- Create an account, or sign in with Google, Microsoft or Apple.
- Type into the box and press Enter.
There is nothing to configure. The interface is deliberately as plain as it looks.
What the plans include
As of August 2026, ChatGPT has a free tier and several paid ones — commonly Go, Plus, Pro, and Business or Enterprise for organisations. Plus, at around $20 a month, is the tier most individuals end up on.
Two things about the free tier are worth knowing:
- It carries advertising in the US. OpenAI began showing ads to free users in February 2026 — a genuine change from how ChatGPT worked for its first three years. Paid tiers are ad-free.
- It runs a smaller model than the paid tiers, with tighter limits on file uploads, images and other tools.
The current model family is GPT-5.6, offered in several variants of increasing capability depending on tier.
What it is genuinely good at
- First drafts. Emails, outlines, job descriptions, cover letters — anything where the blank page is the obstacle.
- Rewriting. Shorter, clearer, more formal, for a different audience.
- Summarising long documents into something someone will actually read.
- Explaining. An unfamiliar concept at whatever level you ask for.
- Code. Writing small functions, explaining unfamiliar code, and — often best of all — reading an error message and suggesting what it means.
- Translation, which is usually better than dedicated translation tools for anything with nuance.
- Structuring. Turning messy notes into an ordered plan.
What it is bad at
This half matters more, because the failures are confident rather than obvious.
It invents facts. Names, dates, statistics, court cases, academic citations and product specifications are all produced with complete confidence whether or not they are real. This is the single most important thing to understand about it. Anything checkable must be checked — and fabricated citations in particular have caused real professional consequences for people who did not.
- Arithmetic and precise counting are weak points, unless it uses a tool to compute.
- Recent events are unreliable unless it searches the web, which not every mode does.
- It agrees too readily. Push back on a correct answer and it will often change it. That is a property of the training, not a sign you were right.
- It does not know what it does not know, so there is no useful signal in how certain it sounds.
Getting better results
Most disappointment comes from prompts, not from the model.
| Instead of | Try |
|---|---|
| “Write about email marketing” | “Write a 300-word introduction to email marketing for a small bakery owner with no marketing background. Plain language, no jargon.” |
| “Fix my code” | Paste the code, the exact error, what you expected, and what happened. |
| “Summarise this” | “Summarise this into five bullet points for someone who will not read the original.” |
Three habits that help more than any prompt template:
- Say who it is for. Audience changes the answer more than any other instruction.
- Give it an example of what good looks like.
- Iterate in the same conversation. “Shorter”, “less formal”, “keep the second paragraph and redo the rest” all work, and are faster than rewriting the prompt.
Privacy: what to think about before pasting
Your conversations may be used to improve the models depending on your plan and settings. Practical guidance:
- Do not paste customer data, medical details, passwords, API keys, or anything under an NDA.
- Turn off training in Settings → Data controls if you would rather your chats were not used that way.
- Temporary Chat keeps a conversation out of your history.
- Business and Enterprise plans have different data commitments — the relevant ones if this is work.
Where else it appears
- Microsoft Copilot runs on OpenAI models and is built into Windows and Microsoft 365. See using Copilot on Windows 11 and ChatGPT and Copilot in Microsoft Word.
- Browser extensions add it alongside search results and web pages — with caveats worth reading first: see choosing a ChatGPT browser extension.
- Not on WhatsApp any more. That service closed in January 2026 — see what happened.
Frequently asked questions
Is ChatGPT free?
There is a free tier, which since February 2026 shows advertising in the US and uses a smaller model with tighter limits. Paid plans start below the well-known $20 Plus tier and remove the ads.
Does it have access to the internet?
It can search the web in the modes that support it, and will usually cite sources when it does. Not every response involves a search, so do not assume any given answer is current.
Can teachers or employers detect ChatGPT text?
AI-detection tools are unreliable in both directions — they miss AI text and flag human text, particularly from non-native English writers. Do not rely on them, and do not assume you are undetectable either. Check the rules where you are.
Will it replace jobs?
It changes tasks more than it removes roles, so far. It is fast at first drafts and summarising and unreliable at anything requiring judgement or verified fact — which is most of what people are actually paid for.
Is my data used to train it?
Depending on your plan and settings, it may be. Turn it off under Settings → Data controls, or use a Business or Enterprise plan with different commitments.
What are the alternatives?
Google Gemini, Anthropic’s Claude, and Microsoft Copilot are the main ones. They are close enough in capability that the deciding factor is usually which ecosystem you are already in.
2 comments
Comments are closed on this guide.