ChatGPT brand guidelines mean giving ChatGPT the brand's rules one of two ways: paste a design brief into its custom instructions, or connect a source it reads on its own, such as a design brief served over an MCP connection. The first works inside a single account. The second reaches the whole team and keeps itself current without anyone retyping anything.
Most teams back into the first method without choosing it. A marketer pastes a hex value and a line about tone into a new chat because it is faster than finding the style guide, ChatGPT turns out a passable draft, and three weeks later a coworker types the exact same thing into their own account, because there was never anywhere else for it to live. Nobody decided to skip the shared brief. It just never got written anywhere but that one chat.
This post covers both: what belongs in that brief, and how to connect it so ChatGPT reads it on its own instead of waiting for someone to paste it in.
Key Takeaways
- There are two ways to give ChatGPT your brand guidelines: paste a design brief into custom instructions, or connect a source over MCP so ChatGPT reads it before it writes.
- Custom instructions in ChatGPT carry over between conversations in one account, but someone still has to write and update them by hand.
- TrueKeep connects ChatGPT, Claude, and Cursor to a portal through the Model Context Protocol, an open standard, so one brief reaches every account on the team.
- A brief that names a banned word, a tone rule, and a logo spec gives ChatGPT the same information a style guide gives a new hire.
- A brief only helps if someone keeps it current. ChatGPT reads whatever version it finds, including an outdated one.
Two ways to give ChatGPT your brand guidelines
The first way is free and takes two minutes. Open ChatGPT's settings, add the brand rules under custom instructions, and ChatGPT applies them in every new chat on that account. It is a real improvement over pasting rules into a single conversation, and it stops at the edge of that one account. The next person on the team starts from a blank settings page.
The second way asks for one thing up front: a connection ChatGPT can read on its own, instead of a page someone filled in by hand. TrueKeep sets this up through an MCP connection, and once it is in place, ChatGPT reads the brand's strategy, guide, and design brief before it drafts anything, the same brief for every account on the team.
Custom instructions are right for one person working alone, on one account, who will not ask a teammate to match the result. A connection is right the moment more than one person prompts ChatGPT for brand work, because then the brief is written and kept current exactly once, not once per person.
What belongs in a brief ChatGPT can use
ChatGPT writes what the prompt tells it to, not what it guesses you meant, so a vague instruction gives it nothing concrete to act on. "Sound premium" could mean anything to anyone. A rule like the one below settles the question before ChatGPT writes a word:
Words we avoid: leverage, synergy, disrupt.
Use instead: use, combine, change.
Formality: warm but not casual, first names only after an introduction.
Color: use the dark teal background only for hero sections.
Four lines like that decide more drafts than a paragraph about personality. Shopify's Polaris content guidelines are a public example of a brand voice section written at this level of detail, down to how it handles buttons and error messages.
A new hire gets this same information in a style guide and can still ask someone to clarify "premium." ChatGPT only gets what is written down.
Connect ChatGPT to your brand guidelines over MCP
Custom instructions still live inside one account, so a new hire, a new device, or a cleared settings page means writing them again. An MCP connection skips that account boundary entirely. TrueKeep serves the portal to ChatGPT, Claude, or Cursor through the Model Context Protocol, an open standard that lets an AI tool read a source of information directly, the same connection no matter which account is asking.
Once connected, ChatGPT reads the strategy, the guide, and the design brief on its own, before it drafts a word, and a final pass checks the result against the brand afterward. Setup happens once, for the whole team, instead of once per account. The AI connection guide walks through those steps.
Custom instructions versus a live connection
Custom instructions and an MCP connection solve the same problem at different scales. Custom instructions live in one person's ChatGPT account, so they cover that person's chats and nobody else's. A team of 10 people doing brand work needs 10 copies of the same text kept in sync by hand, or one connection that reaches all 10 accounts at once.
The tradeoff runs the other way too: custom instructions take minutes to write and need no setup beyond a settings page, while a connection needs that one setup step before it helps anyone. For a single marketer writing alone, the settings page is often enough. For a team, the connection pays for itself the first week two people ask ChatGPT for the same thing and get the same answer.
Before and after: what changes in ChatGPT's answer
Ask ChatGPT for a product announcement with no brand context, and it reaches for the words every brand announcement reaches for:
Before: "Our platform empowers you to revolutionize how your team collaborates."
After: "Comments now show up next to the line they're on, not in a separate tab."
Nothing about the writing improved on its own. ChatGPT had a brief this time, one that named the banned words and showed the replacement, and it followed it. Same request, same model, different answer, because only one version carried the rules with it.
The tradeoff: a brief only works if someone keeps it current
A connection makes sure ChatGPT finds the brief. It does nothing to check whether the brief is still true. ChatGPT reads whichever version sits in the portal and answers with full confidence, even if that version is a year out of date.
A retired logo variant, a dropped tagline, or a word the company quietly stopped using can sit in a brief for months with nobody noticing, until someone spots ChatGPT still writing with the old one. The same fix applies here as to any brand document: give the brief one named owner with a reason to check it regularly. TrueKeep's edit mode lets that person update the guide directly, and the design brief template covers what the ownership line should say.
FAQ
How do I give ChatGPT my brand guidelines? Paste the written rules into custom instructions for that account, or connect a source ChatGPT can read on its own, such as a design brief served over an MCP connection. The second method reaches every account on the team without anyone repeating the paste.
What are ChatGPT custom instructions? Custom instructions are a settings field where a person tells ChatGPT how to respond, and ChatGPT applies that guidance in every new chat on that account. They carry over between conversations, but only for the account that wrote them.
Can I just paste my brand guidelines into ChatGPT each time? You can, and ChatGPT will use it for that conversation. The rules do not travel past the account they were typed into, so a teammate on a different login, or a new chat after a settings reset, starts over.
What is an MCP connection? MCP, the Model Context Protocol, is an open standard that lets an AI tool read a source of information on its own, instead of waiting for someone to paste it in. TrueKeep connects ChatGPT, Claude, and Cursor to a brand portal this way.
What should a brief for ChatGPT include? The words the brand avoids and what to use instead, a line on how formal to sound, and whatever logo or color rule applies to the request at hand. A list like that gives ChatGPT something concrete to follow; an adjective like "premium" does not.
Does TrueKeep offer a free way to connect ChatGPT? No. TrueKeep is one flat monthly price for the whole organization, with no free plan. See pricing for the current numbers.
Give ChatGPT a brief it can actually use
ChatGPT brand guidelines come down to two choices: write the rules into one account's custom instructions, or connect a source that ChatGPT reads on its own across every account. Either way, write specifics instead of adjectives, since "premium" gives ChatGPT nothing to act on and a banned-word list does. Name who owns the brief, because ChatGPT will follow an outdated one exactly as confidently as a current one.
If you want the connection instead of the paste, get started with TrueKeep.