Custom instructions: set up your AI assistant once and for all
Custom instructions are preferences you save in an AI assistant's settings, such as your job, the tone you like and how answers should be formatted, which it then takes into account in every new chat. They spare you from repeating the same details each time.
Data as of October 22, 2026
Key figures
| Level | Beginner |
|---|---|
| Origin | 2023 |
| Introduced by | OpenAI (fonction « Custom instructions » de ChatGPT, annoncée le 20 juillet 2023) |
How it works
Every new chat starts from scratch: the assistant knows nothing about you. Custom instructions fill that gap. A short piece of text, written once, is quietly added behind the scenes to every request you make. Technically, it works much like a system prompt that you write yourself.
OpenAI popularized the feature when it announced it for ChatGPT on July 20, 2023 (official announcement), with two boxes to fill in: what the assistant should know about you, and how you would like it to respond. Other AI assistants have since added similar settings.
Real-world examples
A fourth-grade teacher might write: “I teach fourth grade. Suggest activities that take 20 minutes or less, using vocabulary suitable for 9- and 10-year-olds.”
A self-employed plumber: “I run my own plumbing business. When I ask for an email, keep it polite and short, skip the jargon and end with a placeholder for my contact details.”
For personal use: “Use short bullet points. If you’re not sure about something, say so clearly. I’m vegetarian, so keep that in mind for recipes.”
The prompt generator can help you phrase these instructions clearly; then you just paste them into your settings.
Why it matters to you
Well-written custom instructions save time and make answers more relevant on the first try. They complement the prompt you type in the moment rather than replacing it: the instruction sets the frame, the request defines the task.
They also take up a little space in the context window, the amount of text the model can read at once. A short, precise note works better than a long rulebook.
Common mistakes
- Trying to cover everything. A list of thirty rules quickly contradicts itself. Keep it to the essentials and revisit it now and then.
- Confusing it with memory. An assistant’s memory picks things up over time; your instructions only change when you edit them.
- Forgetting they apply everywhere. A work-only rule like “always answer in formal business English” will also apply to your personal questions. Adjust or switch it off when needed.
- Including sensitive data. Instructions are stored with your account, so only include what you would be comfortable having stored by the provider.
Related terms
- Prompt · The text, question or instruction you give an AI to get a response; how clear it is directly shapes the result.
- System prompt · Background instructions given to a model before the conversation starts (role, tone, rules) that shape every answer that follows.
- AI memory · A feature that lets an AI assistant remember facts about you from one conversation to the next.
- AI assistant · A consumer app such as ChatGPT, Claude or Gemini that answers, writes, summarizes and helps with everyday tasks using AI.
- Context window · The maximum amount of text (in tokens) a model can read and produce at once, including its own answer.
Try it
Full glossary
Frequently asked questions
Where do I find custom instructions?
In ChatGPT, they live in your account's personalization settings. Other assistants offer similar options under different names, such as preferences or saved info. The exact location changes as apps are updated.
How are they different from the assistant's memory?
You write custom instructions yourself and stay fully in control of them. Memory, where available, is filled in by the assistant based on your conversations. The two can work side by side.
What should I leave out of my instructions?
Anything sensitive: account numbers, passwords, detailed health information or details about other people. Stick to what improves the answers: your role, your level, tone and format.
Will the assistant always follow my instructions?
Not always to the letter. It takes them into account, but a specific request in the chat can override them, and it may apply them imperfectly. If something really matters, repeat it in your message.
Data sources
Data as of October 22, 2026
Report an error
Does a figure, a date or a fact look wrong? Tell us: every report is checked.