How to Stop Repeating Your Brand Guidelines in Every AI Prompt

By /Founder, Daoco

Ben House builds Daoco and writes about brand-aware AI workflows, campaign execution, and practical systems for small marketing teams.

Store your brand guidelines once, where the AI reads them on every request, and keep only the new part of each task in the prompt. Four ways to do it, from free to full.

Key takeaways

  • Store what changes slowly: audience, voice with examples, banned words, product facts, and formatting habits.
  • Leave in the prompt only what is new: the goal, the channel, the offer, and the main constraint.
  • In the example, 30 of the prompt's 40 words are brand context that had not changed.
  • Instructions and projects store guidelines but do not publish. A marketing tool with brand memory does both.

A small marketing team stops repeating brand guidelines in every AI prompt by storing them once, in a place the AI reads on every request, and keeping only the new part of each task in the prompt. There are four places to store them: account-wide instructions, a project, a shared assistant, or a marketing tool with its own brand memory.

Which one fits depends on two things. How many people need the same guidelines, and whether the content has to be published from the same place it was written.

Why the pasting happens

A chat assistant starts each conversation knowing nothing about your brand. So the prompt has to carry the tone, the audience, the words you avoid, the product facts, and the request itself. Only the last part is new each time.

The cost is not the typing. It is the drift. One person pastes last quarter's tagline. Another skips the list of banned words because the prompt was getting long. Three weeks later the LinkedIn posts and the newsletter sound like two companies. We wrote about what that costs in the hidden cost of brand drift.

What to store, and what to leave in the prompt

Store what changes slowly.

  • Who you serve and the problem you solve
  • Voice, with two or three real examples of posts you were happy with
  • Words and claims you never use
  • Product names, prices, and facts that must be right
  • Formatting habits, such as emoji and hashtag use

Leave in the prompt what is new this time: the goal, the channel, the offer, and the one constraint that matters most.

Here is the same request both ways.

Before. "You are a copywriter for a neighborhood bakery. Our tone is warm and plain, never salesy. We serve families and office orders. Never say artisanal. Use at most one emoji. Now write a LinkedIn post about our new catering menu."

After. "Write a LinkedIn post about our new catering menu."

The first prompt is 40 words, and 30 of them are brand context that has not changed since last week.

Option 1: account-wide instructions

ChatGPT has custom instructions and Claude has a similar account-wide setting. Claude's help center describes it this way: "Any instructions you add here will be applied to all of your conversations with Claude" (Claude personalization).

This is the quickest fix for one person. The limits are that it holds text only, it applies to everything you ask, including things unrelated to the brand, and it belongs to your account, so a teammate has to copy it into theirs.

Option 2: a project per brand

A project keeps instructions and reference files together and applies them to every chat inside it. ChatGPT Projects work this way. So do Claude Projects, which Anthropic describes as "self-contained workspaces with their own chat histories and knowledge bases" (Claude Projects).

This is the best free or low-cost option for most small teams. Put the voice guide, the product sheet, and five approved posts in the project, and write short instructions that point to them.

Check the sharing rules before you rely on it. Claude's help center says project sharing is for Team and Enterprise plans. On a personal plan, each person keeps their own copy, and the copies drift apart.

Option 3: a shared assistant or instructions page

Gemini has Gems, which hold a name, instructions, and knowledge files, and can be shared with a person, a link, or an organization (Gemini Gems, sharing Gems).

Notion takes a different route. Its agent reads instructions from a Notion page and, in Notion's words, "will always reference your instructions in its interactions with you" (Notion Agent). If your guidelines already live in Notion, that page can be the source. Notion AI is on the Business and Enterprise plans (Notion AI FAQ).

Option 4: a marketing tool with brand memory

The first three options store your guidelines. None of them publishes a post or reads how it performed. You still copy each draft into a scheduler.

Marketing tools with a stored brand voice close that gap in different ways.

ToolWhat it storesWhich planPublishes the content too
JasperBrand voices, knowledge, audiences. Style guide and visual guidelines on Business2 brand voices on Pro at $59 a seat on yearly billingNo native social or email publishing listed
HubSpotBrand voice with tone, mission, terms to avoid, and writing samplesProfessional or EnterpriseYes, social, blog, and email
StoryChiefTone, values, audiences, content pillars, and preferred termsNamed on its AI agent pageYes, social and blog platforms
DaocoAn editable brand memory: voice, positioning, visual identity, audience, products, competitorsEvery plan, including the trialYes, 8 social networks and newsletter drafts

Sources: Jasper pricing, HubSpot brand voice, StoryChief AI marketing agent. Read on October 11, 2026.

Daoco is our product. It builds the brand memory from your website, social profiles, and files, with the source for each fact, and you can edit any of it. Once a brand has at least 10 imported posts, it also learns how you write from them, and fields you edit are kept when that profile refreshes (learned brand voice). The specialists that plan, write, and design all read the same memory. You can see what it finds on your own site with the free brand memory tool, which needs no account.

Jasper has the most governance controls of the four, and HubSpot is the natural choice if your contacts and email already live there.

Which option fits

Your situationStart with
One marketer, occasional contentAccount-wide instructions, then a project
Two to five people on one brandA shared project or Gem, on a plan that allows sharing
Guidelines already live in NotionA Notion instructions page
Several channels every week, with publishing and reportingA marketing tool with brand memory
Several brands or clientsA tool that keeps one memory per brand

Keep it current

Stored guidelines go stale the same way a brand book does. Three habits help.

  1. Name one owner. One person edits the stored version. Everyone else suggests changes.
  2. Save corrections, not only drafts. When a reviewer says "we never say this", add it to the stored guidelines that day.
  3. Reread it each quarter. Check prices, product names, and the examples.

Then keep prompts short. If you catch yourself pasting context into a request, that context belongs in the stored version. For how short a prompt can get once the context has moved out, see stop writing essays in your AI prompts.

Frequently asked questions

How can a small marketing team avoid repeating the same brand guidelines in every AI prompt?

Store the guidelines once where the AI reads them on every request. The options are account-wide instructions, a project with instructions and reference files, a shared assistant such as a Gem, or a marketing tool with its own brand memory. Then keep prompts to what is new for the task.

What brand information should I store for an AI tool?

Who you serve and the problem you solve, your voice with two or three real examples, words and claims you never use, product names and prices that must be right, and formatting habits such as emoji and hashtag use.

Can a ChatGPT or Claude project be shared with my team?

It depends on the plan. Claude's help center says project sharing is available on Team and Enterprise plans. Check your plan before relying on a shared project, because separate personal copies drift apart.

How can I stop my brand voice from changing when I adapt content for email, social media, and blog posts?

Write every channel version from the same stored brand context and the same approved source piece, then review the versions together. When a reviewer corrects the voice, add the correction to the stored guidelines so the next draft starts from it.