ChatGPT Basics

What ChatGPT Is

ChatGPT is an AI assistant built by OpenAI. You type what you need in plain language and it answers, and you keep going from there. OpenAI describes it as an assistant for everyday tasks like brainstorming, writing, studying, planning, math, coding and analysing images and files, trained to follow your instructions in a conversation.

You can use it on the web at chatgpt.com and in the iOS and Android apps, and in supported regions you can try it without an account at all. Signing in is what unlocks the features most of this page describes, because nearly all of them are attached to your account rather than to the browser tab.

It helps to know what it is not. It is not a search engine that hands you ten links, and it is not a database looking up an answer somebody stored earlier. It reads what you give it and reasons about it, which is why supplying the right material matters more than phrasing the question perfectly.

Chats, and How a Conversation Works

Everything happens inside a chat. ChatGPT keeps context within a chat, so follow-up questions build on what was already said and you can ask "make it shorter" without repeating yourself. Start a new chat when you change subject — a fresh chat is a clean slate, and a long chat that wanders gives worse answers rather than better ones.

Chats are saved in your history until you delete them, so treat each one as a container for a distinct task: one for a document you are drafting, one for a topic you are learning, one for a bug you are chasing. You can also share a chat as a read-only link, and you can invalidate that link later from the share dialog — deleting the chat itself is a separate action.

Plans

There is a free tier, and it is a real product rather than a trial. Paid plans — Go, Plus and Pro for individuals, Business and Enterprise for organisations — mainly buy higher limits and earlier access to the newest models and tools. Message limits vary by plan and by model and change over time, so the honest way to think about a plan is "how much and how new", not "which features exist".

ChatGPT Go is the low-cost plan, and it is worth knowing it is not available everywhere. Its availability is a per-country list, and India is on it.

Custom Instructions

Custom instructions are where you tell ChatGPT things you would otherwise retype every time — who you are, what you are working on, how you want replies formatted. They are available on all plans, on web, desktop, iOS and Android, and OpenAI states they are applied immediately to all chats.

Two details catch people out. The text fields cap at 1,500 characters, so this is a short standing brief and not a document. And updates apply to future conversations only — changing your instructions does not rewrite answers you have already received.

Memory

Memory is the other half of personalisation, and it works differently. With memory on, ChatGPT remembers useful context from your chats to personalise later responses, so you are not reintroducing yourself every time. You can read what it has retained in the memory summary, edit it by typing the correction, or delete it.

The distinction from custom instructions is the one to hold on to: custom instructions are what you state explicitly, memory is what ChatGPT picks up from conversations. One is a brief you write; the other is a record it keeps.

Deleting a memory is not the same as deleting its source. To fully remove something ChatGPT may know about you, you have to delete every place it appears — past chats, archived chats, files and the memory summary itself.

Projects

A project is a workspace that keeps a long-running effort in one place: grouped chats, uploaded reference files, and project instructions that apply inside that project only and override your global custom instructions. Projects are available on free and paid plans, and you must be signed in to use them.

Use one whenever you will return to the same subject repeatedly — a syllabus you are studying, a report you are drafting, a launch you are planning. The payoff is that you upload the material once instead of re-attaching it to every new chat.

File Uploads

Attaching a file changes the kind of question you can ask. Instead of describing a document you supply it, and OpenAI groups what follows into three jobs: synthesis (combining or analysing material to make something new), transformation (reshaping it without changing its essence, such as summarising a research paper in plain terms) and extraction (pulling out specific details, quotes or sections).

Two limits are worth remembering because you will meet them. Free accounts are limited to three file uploads per day. And every file has a hard ceiling of 512MB, whichever plan you are on.

The Tools Around the Chat

• Search. ChatGPT can search the web and cite sources when it needs current information.
• Deep research. For multi-step questions, it proposes a research plan you can edit, then returns a structured report with citations you can check.
• Canvas. A side-by-side workspace for writing and coding, where you can highlight a section to say exactly what to work on, edit the text yourself, and restore earlier versions.
• Images. You can generate an image from a description and edit an existing one by describing the change.
• Voice. You can speak to ChatGPT and hear the reply, inside the same chat, so you can listen while reading along.
• Study mode. Instead of handing over a final answer, it asks questions, explains in layers and checks your understanding. It is on all plans, and for exam preparation it is the mode that matters.

Privacy and What You Control

Data Controls let you decide whether your conversations are used to improve OpenAI's models. Turning that off is one toggle, it applies to your whole account across devices, and your chats still appear in your history — the setting governs training, not storage.

Temporary Chat is the stronger tool for a single sensitive conversation. A temporary chat does not appear in your history, does not use or create memories, is not used to train models, and is deleted after 30 days.

What to Check

Treat ChatGPT as a fast, well-read first draft rather than a final authority. OpenAI's own guidance is direct about it: ChatGPT can make mistakes, and you should check important information — especially anything date-, time- or location-sensitive.

Two habits do most of the work. Verify what matters — legal, medical, financial, or anything you are about to publish — against a source you trust, and prefer answers that cite one. Give it the material instead of relying on recall, because a document you supply cannot be misremembered.

One more thing to expect: models change. OpenAI periodically retires older models as newer ones arrive, so the voice and style of replies can shift under you. Your prior context carries over, but a workflow tuned to one model's habits may need a nudge.

Getting Better Answers

OpenAI's own prompting guidance comes down to two habits. Be clear and specific — give enough context, avoid ambiguity, say who the answer is for and what format you want. And refine iteratively — start with a prompt, read the response, then adjust the wording, add context or simplify. The first answer is a draft.

Where to Go Next

Prompt Engineering Fundamentals goes deeper on writing prompts, and Advanced Prompting Techniques further still. GPT Models Overview covers the models behind the assistant. If you are comparing tools, Claude.ai Basics covers the same ground for Anthropic's assistant and Gemini Basics for Google's. AI Agents Basics covers what changes once an assistant is given tools and left to work.

Sources

Official OpenAI Help Center pages. What is ChatGPT: FAQ · ChatGPT Capabilities Overview · Memory FAQ · Projects in ChatGPT · ChatGPT Custom Instructions · Data Controls FAQ · File Uploads FAQ · Using Study Mode in ChatGPT · Deep research in ChatGPT · ChatGPT Voice · What to expect when models change