The Only AI Writing Guide You Need in 2026 (Claude + ChatGPT)
A folder system, not another prompt list — how to make Claude or ChatGPT write in your voice by giving it a home base of context instead of a one-off prompt.
Posted
3 weeks ago
Duration
Format
Tutorial
educational
Views
614
42 likes
57 · 43
Big Idea
The argument in one line.
Generic AI writing is a context problem, not a prompting problem, and the fix is a persistent folder system — an About Me file plus brand voice, reference, and editing-rule subfolders — that Claude or ChatGPT reads before every draft instead of starting cold each time.
Who This Is For
Read if. Skip if.
READ IF YOU ARE…
You use ChatGPT or Claude to draft LinkedIn posts, emails, articles, or scripts and the output reads generic or off-brand.
You already have past writing, templates, or a rough style guide sitting unused in scattered docs and want AI to actually use it.
You want a repeatable setup you build once instead of re-explaining your voice in every new chat.
SKIP IF…
You're looking for prompt phrasing tricks — this video is about a folder/context system, not a prompt library.
You don't use a tool with local file or project-folder access (Claude Cowork, ChatGPT Projects/Work) — the whole method depends on the AI reading a folder.
TL;DR
The full version, fast.
Most AI writing sounds generic because people type a vague prompt into a fresh chat and get a vague answer back — the model did exactly what it was asked, with zero context to work from. The fix is a small folder system: one home-base folder with an About Me.md briefing document, plus Brand Voice, References & Templates, and Copy Editing Rules subfolders, optionally connected to live-data tools like Exa via MCP. Point Claude Cowork or ChatGPT Work at that folder and it drafts from your actual voice, past examples, and rules instead of guessing. Build the folder once, in under an hour, then save every output that works back into it so the system compounds — each session gets sharper than the last, but the writer still owns the angle, the story, and the judgment calls.
Free for members
Chat with this breakdown — free.
Sign in and you get 23 free chat messages on us — ask for the hook, quote a framework, find the exact transcript moment, generate a markdown action plan. Bring your own key when you want unlimited.
Cold open reframes generic AI output as a systems problem, not a model problem, and Eddy establishes credibility (4+ years, Jasper, 4,000+ students coached).
01:25 – 03:51
02 · Improve AI Output with Context
Side-by-side comparison: a bare prompt ('write me a LinkedIn post about AI writing') produces generic corporate output, while the same request loaded with context, a hook structure, and a brand-voice reference produces a specific, personal post.
03:51 – 06:51
03 · Set up AI Folder System
Introduces choosing one tool (Claude Cowork or ChatGPT Work) and creating a home-base folder with an About Me.md file as the foundation everything else is built on.
06:51 – 08:25
04 · Organize your AI writing folder
Builds the three subfolders — Brand Voice, References & Templates, Copy Editing Rules — then shows connecting outside tools via MCP, using Exa as the example.
08:25 – 09:59
05 · Use AI to Draft Content
Live demo with a fictional HVAC business owner ('Mike'): Claude Cowork reads his folder, interviews him about a work accident, then drafts a LinkedIn post pulling from his About Me, brand voice, and post template.
09:59 – 13:58
06 · Always improve AI output
Explains the compounding loop (save outputs back into the folder), warns that AI doesn't replace the writer's judgment, then gives a same-day checklist and points to a free folder-system template.
Atomic Insights
Lines worth screenshotting.
A generic AI writing prompt isn't a failed prompt — the model did exactly what was asked; the input just had no context in it.
The fix for bad AI writing output is a folder system, not a better one-off prompt.
An About Me.md file works like a briefing document handed to a new employee on day one — skip it and every draft defaults to generic.
The full folder architecture is only three subfolders: Brand Voice, References & Templates, and Copy Editing Rules.
Claude Cowork and ChatGPT Work both read and write inside your actual local folders, so they function as a writing partner rather than a stateless chatbot.
MCP (Model Context Protocol) connections let the AI pull live web data for research and fact-checking instead of relying on stale training knowledge.
Asking the AI itself how to connect a new MCP tool works better than searching for setup instructions — it can walk you through the connection directly.
The system compounds: every output you save back into the folder becomes future context, so voice and rules get sharper with each session.
AI does not make someone a better writer automatically — used carelessly, it lets instinct for hooks and structure atrophy.
The writer is still driving: the folder system controls tone and rules, but the topic, angle, and judgment calls remain the human's job.
The full folder system (About Me file plus three subfolders) takes under an hour to build the first time.
Takeaway
A folder system beats a better prompt every time.
WHAT TO LEARN
Generic AI writing is a context problem: build one home-base folder with an About Me file, brand voice, reference templates, and editing rules, and every future draft gets sharper because the AI reads it before writing.
01AI Writing Intro
A generic prompt produces generic output because the model did exactly what it was told with no other information to draw from — that's not a model failure.
Credibility built over years (4+ years of AI writing, early work with a GPT company, coaching thousands of students) is worth stating up front so the method reads as tested, not theoretical.
02Improve AI Output with Context
Comparing a bare prompt against the same prompt loaded with context, structure, and a voice reference makes the value of context immediately visible rather than abstract.
A single reference example — one converting post, one working template — teaches an AI system more than a paragraph of instructions describing your style.
03Set up AI Folder System
Before building any AI content system, write the About Me file first: it is the foundation every other file depends on.
Pick one tool with local folder access (Claude Cowork or ChatGPT Work) and commit — both work the same way against the same folder structure.
04Organize your AI writing folder
Keep the folder structure to three purpose-built subfolders (voice, references, editing rules) — resist the urge to over-engineer it into more categories.
Connecting outside data tools (MCP-based research tools like Exa) closes the gap between AI drafting and AI that can fact-check itself against current information.
Ask the AI directly how to connect a new tool instead of searching documentation — a modern assistant with tool access can usually walk you through its own setup.
05Use AI to Draft Content
Letting an AI interview you with specific questions (who you serve, your credibility, what to never associate with your brand) produces sharper source material than trying to write a bio cold.
A real example grounded in specific numbers and details reads as more credible and personal than a polished, vague version of the same story.
06Always improve AI output
Every output that performs well should be saved back into the folder — the system's quality compounds only if wins get archived, not just produced.
Treat AI-assisted writing as a risk to instinct if used passively: relying on it without thinking about hooks and structure lets those skills atrophy over time.
The entire buildout (About Me file, three subfolders, one connected tool) is scoped to take under an hour, which removes the excuse to keep procrastinating on it.
Glossary
Terms worth knowing.
Claude Cowork
A mode of Claude that works directly inside a user's local file system or connected folders, reading and writing files rather than only responding in a chat window.
ChatGPT Work
OpenAI's equivalent to Claude Cowork — a project mode that connects to local files or services like Google Drive and works from those documents instead of a blank chat.
MCP (Model Context Protocol)
A standard that lets an AI tool connect to outside data sources or services — such as live web search — so it can pull current information instead of relying only on what it was trained on.
Exa
A web search tool built for AI agents, used here as an MCP connection so Claude or ChatGPT can research and fact-check with live web results.
About Me.md file
A markdown document describing who you are, what you do, who you serve, and your business context — meant to be the first file an AI reads before drafting anything.
tight, quotable rule that survives outside the video↗ Tweet quote
11:40
“You're still driving. The AI is just the engine.”
clear metaphor, good pull-quote for the responsibility caveat↗ Tweet quote
12:10
“AI doesn't make you a better writer automatically. Use it wrong and your instincts atrophy.”
the video's explicit warning, useful as a counterweight clip↗ Tweet quote
The Script
Word for word.
Read-along
Don't just watch it. Burn it in.
See every word as it's spoken — crank it to 2× and still catch all of it. The same dual-channel trick behind Amazon's Kindle + Audible.
17px
metaphor
If you're using AI for writing and it sounds like everybody else, it's not an AI problem. It is a systems problem. And in this video, I'm gonna show you the ultimate AI writing guide, and it's not gonna be another prompt video.
I'm gonna show you how to create a folder system that makes AI actually write like you. And the reality is that most people are using AI wrong when it comes to writing. They open up ChatGPT or Claude, they type in a very generic prompt, and they're gonna get a generic output.
And then they're gonna question themselves of why does it not sound like me? AI sucks. This doesn't work.
Why does it work for others? But, again, in this video, I'm gonna show you the exact system that completely fixes it. So why listen to me?
Well, my name is Eddie. I've been using AI for over four years now. I had the privilege of working on the Jasper Mastermind, which is one of the first AI companies to use ChatGPT as a rapper.
And I got to work for a community called Blog Earth Engine that has over 4,000 students, and the primary goal of that community was for me to help them with content writing, creating systems, and showing up on Google results using AI.
And funny enough, I'm still in the trenches daily. I do have a website that ranks on the first page of Google for specific keywords, And I use AI to help me with my community posts, my emails, and my own work.
So I know this still works, and by the end of the video, you should have a workable system or something at least to help you get along with AI. So when it comes to AI prompting, it's important to understand what is actually bad and what is actually good.
Most people will type in a very generic prompt, and they can say they're just starting, and they're typing in, write me a LinkedIn post about AI writing. That is, like, terrible because it's gonna give you very generic fluff information, doesn't have any other context of about what you actually wanna write about, and you didn't give it, like, a hook structure.
You didn't give it any other frameworks to follow. So you're gonna get a very corporate SIFF and generic output.
Nobody's gonna read it and nobody's gonna care. And from this example alone, you can see that while it does provide some text, it doesn't have any context behind this.
You can see that there's, like, clearly AI written verbiage. There's em dashes, and this is not gonna do so well.
However, now we add more context and I write something like, I haven't posted on LinkedIn in eight months. I just got back. Write me a LinkedIn post in my voice using no fluff that I've been heads down building automations in Claude, and then provide one valuable AI prompt based on my topic for my audience.
And then make sure that you use my brand voice in any memory of you that you have of my prior writing. And at the end of the generation, create a sketch image that would pair well with my return. And this is so much better.
You can see that it's actually using my style for LinkedIn posts. If you ever seen any of my LinkedIn posts before, it's saying it's been eight months. No.
I didn't quit LinkedIn. I just stopped posting. Instead, I went heads down building hundreds of hours testing clon, breaking prompts, rebooting workflows.
There's a lot more here. And then at the end, over the next weeks, I'm sharing some practical things, like no fluff, no 10 AI tools that you need, just practical automations, cloud workflows, prompt engineering, the whole shebang here.
And then it even gives that valuable AI prompt that anyone can just literally steal and use for themselves. And I even had create, like, a image of what that would look like if we put it all together, and I think this is so, so cool.
Of course, you know, that isn't me right there. But as a little kind of overview what that would look like, you can have Chattypiti recreate this for you.
So, I mean, this is just insane when it comes to putting context into your AI writing. If you have references, if you have other documents, or if you have templates that you use, or you just speak to Claude about what you want, whether it's ChatGPT or Claude, you'll get way better outputs than just saying, write me a LinkedIn post.
So like I said, you can copy this prompt right here on the screen, but it's really not about the tool. It's just having context. And I think this is really important to understand that you just should focus on context prompting, which is in prompting and then adding more context to that.
Now I'm gonna dive a little bit deeper here, and I want you to choose one tool. It doesn't matter what you choose at the end of the day because both of them now do the exact same thing. So you can use Claude Cowork and create a very practical folder system that has your brand voice, everything that you need because it lives in your local files.
And now there's ChatGPT work, is a direct competitor of Cloud Cowork, and it does the exact same thing. You can work in your local file system. It also connects to a bunch of different apps like your Google Drive if you have already templates and articles or YouTube scripts that you've ever read in the past.
Cloud Cowork or ChatGPT work can grab all of those files directly with connections. Or if you already have it in a folder system, then you can use that as well. So both work inside of your actual folders.
They're not just a regular chatbot like we've been used to, and it becomes a writing partner for your business. I literally use it for almost everything now. So use what you already use.
If you are a huge ChatGPT person, stick with that. I'm not gonna tell you to move to Claude. And if you're with Claude, then you can stick with that as well.
The great thing about this is that you can use any folder system and it works the exact same for any of these tools. So before I dive into actually showing you, I need to make sure that you actually understand the steps. So we need to create a home base.
So we're gonna have we're gonna create one folder where everything lives. And the cool thing is that you can actually create this directly into Cloud or ChatGPT work, and you're gonna have it create an about me markdown file. This is gonna talk about who you are, what you do, who you serve, what your business is about, and it's essentially a briefing doc that you hand over to Cloud as it feels like a new employee.
If you don't have an about me file, you're most likely gonna get generic output every time, or you're gonna have to provide more context. Right? So once we created that about me file, it's gonna live in these subfolders here.
This is why it's gonna start to compound and become better every time. We're gonna be able to create our brand voice, which is essentially our tone, our vocabulary, our dos and don'ts.
And if you have any references and templates, so let's say you're writing, you know, articles on Google and you have specific blog post templates that you follow that have a specific heading hierarchy or you like how you have your frequently asked questions or your hook in the first paragraph, things like that, you can include your references and templates.
And if you have any copy editing rules such as, like, fixing grammar issues, no em dashes, what your style guide is, you can have all those things and create a entire folder system of how you write. And then once we created the folder system, you wanna connect the tools that you already use.
So this could be Google Drive, this could be Notion, your, like, your content planner. This could be other MCP tools like Exa, which is the one I use for all of my web search for better fact checking, for better research, and just for honestly understanding information more easier.
I promise you, becomes easy. You just ask the AI, how do I connect this MCP to you? And it literally will walk you through it or sometimes even just give you the connection to connect.
So this is exactly how it's gonna look like once we build it. You're gonna have your home base of all your folders. You're gonna have your about me folder, your brand voice, your templates and references, any editing rules, and an archive as well.
I'm gonna give you a sample that you can see as well with a truck driver named Mike. He's not a rider, but that doesn't matter. It's gonna contain a swipe file for him, and it's also gonna contain an archive.
So for our demo here with Mike, you can see that this is what's gonna look like. It's gonna have a folder structure, and this README template gets built once you've created your about me markdown file. Whenever you use ChatTubt or Clon, it's going to read this entire README file so it understands the context so it never has to ask you questions.
Right? So it's gonna say, here's my AI writing folder. Here's what I built.
I have my, um, all my, you know, brand templates, references, my copy editing rules, and archives. I wanted to always review this about me MD file and add anything I miss in the future, read my brand voice, save any content I perform as well, and then update to improve any file. And then here, it's gonna contain all that information about Mike.
So it's gonna have the short bio here, who he serves, which is home service owners, credibility, what he's building, so a newsletter, YouTube channel, and a paid community.
Never associate my brand with debt rich quick private equity, so that's pretty cool. And then we have the brand voice template. We have the copy editing rules to follow, and everything lives in here.
So you can continue adding and improving it over time. So we're gonna open up Cloud CoWork. You can do the same thing in Chateactbt work.
And I created that folder, right, for demo using Mike. And you wanna make sure that it's pointing to the specific folder when you're actually creating the content so it can continue refining, improving it, and then saving everything in there.
And all I'm gonna say is let's begin. Cool. So the first thing it's gonna say, hi, Eddie.
Ready when you are. What are we working on? An email draft for YouTube titles, a research brief.
So it already knows. I'm gonna say here, help me write a LinkedIn post for a recent accident that I encountered for when I was on the job. So I'm gonna have a say, help me write a LinkedIn post for a recent accident that I encountered when I was the job.
Cool. So there's a thought process behind this, which is awesome. It's basically asking us before we draft anything, what happened?
What was the accident? What's the takeaway? And any details you'd rather leave out?
You can see here once you start to add information, it's going to include all of Mike's AI writing folders, which is his about me markdown, brand voice, LinkedIn post template, and copy editing rules.
So this is pretty cool. Uh, it came out with that text. One of my texts fell through an attic ceiling in June.
He's fine. The customer's dining room was not. Second week of Phoenix heat wave, board was slammed, every truck out, and we were running calls forty five minutes long that than it should have taken.
He was rusty attic. Here's the part that was actually my fault. I built a schedule that only worked if nobody slowed down.
So, yeah, you can see that this is pretty good based off of what Mike's business is and his brand voice. You can see that it includes all that information here, and it also includes the LinkedIn post template that's within that folder. So as you can see, can start to build more references, more templates, and it just creates something amazing with your writing style.
So now just imagine that you write with AI like that where it's saving your folders, it's saving your prior outputs. It's going to continue getting better every time because I prompted it to make sure that it follows a system where it continues to update things that went well. So you write something, you save it back into the folder, you provide more context, a sharper voice, tighter rules, and it continues to compound after that.
So before every session that you write with AI, you have your unique angle, your key points, your opinion, and then you just hand it off to ChatGPT work or Cloud Cohort. So I'm gonna go ahead and leave that prompt and folder system for you in the bottom of this video in the description. So now you can make your own BrandVoice Bounty templates and everything and have that same system that I just showed you and the one I use for my own business.
So to make this work, what you're gonna do is you're gonna create a new folder, and you can name it whatever you want, but we're just gonna say AI writing system. And you're gonna drag that start here m d file in here. Now, what we're gonna do is we're gonna point to that new folder and then click open.
And then once you've done that, you can go ahead and start creating your folder system. So keep it really simple. You're gonna wanna say, using that markdown file, help me create my folder system inside of the AI writing system folder, and then now it's gonna interview me.
Now this takes about twenty to thirty minutes, something longer, but what you're gonna wanna do is you're gonna wanna answer all of those questions. And try to write as much as possible. Or if you're using, like, a voice command feature like Whisper Flow or Super Flow or something like that, or Super Whisper, you can use that as well.
But you can see it's gonna ask us what our name is, what we'd actually do, not the LinkedIn version. Who do we help? What results do we get people?
What's the credibility? How many years have we been doing it? Do we have any companies, results?
What are we currently building? What do you wanna be known for in two years? What should we never associate your brand with?
And then once you've created that, that's just section one out of five. It's gonna ask you about including your brand voice templates, if you'd have any prior documents or frameworks that you use. You're gonna have to include all of those things in here, and then you'll come up with another similar folder system like the one I showed you for Mike.
So go ahead and build that out for yourself. Spend the next thirty minutes or so doing that, and then you'll have a perfect structure for your AI writing. And I do need to put this out as a warning.
AI doesn't make you a better writer automatically. If you use it wrong, it's gonna make your results even worse. So what this does, it just helps you stop thinking about hooks, the structure, and it's gonna teach you what actually lands.
You're still basically the person driving because it's gonna just have your structure and how you write, but you're the one still thinking about the topic, adding the references, adding the information. It's still your voice and then your judgment call.
Once you've made some edits to the output from AI, it's gonna take way less effort to do that. So go ahead and build this out for yourself. Create the folder, write the about in the MD file, build your subfolders, and then connect to, like, an AI tool like Exa or another MCP that makes sense for your business.
And it's just as easy as asking it to connect to those tools and say, what are current MCP connections that you have right now that I can leverage for my AI writing? And then help me connect to those tools. So you can see here that it it's kind of doing some research.
It knows what MCP connections I'm already using, like Kit, Notion, Canva, and Google Drive.
And then it's asking me it's gonna check the MCP registry from Claude and if there's anything else that we should could include. So, I mean, I already have all of my connections that connected already. So you would just ask Claude to add a connection that you want, and then it'll help you do that.
So there you go. You can go ahead and grab that prompt and markdown file in the bottom of this video so you can create this yourself like I stated. If you have any questions, feel free to comment in the bottom of this video.
I'm also launching my community soon where it's gonna give you prompts, templates, frameworks, everything that I use for AI marketing, such as Claude Cowork and Tat GPT work, how I create folder systems, how I market my business, how I do it for others. So I'll see you guys in there.
I will let you know more about it. You can watch this video next that YouTube is gonna recommend you. I'll see you guys in that video.
The Hook
The bait, then the rug-pull.
Eddy Ballesteros opens with a flat rejection of the usual complaint — AI writing doesn't sound generic because the model is bad at writing, it sounds generic because it was handed nothing to work with. What follows isn't another prompt list; it's a folder system meant to live behind every future draft.
Frameworks
Named ideas worth stealing.
04:51model
The AI Writing Folder System
About Me.md (home base)
Brand Voice
References & Templates
Copy Editing Rules
Archive (saved wins)
One home-base folder containing a briefing-doc-style About Me file, plus three purpose-built subfolders that hold tone/vocabulary rules, past converting examples and templates, and grammar/style constraints — connected to a project-aware AI tool so every draft pulls from real context instead of a blank prompt.
Steal forany recurring AI writing task — LinkedIn posts, emails, articles, YouTube scripts — not just this creator's workflow
CTA Breakdown
How they asked for the click.
VERBAL ASK
12:30newsletter
“I'm gonna go ahead and leave that prompt and folder system for you in the bottom of this video in the description.”
Soft CTA delivered as a practical resource drop (free folder-system template + prompt via a Kit signup link) rather than a hard sell, reinforced on-screen by a 'community launching soon' note.
A creator connects Higgsfield to Claude Code's Fable 5 model over MCP, turns one detailed prompt into a full portfolio site, and publishes it live on a custom domain before the twenty minutes are up.
A two-tool loop — ChatGPT Sol for the mockup and real images, Claude Design for the working code — that turns one prompt into an animated, on-brand website in under fifteen minutes.
Brandon Doyle gave Claude, ChatGPT, Gemini, Grok, and Perplexity $1,000 each of real money and told every one of them to win or he'd cancel the subscription — six months later Claude has more than doubled its stake.
Mobbin's new MCP connector lets Claude search 600,000+ real, shipped app screens as design references — trading the generic serif-and-purple-gradient AI look for interfaces grounded in production apps.
00:00
00:26
00:37
00:47
00:58
01:09
01:19
01:40
01:51
02:01
02:11
02:22
02:32
02:43
02:53
03:04
03:14
03:24
03:45
03:56
04:06
04:17
04:28
04:38
04:59
05:10
05:21
05:31
05:52
06:03
06:24
06:35
06:45
06:56
07:06
07:17
07:27
07:38
07:48
07:58
08:09
08:15
08:30
08:40
08:51
09:12
09:22
09:32
09:43
09:50
10:04
10:14
10:25
10:35
10:43
10:56
11:06
11:16
11:27
11:48
11:57
12:08
12:19
12:40
12:49
13:00
13:11
13:25
13:32
13:42
13:52
20:00
13:36
36:26
09:00
09:59
11:14