--- title: Load Your Brand Into a Claude or ChatGPT Project — Mustafa Alfredji description: The exact steps to turn brand documentation into a project your AI actually uses — Claude and ChatGPT, click by click, plus how to test that it worked. url: "https://mustafaalfredji.com/blog/load-your-brand-into-a-claude-or-chatgpt-project" type: static generatedAt: "2026-08-07T05:10:56.609Z" --- ![A stack of brand documents flowing into an AI chat panel](/_image?href=%2F_astro%2Fhero.DNqi1e3n.svg&w=1200&h=630&f=svg) Everyone has the same models now. Ask a generic chat for “10 hooks about discipline” and you get the same 10 hooks as everyone else who asked, because the model is starting from the same place: the average of the internet. The fix is not a better prompt. It is context — a project that holds your brand’s documentation, so every chat starts from *your* audience, *your* voice, *your* strategy. Both Claude and ChatGPT ship this as a first-class feature, both call it a Project, and loading one takes about ten minutes. This is the exact setup I use for my own brands and for client work. It works with any brand documentation; the better the documents, the better the output. ## What you need - **An AI account that supports Projects.** Both Claude and ChatGPT include Projects on free plans with limits; paid plans raise the file and usage caps. A full documentation stack fits comfortably on either paid tier. - **Brand documentation worth loading.** Files that record decisions: who your audience is and what they secretly want, what you stand against, how you talk, what your hooks do, what each content pillar is for. A logo kit and three tone words will not change the output — psychology and rules do. - **Ten minutes.** The setup is genuinely one-time. Updates are drag-and-drop. What makes documentation 'loadable' AI uses documents that make decisions for it. “We sound premium” decides nothing. “Short sentences. No exclamation marks. Never say ‘unlock’ or ‘game-changer’” decides every line it writes. If your documents read like a mood board, tighten them before uploading — or generate them properly with a system built for it. ## Setting up a Claude Project 1. **Create the Project.** In Claude’s sidebar, open **Projects** → **New project**. Name it after the brand — “Acme — Brand”, not “Test 3”. The name anchors every chat you start inside it. 1. **Upload your documents to project knowledge.** Open the project’s **knowledge** panel and add your brand files — markdown or plain text work best. Claude’s project knowledge is generous: a full 14-document stack fits with room to spare. 1. **Set the project instructions.** This is the step most people skip, and it is the one that makes the files matter. In **Instructions**, tell Claude what the files are and when to use them — for example: “You write content for this brand. Before writing anything, consult the voice document for style rules and the audience document for psychology. Never contradict them.” 1. **Start a chat inside the project and test it.** Ask for something small — one hook, one caption. Then ask: “Which documents did you use, and which rules did you apply?” If it can answer specifically, the project is live. ## Setting up a ChatGPT Project 1. **Create the Project.** In ChatGPT’s sidebar, click **New project**, name it after the brand, and open it. 1. **Add your files.** Use the project’s **Files** area. ChatGPT caps the number of files per project (the cap depends on your plan), so if your stack is large, upload bundled files — several documents merged into one — rather than dozens of small ones. 1. **Set the project instructions.** Same principle as Claude: tell it what the files govern. “Voice rules live in the voice bundle. Audience psychology lives in the audience bundle. Consult them before writing; when in doubt, quote the rule you are following.” 1. **Know how retrieval works.** ChatGPT reads your files through retrieval — it pulls relevant chunks rather than holding every file in mind. This is why one compiled master context file matters more here: a single document that summarises the whole stack gets retrieved intact and keeps answers coherent. 1. **Test it the same way.** One small ask, then “which rules did you just apply?” Vague answer, vague setup. ## How to know it worked Run the same request twice — once in a fresh generic chat, once inside your project: > Write me three hooks about [your core topic]. The generic chat gives you competent, interchangeable lines. The project should give you hooks that use your vocabulary, aim at your audience’s actual tension, and sound like something you would post. If you cannot tell the two apart, the problem is almost never the AI — it is that the documents do not contain decisions. ## Troubleshooting - **It ignores the files.** Instructions, not uploads, are what bind the AI to the documents. Restate them: name each file and the decision it owns. - **It sounded right last week and drifted.** Long chats dilute context. Start a new chat in the project — projects exist so that starting fresh costs nothing. - **You updated a document and nothing changed.** Re-upload the file. Both platforms snapshot files at upload; editing your local copy does not reach the project. - **ChatGPT keeps missing details.** That is retrieval doing its thing. Merge your documents into fewer, larger bundles, and lead each one with the rules that matter most. ## The honest prerequisite None of this works without documentation worth loading. The project is a delivery mechanism; the advantage is what is inside the files — the audience psychology, the voice rules, the hook strategy that only you could have written down. If your brand has never been documented at that depth, that is the actual work — and it is exactly what [The Brand Doctrine System](/products/brand-doctrine) builds: a guided interview that runs inside the project you just learned to set up, generates the full 14-document stack for your niche, and compiles the master context file that makes any fresh chat instantly yours. ## Quick answers What is a Claude or ChatGPT Project?+ A workspace inside Claude or ChatGPT that holds files and instructions every chat in it can use. Upload your brand documents once, and every new conversation starts already knowing your audience, voice, and strategy — no re-pasting, no re-explaining. Do I need a paid plan to use Projects?+ Both Claude and ChatGPT include Projects on free plans with limits on file count and usage; paid plans raise those limits. For a full brand documentation stack you will usually want a paid tier of either — check your plan before uploading. What files should I upload to make AI write in my brand voice?+ Documents that describe decisions, not vibes: who your audience is and what they secretly want, what you stand against, how you talk (sentence length, banned words, signature phrases), your hook patterns, and your content pillars. One compiled master context file plus the source documents is the strongest setup. Why does my AI still sound generic after uploading files?+ Usually one of three things: the files describe aesthetics instead of decisions, the project instructions never tell the AI to consult them, or the request itself overrides them ("write a viral hook" invites the average of the internet). Fix the instructions first: tell the AI which file governs which decision. Should I use one big file or many small documents?+ Both. Keep the full documents in the project for depth, and add one compiled master context file that summarises the stack. Claude can hold the whole stack in context; ChatGPT retrieves chunks, so the single compiled file is what keeps answers coherent there. --- ## Keep reading ![A document with highlighted lines feeding into a node graph](/_image?href=%2F_astro%2Fhero.B08py7GJ.svg&w=1200&h=630&f=svg) Build the System · Build the Brand ### [How This Site Talks to LLMs](/blog/how-this-site-talks-to-llms/) Every page here ships a markdown mirror, an llms.txt index, a served AGENTS.md catalog, and answer-ready JSON-LD. Here is the whole setup, end to end. July 26, 2026 · 2 min read ![Concentric rings on a warm paper grid, echoing the site mark](/_image?href=%2F_astro%2Fhero.B3wvGiZg.svg&w=1200&h=630&f=svg) Build the System ### [Building This Site on a Production-Grade Foundation](/blog/building-this-site/) Why a two-page personal site runs a tokenized design system, feature folders, layout-owned SEO, and a typed analytics catalog — and what that buys later. July 26, 2026 · 2 min read [← All posts](/blog/)