Complete guide to ChatGPT
Getting started
Go to chatgpt.com and create a free account with email or Google. When it opens, a new chat is already waiting for you: type your first question or instruction (called a "prompt") in the text box. You can type text, but you can also attach an image, an audio file or a document for it to analyze. You don't need to configure anything to get started: the more specific you are about what you ask for (what you want, for whom, in what format), the better the response will be. If the first response isn't exactly what you were looking for, keep the conversation going by asking for adjustments instead of rephrasing from scratch: ChatGPT keeps the context of the whole chat.
Use cases
Writing and editing text
Write drafts of emails, posts or reports, and ask ChatGPT to review, summarize or adapt them to a different tone (more formal, friendlier, shorter).
Coding and debugging
Paste a code snippet and ask it to explain it, optimize it or find the bug. It works with most programming languages.
Analyzing documents and data
Upload a PDF, Excel or Word file and ask for a summary, extraction of specific data or answers to questions about the content.
Generating images
Describe an image in detail (style, colors, composition) and ChatGPT generates it with GPT Image 2 built directly into the conversation (successor to DALL-E 3, retired in 2026).
Learning and studying
Ask it to explain a complex concept with simple examples, or to quiz you on a topic with exam-style questions.
Advanced setup
- Custom instructionsIn Settings → Personalization, fill in what ChatGPT should know about you (your role, your level, your goal) and how you prefer responses (tone, length, format). They apply automatically to all new conversations. Limit: 1,500 characters per field.
- MemoryUnlike custom instructions (fixed), memory learns details over time: your name, your projects, your preferences shown in past conversations. Turn it on in Settings → Personalization → Memory, and you can review or delete what it remembers at any time.
- ProjectsGroups conversations on the same topic, with files and instructions shared across all the chats in that project. Useful if you work continuously on something specific (a thesis, a product, a client) and don't want to repeat context every time.
- Choosing the right modelGPT-5.5 Instant is fast and free, designed for everyday use. Reasoning models (o3, o4-mini) take longer but think step by step, which makes them better for math or complex problems. Switch models from the selector at the top depending on the task.
Tips
- Tell ChatGPT what you DON'T want, not just what you do want: "don't use bullet lists" or "avoid corporate language" improves the result far more than giving only positive instructions.
- Use the PTCF framework (Persona, Task, Context, Format) when writing a prompt: tell it who it should be, what it needs to do, with what context, and in what format you want the response.
- If you need to keep a fixed style or set of rules across many conversations, put them in Custom instructions instead of repeating them every time.
- For tasks that require a lot of thinking (math, logic, complex code), switch to the reasoning model instead of the fast default model.
- When you upload a long document, ask for a general summary first before asking specific questions: it helps make later answers more accurate.
FAQ
Is ChatGPT free to use?
Yes, it has a free plan with unlimited text chats and access to GPT-5.6 Luna (the fast default model since August 2026, which replaced GPT-5.5 Instant). Limits still apply to file uploads, images and other tools. The Plus plan (from $20/month) adds more advanced models and image generation with fewer restrictions.
How do I cancel my subscription?
From your profile icon → Settings → Billing → Cancel plan. Important: cancel at least 24 hours before the renewal date to avoid being charged for the next month, and note that if you subscribed through the App Store or Google Play, you must cancel there, not on the web.
Can ChatGPT make mistakes?
Yes. It can generate incorrect information with apparent confidence ("hallucinations"), especially with very recent or very specific data. Always verify critical information against a reliable source.
Are my conversations used to train the model?
By default they may be used unless you turn that off in Settings → Data → Improve the model for everyone. Team and Enterprise plans do not use your data for training by default.
Based on help.openai.com and openai.com/academy