Quick Answer
ChatGPT raised its custom instructions limit to 5,000 characters, up from 1,500, for Plus, Pro, Enterprise, Business, and Education users, while Free and Go users keep 1,500. Custom instructions are a standing brief that shapes every answer, so the extra room lets you state your role and output format once. Set them under Settings, then Personalization.
Key Takeaways
- The limit rose from 1,500 to 5,000 characters on July 15, 2026, for paid ChatGPT tiers, while Free and Go stay at 1,500
- Custom instructions apply to every new chat automatically, unlike a prompt you type once
- The best use is stating your role, your format preferences, and what to avoid, not padding for its own sake
- Keep sensitive personal, health, and work data out of the box, since it is stored and reused
- More characters help only if they are specific; vague filler can make answers worse, not better
What changed with ChatGPT custom instructions?
ChatGPT tripled the space available for custom instructions, raising the limit from 1,500 to 5,000 characters. OpenAI rolled the change out on July 15, 2026 for Plus, Pro, Enterprise, Business, and Education users, per its release notes, giving paid subscribers far more room to shape how the assistant responds. Free and Go users can still save up to 1,500 characters, according to OpenAI’s custom instructions FAQ.
Custom instructions are a standing personalization setting, not a one-off message. They sit in the background and apply to every new conversation automatically, which is what makes them different from a prompt you type into a single chat. Set them once, and every future answer follows them without you repeating yourself.
The old 1,500-character ceiling forced hard choices, since a detailed role description and a list of format preferences could not both fit. At 5,000 characters, roughly the length of a page of text, you can describe your work, your preferences, and your rules for output in one place. The catch, covered below, is that more room only helps if you fill it with specifics rather than padding.
What are custom instructions, and where do you set them?
Custom instructions, documented in OpenAI’s help center, tell ChatGPT what it should know about you and how you want it to respond. OpenAI’s help center describes a single Custom Instructions field in the personalization settings, along with an Enable customization toggle that must be on for the instructions to apply. The instructions act as a persistent brief applied on top of whatever you ask in the moment, and they are available on all ChatGPT plans on web, desktop, iOS, and Android.
Setting them takes under a minute:
- Open ChatGPT and select your profile, then Settings.
- Open Personalization on web and desktop, or Customize ChatGPT on iOS and Android.
- Make sure Enable customization is toggled on.
- In the Custom Instructions field, describe your role and context, then state how you want responses formatted, toned, and structured.
- Save, and the instructions apply to every new chat from then on.
Separating context from behavior still matters inside a single field. A short section on what the assistant should know about you, followed by a section on how it should respond, keeps the two kinds of guidance from blurring together and tends to produce cleaner results.
What should you put in the extra space?
Use the room for specifics that change how answers come out, not for a biography. The goal is to remove the instructions you find yourself retyping in chat after chat. If you keep telling ChatGPT to skip the preamble, that belongs in custom instructions permanently.
These are the categories worth including:
- Your role and field, so answers assume the right level of expertise
- Output format, such as short bullet points, no preamble, or code with comments
- Tone, such as direct and concise, or warm and detailed
- Things to avoid, such as filler openings, hedging, or repeating the question back
- Recurring context, such as the tools you use or the audience you write for
The single highest-value entry for most people is a clear statement of what to stop doing. Telling ChatGPT to skip long introductions and get to the point, once, in custom instructions, fixes a recurring annoyance across every future chat. Our guide to ChatGPT prompts for work covers the kinds of task-specific instructions that pair well with a strong standing brief.
What should you keep out of custom instructions?
Keep sensitive information out, because custom instructions are stored and reused, not typed and forgotten. Anything you put in that box persists on your account and is fed into your conversations, so it deserves the same caution as anything else you would not want stored. OpenAI also states that information from custom instructions can be used to improve model performance unless you opt out of model training in your data controls.
Specifically, leave out full names of others, financial details, health specifics, account numbers, passwords, and confidential work information. The risk is not that a single answer leaks, but that sensitive data sits in a persistent setting and flows into every session. This is the same principle behind keeping regulated data out of consumer AI generally, which we cover in our guides to whether ChatGPT Health is safe and to broader cybersecurity and data privacy.
There is also a sharing angle worth knowing. OpenAI states that your instructions are not shared with people who view a shared link, but the instructions still shape the answers those viewers can read. Shared AI chats have turned up in search results before, an issue we covered in our piece on shared AI chats appearing in search. Treat the custom instructions box as durable, reusable text, and write it accordingly.
Does more space actually make answers better?
Only if the extra characters are specific, and this is the counterintuitive part. More instructions do not automatically mean better answers. Vague or contradictory filler can pull responses in competing directions and make them worse than a short, sharp brief would.
The reason is that every instruction competes for the model’s attention. A tight 500-character brief that says exactly what you want will usually beat a rambling 5,000-character one full of hedged preferences. So treat the higher limit as a budget you do not have to spend. Use the room when you genuinely have distinct, specific things to say, and stop when you have said them.
A practical test helps here. Read your instructions back and ask whether each line would change an answer if removed. Lines that would not change anything are padding, and padding dilutes the lines that matter. Quality of instruction beats quantity every time, which is why a thoughtful brief outperforms a maxed-out one.
How do custom instructions fit with memory and projects?
Custom instructions are one of several personalization layers, and knowing how they differ keeps you from duplicating effort. ChatGPT also has a memory feature that learns preferences over time, and Projects that hold instructions scoped to a specific body of work.
The distinction is about scope, and this table shows where each one fits:
| Feature | Scope | Best for |
|---|---|---|
| Custom instructions | Global; every chat | Rules you want everywhere, like format and tone |
| Memory | Automatic; learned over time | Preferences ChatGPT picks up as you use it |
| Project instructions | One project only | Rules tied to a single client, topic, or codebase |
A rule you want everywhere, like your preferred format, belongs in custom instructions. A rule that applies to one client or codebase belongs in that project. Putting a project-specific rule in the global box is a common way to get instructions that fight each other.
Two newer controls change how these layers interact, according to OpenAI’s release notes. Since August 27, 2026, you can start a temporary chat as personalized, which lets it use your memories, custom instructions, and plugins without creating new memories; temporary chats remain non-personalized by default. Since August 14, 2026, eligible unshared projects can also switch between default and project-only memory without starting a new project, which helps when a project’s context should stay separate from the rest of your account.
For getting more from ChatGPT overall, our guides to searching your ChatGPT history and whether AI actually makes you more productive cover the features and habits that compound with a good set of instructions.
Is this worth setting up if you use several AI tools?
Yes, and it is worth doing on each tool you use regularly, because the payoff repeats with every chat. Most people now use more than one assistant, and each has its own version of standing instructions or personalization. Setting them once per tool removes the same friction across all of them.
The habit that pays off is writing one clear brief and adapting it to each assistant’s settings, rather than starting fresh in every chat. The specifics differ by platform, but the principle is constant: tell the tool who you are and how you want it to respond, once, so you stop repeating it. For how the major assistants compare on everyday tasks, our Claude versus ChatGPT versus Gemini overview is a useful starting point, and students in particular may want our roundup of the best AI tools for students.
FAQ
How many characters can ChatGPT custom instructions hold now?
ChatGPT custom instructions now hold up to 5,000 characters for Plus, Pro, Enterprise, Business, and Education users, raised from 1,500 on July 15, 2026, while Free and Go users can still save up to 1,500. The extra room lets you state your role, preferences, and output format in one place instead of repeating them in each chat.
Where do I set custom instructions in ChatGPT?
To set custom instructions in ChatGPT, open Settings, then Personalization on web and desktop or Customize ChatGPT on iOS and Android, and make sure Enable customization is toggled on. Enter what ChatGPT should know about you and how you want it to respond in the Custom Instructions field, and the instructions then apply automatically to your conversations.
What should I put in ChatGPT custom instructions?
Put your role, your preferred output format, your tone, and things to avoid such as long preambles or hedging, since the most useful entry for many people is a clear statement of what to stop doing. Focus on instructions you find yourself repeating in chat after chat, because those benefit most from being permanent.
Is it safe to put personal information in custom instructions?
Keep sensitive information out, since custom instructions are stored and reused across every conversation rather than typed once. Leave out financial details, health specifics, account numbers, passwords, and confidential work data, because that data would otherwise sit in a persistent setting and flow into every session.
Do longer custom instructions make ChatGPT better?
Only if they are specific, because more characters do not automatically improve answers and vague filler can make them worse by pulling responses in competing directions. A short, sharp brief usually beats a rambling one, so treat the 5,000-character limit as a budget you do not have to spend.