ChatGPT Custom Instructions: Complete Setup Guide

Last Updated: September 22, 2026

Summary: Custom Instructions let you set ChatGPT preferences once instead of repeating them every conversation. Two fields control what it knows about you and how it responds. A July 2026 update expanded the character limit from 1,500 to 5,000 per field, more than triple the old cap, letting existing setups add real detail that wouldn't have fit before.

ChatGPT Custom Instructions let you set persistent preferences once. You don't have to repeat them in every new conversation. Tell it your job. Tell it your preferred tone. Tell it what to avoid. It remembers automatically, going forward. Here's exactly how to set it up. This covers what actually goes in each field, plus a real change from mid-2026 that most existing guides haven't caught up with yet.

💡 Not sure which AI tool is actually right for your business?

Get the free guide, Which AI Tool Should You Actually Use? — a straightforward breakdown of the leading AI tools to help you pick the right one for your needs.

Subscribe to AI Business Weekly for the guide, plus daily coverage of AI trends, acquisitions, and product launches.

Get the Free Guide →

Custom Instructions solve a specific, recurring annoyance: repeating the same context in every new conversation. Without them, explaining your job, your preferred tone, and your formatting preferences becomes a repeated task at the start of nearly every chat. With them set up once, ChatGPT already knows this by the time you type your first message.

Setting up Custom Instructions takes about two minutes. It lives in the same place across every ChatGPT plan. Click your profile picture in the bottom-left corner. Then select Settings. From there, go to Personalization. Then click Custom Instructions. Two text boxes appear once you get there. One asks what you'd like ChatGPT to know about you specifically. The other asks how you'd like it to respond to you going forward.

Fill in both fields. Then save. Every new conversation from that point on automatically applies these preferences. You don't need to repeat them. This differs from a single conversation's memory. That resets each time you start a new chat. Custom Instructions persist across every conversation, until you go back and change them.

The first field asks what ChatGPT should know about you. It works best with real, specific context, not vague generalities. Useful details include your profession. Your experience level with topics you frequently ask about. Any specific interests or focus areas relevant to how you actually use ChatGPT. Keep this factual and relevant. It doesn't need to be a full biography.

A marketing manager might write that they work in B2B SaaS marketing. They could add that they want examples relevant to that industry specifically. A developer might note their primary programming language and experience level. That way, code suggestions land at the right complexity from the first response. Specific, concrete details in this field consistently produce better results. A generic line like "interested in technology" doesn't do much.

The second field asks how ChatGPT should respond. This controls tone, length, and format, not subject matter. This is where you specify a preferred tone. Whether you want brief or detailed answers by default. Specific formatting preferences. Any topics you'd rather it avoid bringing up unprompted.

A common, effective example: ask for concise responses that use bullet points over long paragraphs. Skip unnecessary caveats or disclaimers unless the topic genuinely requires one. These preferences apply automatically across every future conversation. You save the repeated effort of restating the same formatting request every single time.

Here are a few real example instructions that work well, broken out by field, to make the two-box structure concrete.

Notice the pattern across all four examples. Each one is specific enough that ChatGPT can actually apply it consistently, rather than a vague preference that's open to interpretation on every single response. Vague language leaves too much room for guessing, and guessing is exactly what these fields are meant to eliminate.

Here's a real, recent change most existing guides on this topic haven't updated for yet. The character limit for Custom Instructions expanded significantly on July 14, 2026. Each field previously capped at 1,500 characters. It now allows up to 5,000 characters per field. That applies across Plus, Pro, Business, Enterprise, and Education accounts.

That's more than triple the previous limit. It meaningfully changes what's actually possible here. The old 1,500-character cap forced genuinely tight, compressed instructions, every time. The new 5,000-character limit allows much more detailed context. You can include multiple examples of preferred output. You can add more nuanced tone guidance. You can spell out specific edge cases you want handled a certain way. If you set up Custom Instructions before July 2026 and haven't revisited them since, there's real room now. You can add detail that simply wouldn't have fit before.

A few common mistakes show up repeatedly when people set these fields up for the first time. Writing vague preferences like "be helpful" or "be smart" wastes the field entirely, since ChatGPT already tries to do both by default. The instruction needs to specify something ChatGPT wouldn't already do on its own.

Overloading the fields with contradictory instructions is another common issue. Asking for both "always be extremely detailed" and "keep responses short" in the same field creates a genuine conflict ChatGPT has to guess how to resolve. Pick one clear direction per preference instead of layering in competing requests that pull against each other.

Trying to cram too many unrelated preferences into one field is a related, third mistake worth avoiding. If you find yourself listing five or six separate, disconnected rules in a single box, consider whether some of those preferences might actually belong in a dedicated Custom GPT instead, built specifically around that narrower combination of rules.

Custom Instructions and Custom GPTs solve related but genuinely different problems. Mixing them up means setting up the wrong one for what you actually need. Custom Instructions apply globally. They cover every conversation in your account, all the time. A Custom GPT is different. It's a separate, standalone setup you build for one specific repeated task, like a particular writing style for one project. It stays kept apart from your everyday conversations.

If you want ChatGPT to consistently know your job and preferred tone everywhere, Custom Instructions is the right tool for that. If you need a completely different, specialized setup for one narrow recurring task, without touching your default experience everywhere else, a Custom GPT fits better. Many people use both together. Custom Instructions handle baseline preferences. One or two Custom GPTs cover specific, narrower jobs on top of that.

Custom Instructions sync automatically across the web app, the desktop app, and the mobile app, since they're tied to your account rather than any single device. Set them up once on your phone, and they apply the next time you open ChatGPT on a laptop, with no extra setup needed on the second device.

This also means changing or removing Custom Instructions on one device updates everywhere else immediately. There's no separate configuration to manage per platform, which is a genuine convenience compared to settings that only apply on one specific app.

One small exception worth knowing: Custom GPTs you've built or saved don't automatically carry your Custom Instructions into them unless you explicitly configure that GPT to reference your profile information. Treat the two features as related but separately managed, even though your baseline Custom Instructions sync everywhere your regular ChatGPT conversations do.

Custom Instructions differ from Memory, ChatGPT's other personalization feature. Confusing the two leads to duplicated or conflicting settings. Memory builds up gradually. It picks up things you mention in regular conversations, like a project you're working on, or a preference you stated once in passing. Custom Instructions work differently. They're deliberate, written upfront, and they don't change unless you edit them directly yourself.

Both features work together. Neither one replaces the other. Memory picks up context you never explicitly stated as a formal instruction, almost by accident. Custom Instructions cover the stable, foundational preferences. These apply every time, no matter what's come up in recent conversations. Check both settings every so often. That avoids a situation where Memory has picked up something that now contradicts what you wrote in Custom Instructions.

Custom Instructions work well for setting durable preferences. They're not the right tool for a one-off request, though. Worth remembering before you overload the fields. If you only need a specific tone or format for a single conversation, just ask for it directly in that chat. Don't update your global settings for something temporary.

Review your Custom Instructions every so often, especially if your job or primary use case has changed since you first set them up. Preferences that made sense six months ago, written under the old 1,500-character limit, may no longer reflect how you actually use ChatGPT today. The expanded field length now gives you real room to update them properly.

A quick way to check whether your current instructions still make sense: read them fresh, as if a stranger had written them for you, and ask whether they'd still make sense given how you actually use ChatGPT right now. A lot of instructions written months ago reference an old job, an old project, or a preference that's simply changed since then without anyone updating the settings to match.

Custom Instructions let ChatGPT remember your job, tone, and formatting preferences across every conversation, set up once through Settings, then Personalization. A July 2026 update expanded each field from 1,500 to 5,000 characters, more than triple the old limit, giving real room for detailed context that didn't fit before. They differ from Custom GPTs (task-specific setups) and Memory (context picked up automatically), and work best when reviewed periodically as your needs change.

How do I set up Custom Instructions in ChatGPT?
Click your profile picture, select Settings, then Personalization, then Custom Instructions. Two fields appear: one for what ChatGPT should know about you, and one for how it should respond. Fill in both and save. Every new conversation automatically applies these preferences afterward, without needing to repeat them each time.

What is the character limit for ChatGPT Custom Instructions?
Each field allows up to 5,000 characters as of July 14, 2026, up from a previous 1,500-character limit, across Plus, Pro, Business, Enterprise, and Education accounts. This is more than triple the old cap and allows much more detailed context, including multiple examples and specific edge cases, than was previously possible.

What is the difference between Custom Instructions and Custom GPTs?
Custom Instructions apply globally across every conversation in your account. A Custom GPT is a separate, standalone setup built for one specific repeated task, kept apart from your everyday conversations. Most people use both together: Custom Instructions for baseline preferences, and one or two Custom GPTs for narrower, specific jobs.

Do Custom Instructions sync across devices?
Yes. Custom Instructions are tied to your account, not a single device, so they sync automatically across the web app, desktop app, and mobile app. Set them up once on any device and they apply everywhere else immediately, with no separate configuration needed per platform.

What is the difference between Custom Instructions and Memory in ChatGPT?
Memory builds up gradually from things mentioned in regular conversations, almost automatically. Custom Instructions are deliberate and written upfront, changing only when edited directly. Both work together rather than replacing each other, covering stable, foundational preferences versus context picked up incidentally along the way.

By Sameer Khan

This article was AI-assisted, then reviewed by Sameer Khan before publishing.

Sameer Khan is the founder of AI Business Weekly. He has a background in research and advisory, working with HR leaders and executives across Canadian public-sector and enterprise organizations on research and AI adoption. He holds an MBA from the Ted Rogers School of Management and has spent nearly a decade in B2B sales across SaaS, research and advisory, and AI.