One of your clients emails on a Tuesday morning.

Subject line: Quick Question…

Two words doing a lot of work to avoid saying what this actually is about.

I heard from a friend at the gym that agencies like yours use Claude to write blog content now.

Before he typed a word to you, he searched it himself. Was that true of agencies in general? Then he asked about his account.

You could explain the strategy behind it… The research that goes into it, the hours it takes editing, let’s not mention the four rounds it took to get the angle right before that article got approved. But he’s already done the math on his own. Claude Pro costs $20 dollars a month. His retainer costs somewhere between $2,000 and $15,000.

That difference is the whole conversation now. Whether you get to say it out loud or not.

5-6 years ago, checking used to take real digging. Now it takes thirty seconds, a search bar, and nowhere left to hide.

This isn’t one client being difficult. It’s showing up in inboxes across the industry, for two different reasons. Both land on your desk looking like the same problem.

The Client Already Paid for Work That Looked Finished and Solved Nothing

Some version of this has already happened to him before you.

The last agency he hired delivered content that looked professional. At a glance, the format looked nice and the keywords were right. He even ran it through Surfer SEO, and it passed every check. A couple months after publishing, no calls. No new customers found him because of it. He paid on time, every month, for work that solved nothing. Eventually he stopped renewing. He never fully explained why.

He came to you already burned once. Burned isn’t a feeling that fades quietly. It changes how a person reads every invoice that comes after it. Slow weeks becoming the new normal again, that’s the real fear underneath this. It’s the same fear that was there right before he fired the last agency.

Now there’s a twenty-dollar line item sitting right next to that memory, wearing a smaller price tag this time.

He likes the work you send him, even now. He’s embarrassed that he can’t always tell which draft is the good one. Part of him is hoping you give him a reason to stay instead of a reason to leave.

On top of all that, a competitor’s pulling ahead. This one’s costing him customers.

The Competitor Who Never Ranked Now Shows Up First and Takes the Call

He’s got John two blocks away. A competitor who never posted anything online.

Not once!

Now John’s articles show up in press releases. People share them on their own feeds.

Everywhere he looks, John’s there, first page. Sometimes first result. Articles that read like John is speaking straight to the customers he wants in his own store. How did John go from invisible online to this?

The customers who used to call his business are calling John’s instead.

Now he’s opening Claude himself. At this point, he’s not emailing you about it at all.

He’s already told you, more than once, that he doesn’t have the time or the interest to become a content strategist. But twenty dollars and an evening is still a cheap bet. It’s cheaper than vetting another agency so soon after the last one burned him.

So what do you do? How can Claude help you save the relationship?

Build the System That Makes the Copy Compound

You do not save the relationship by explaining that the brief used to take a junior writer four hours.

The client is not paying for four hours. He is paying for an article that understands his customers, sounds like his business, and gives someone a reason to call him instead of John.

So show him where that difference comes from.

Bring the reviews into the process. Bring in the sales calls, the competitor research, the objections his team hears every week, and the things the business knows that no generic prompt could uncover. Then let Claude help your team turn that material into the brief faster.

Do not hide the tool, and do not sell the time it replaced. Make the strategy around it visible.

A client can open Claude in minutes. What he cannot reproduce in minutes is the system that knows what to feed it, what to reject, and what the finished work needs to accomplish.

I’ve done this with real client work.

Reviews, press coverage, customer language, and competitor research all went into one process. Claude helped turn that material into real articles faster than any single writer could have managed alone.

One project built that way supported a celebrity product launch in one of the most crowded and regulated markets online. The article reached page one for several competitive launch searches, drove 1,473 store visits, and was later cited on the celebrity’s Wikipedia page.

The result did not come from producing words faster. It came from researching what the audience was already looking for, finding an angle beyond the recycled press coverage, and building the article to keep working after the launch was over.

Adoption is not a niche story anymore, either.

Almost nine out of ten organizations worldwide now use AI in at least one part of the business, as McKinsey’s research on the state of AI shows. At the same time, Salesforce’s guide to AI tools for small businesses points to how far access has spread. A three-person shop can now tap into the same kind of computing power as a Fortune 500 company.

Access stopped being the advantage 3 years ago.

The tool is available to everyone. The reviews, customer insight, strategic judgment, and working process are not.

That is what the client is actually paying you to bring together.

A small team can now pull data, culture, and a real voice together at a level this work has not seen before, and at a speed that changes what is possible.

A Single Claude Workflow Protects the Margin on Every Account That Follows

I already built that system.

Mostly because I got tired of watching agencies treat Claude like a faster typist instead of an actual system.

It runs three stages.

The first sets up the account itself. It feeds Claude the business context, the voice, the strategy. Every draft after that starts from real material instead of a blank prompt.

The second stage builds the content engine itself. Research, drafting, and review work together. Nobody’s just handling it whenever they have time that week.

The third stage moves past chat entirely. It uses Claude Code for full integration. The workflow runs inside the tools your team already touches every day.

Build it once for one account, and something changes. The next account doesn’t start from a blank prompt either. Neither does the one after that.

That’s what actually protects the margin here. A system that keeps producing the same result, account after account, without needing to be rebuilt from scratch each time.

I put the full breakdown, three weeks, the exact plays, a way to measure the shift before and after, into the Claude Content Playbook → Get access here.

What Agency Owners Still Ask Before They Trust a Claude Workflow

A few things agency owners ask once they’ve read this far.

How much of this should I actually show a client?

Enough for him to see the difference. Show him the research that shaped the piece, the reasons behind the angle, and how it connects to what his customers actually search for.

What’s stopping a client from doing this with Claude on their own?

Time, mostly. He can open Claude and get a draft today. What he can’t do alone is keep feeding it the research, the customer language, and the angle work every single week while he’s also running his business. That’s the part that actually takes an agency.

How long does it actually take to build a workflow like this?

Three weeks, done in real stages instead of all at once. You keep working client accounts the whole time, nothing pauses for it. It’s built to fit inside a normal month, not replace one.

Do I need to change what I charge once clients know Claude is involved?

No. If anything, this could help grow your margins. Once a client can see the research and the angle behind the work, a discount stops being the easiest way to keep the account. The system becomes the reason instead.

PS... If you want to talk through what this looks like for your specific accounts, before the next invoice question lands in your inbox → let’s chat.

I didn’t like Claude at first.

New tool, moving too fast, changing every few weeks. I’m old school. Give me a process I trust and I’ll run it for years without asking too many questions.

For a long time, that process was Textbroker. Put two or three writers on an article, then go back to running the business while they worked. Check back in two to three weeks and a draft would be sitting there. A round of revisions if the brief wasn’t tight enough going in. But if I’d done my part and written a real brief, I’d get something publishable back most of the time.

A good brief back then had the basics. Who the article was for. What I actually wanted said. How I wanted it structured. Hand a writer that much, and even an average one usually came back with something usable.

That system worked. It still works, for anyone who wants to run it that way.

That is not why so many businesses publishing content right now are watching their traffic drop.

Here’s what’s actually happening.

Everyone found a faster version of the same shortcut.

Hey Chat, can you write me a 1,500-word article about this _______ topic? Use these keywords, make it sound professional, and don’t use em dashes.

Then they publish dozens, sometimes hundreds, of articles built from prompts just like that. Traffic climbs for a few months because the page is new and the keywords are close enough. Then, around month four or five, the traffic falls off a cliff. And most businesses never figure out why.

I know why, because I did the exact same thing the first few times I used Claude for real content work. What came back sounded exactly like everyone else’s content. That’s because I was pulling the same prompts everyone else was, off X threads and from YouTube marketers claiming to give away their secret sauce.

So I stopped asking Claude to simply write the article and started giving it what I used to give a writer at Textbroker, a real brief built from my sources, the structure I wanted, and my own take on the topic.

Here’s what changed once I did, and what I’ve built since.

Most People Never Get Past the First Prompt With Claude

A topic and a word count is not a brief. It’s a placeholder for one.

That’s the entire method most people use with Claude for content, and it’s the entire reason so much of what gets published now reads like it came from the same account. Claude didn’t get lazy or generic on its own. It answered exactly the question it was asked, and the question was thin.

Ask a stranger to write about your business in fifteen hundred words with no context, no proof, and no point of view, and you’ll get something that could belong to any business in your niche.

There is nothing technically wrong with it. The writing is clean, the services are clear, and it sounds professional.

It could also belong to almost any cosmetic dentist in Brooklyn.

Now give Claude the dentist’s actual story, why she chose this work, how she wants patients to feel, and what makes her approach personal. Now the copy changes.

The second version gives the reader something to remember. It sounds like it came from a real person because it started with information only that person could provide.

That’s not a Claude problem. That’s what happens whenever you hand someone a topic instead of a brief, whether that someone is an AI model or a freelancer you found last week on a job board.

Most people run that one flat prompt, get back something generic, and decide the tool itself is the ceiling. Fewer people stop to ask why the result was flat in the first place. That’s the mistake, and it has nothing to do with which model you’re using.

I have seen this play out while fixing content for businesses. The traffic might have looked fine at first, but the articles were not creating anything beyond the click. People were not sharing them, linking to them, or reaching out, because the content never gave them anything of value or a good enough reason to trust the business.

Once I looked closer, the pattern was usually the same. The team had followed the most common process they had seen online, starting with a thin prompt and publishing whatever came back. They were not choosing a weak method on purpose. They were repeating the only version of AI content production anyone had shown them.

Claude Followed My Structure and Still Found Stronger Points

That pattern made me rethink the way I was using Claude in my own work. Instead of asking it for a finished article from a thin prompt, I wanted to see what would happen if I gave it the same level of direction I used to give a writer I trusted.

I began by prepping for a new article, doing the kind of work I always do before writing anything that matters. I pulled the sources and research I knew I’d need, forming an actual point of view instead of just reporting on the topic, and sketching the structure in the order I wanted each point to land.

It was normal prep work, the kind I used to put into a brief and hand off to a writer before waiting two or three weeks to see whether they got it right.

This time, I handed all of it to Claude instead. My sources, my take, and the shape I wanted the article to follow from one section to the next.

The draft came back in 3 to 6 minutes.

Not a rough pass I’d need to gut and rebuild. The research went deeper than what I’d pulled myself, filling in context I hadn’t gotten to yet. The points were sharper than the ones I’d sketched out.

And it followed the structure I’d given it exactly, section for section, in the order I’d laid it out, instead of wandering off into whatever shape felt natural to write in.

I read it twice before I fully believed it, the way you check a number twice when it’s better than you expected.

I sat with that for a minute, because it wasn’t what I expected. I’d been treating Claude the same way I’d treated every writer who came back with something generic, handing it a topic and hoping for the best. This time I’d handed it what I actually give people I trust to do real work.

The Project Is Doing the Work Before You Ever Type a Prompt

None of that came from a cleverer prompt. It came from a Claude Project I’d already built before I ever opened that conversation.

A Project is a workspace that holds everything Claude would otherwise need explained fresh, every single time. Mine holds four things specifically.

Context for what I actually do and who I’m writing for. My brand voice, so it doesn’t default to generic business copy. The writing style I want followed, down to how I like paragraphs broken. And instructions on exactly how Claude should handle whatever I feed it, whether that’s a pile of research or a half-formed idea typed on my phone.

All four of those were already sitting there when I dropped in the sources and the structure. I wasn’t teaching Claude who I was in that conversation. I’d already taught it, once, and it remembered every time since.

This matters even more when several people are working on the same content. Everyone using the same Project follows the same brand voice, instead of five people quietly creating slightly different versions of the business.

The draft itself came back as something I could actually work inside of too, not a slab of text buried three replies deep in a chat thread. I could open it, edit it directly, move sections around, and hand it off looking like a real document instead of a transcript of a conversation with a robot.

This is the same move I used to make with a good freelancer, back when good freelancers were the whole system. Send the style guide once. Send examples of the tone I wanted. Make sure they understood the reader before they ever wrote a sentence.

The only thing that’s changed is where that information lives now, and how long it stays useful once it’s there. A freelancer forgets, moves on, or needs re-briefing every time someone new joins the account. A Project doesn’t forget anything, and it doesn’t need replacing.

A Skill Means You Never Explain the Same Structure Twice

The Project handles who I am. A Skill handles how I want something specific built.

I’ve written a handful of Skill files, plain instructions that tell Claude exactly how I want a certain kind of piece structured, what to check before calling it finished, and what I never want it to do. Once one exists, I don’t retype those instructions the next time. Claude reads the Skill and applies it without me asking twice.

Before Skills, I was re-explaining my preferences in nearly every conversation. Keep paragraphs short. Watch the rhythm. No em dashes, and I mean it this time. I repeated those instructions section by section, sometimes line by line. By draft four, even a good freelancer can miss details like these when they are scattered across old conversations instead of built into the process.

None of it needs to be complicated to work. A Skill is a handful of plain instructions in a text file, not a piece of software. The value isn’t in how sophisticated it looks. It’s in the fact that it exists at all, sitting there the next time you need it instead of living only in your head.

Your Research Doesn’t Have to Leave the Docs It’s Already In

I keep research, notes, and source material in Notion. Most of what I’d pulled for that article was already sitting there before I ever opened Claude.

I connected Claude to Notion once, through what’s called a connector, and now it can go find that material directly instead of me copying it into the chat by hand every time. No exporting a page to a doc first. No pasting half-formatted notes and hoping the formatting survives. No losing track of which version was current and which was three edits old.

That sounds like a small convenience until you actually notice what it replaces. It replaces the fifteen minutes before every draft where you’re hunting through old docs for the right source, copying the right section, and hoping you grabbed the current version and not an outdated one sitting two folders over.

Fifteen minutes doesn’t sound like much on its own. Multiply it by every article you write in a year, and it stops being a small inconvenience. It’s hours that never needed to be spent.

Every Brief After This One Gets Shorter

Setting up the Claude Project and writing the first Skill for it took real time. Connecting Notion took a few minutes that I probably could have spent on writing. None of that was from a click of a button, and I won’t pretend it was.

But that time cost only gets paid once. The next article didn’t need me to re-explain my brand voice, because the Project already had it. The one after that didn’t need me to redescribe the structure I like, because the Skill already covered it. Each brief I wrote got shorter than the one before it, because less of it needed explaining from scratch every time.

That’s the actual difference between a fast draft and a system that compounds. A fast draft is a nice result on one Tuesday. It’s also exactly what everyone chasing “write me fifteen hundred words” is hoping to stumble into, right before their traffic drops six months later because the fast draft never had anything valuable or different behind it in the first place.

A system that gets cheaper to run every time you use it is a different thing entirely. It’s the thing that was missing from every one-line prompt I ever tried, and it’s the thing nobody’s selling you when they sell you a faster way to type a prompt.

Here’s the chain, laid out plainly. A vague prompt produces a generic draft. The generic draft gets published anyway, because rewriting it from scratch feels like more work than it’s worth. The draft sits there doing nothing for 4 to 6 months, and then the traffic proves it. Nobody planned for that outcome. It’s just what happens when the setup never gets built, article after article, each one starting from zero.

That’s not a knock on anyone stuck in that loop. Nobody sits down and decides to build something disposable on purpose. They just never had a reason to stop and ask why the draft came back so thin in the first place.

You Don’t Need Budget Authority to Build This

None of what I just described needed a purchase order.

I didn’t hire anyone. I didn’t wait on a budget approval or a tool rollout from someone three levels above me. A Project, a Skill, and a Connector are things one person builds in an afternoon, on a plan you likely already have access to.

That matters when you can suggest a better way of working but do not have the authority to approve a budget or tool rollout. You do not need permission to test this setup. You need an afternoon and a real brief for the next piece you already plan to write.

Build it once on your own work, and you’ll have something better than an opinion to bring into the next meeting. You’ll have a draft that came back sharper than what you handed it, built from a system you set up yourself, on your own time, with nothing to approve.

There’s a big difference between telling a room you think the team should use AI more, and putting an actual draft in front of them that already sounds like the business, already followed a real structure, and already took a fraction of the usual time. One is an opinion. The other is proof, and proof is what actually moves a meeting.

The System Gets Faster Every Time You Use It

The old process took two to three weeks and ended with an article I hoped would rank. The new one takes about a day, and the day keeps getting shorter the more I use it.

One good draft is what everyone still chasing “write me fifteen hundred words” is hoping for, right before their traffic drops off because the draft never had real research or a real point of view behind it.

What I built wasn’t a better draft. It was a setup that keeps a real brief, a real voice, and real research already in place, so every draft after the first one starts further along than the last.

The skepticism I started this piece with hasn’t fully gone away, and it probably shouldn’t. It’s the same instinct that makes you check a draft instead of publishing it blind. What changed is that I finally had a system worth trusting, one I built once that keeps getting cheaper and faster every single time I use it.

That’s the difference between publishing content and building something that keeps paying you back for work you already did. It’s Copy That Compounds in miniature, actually. Build the asset once, and let it keep working long after the moment that made you build it in the first place.

Once you’ve felt that difference, you don’t go back to typing a topic and a word count and hoping for the best.

What People Ask Before They Set This Up

Does Claude replace a writer?

No, Claude replaces the blank page, not the judgment. It still needs your sources, your point of view, and your structure to produce something worth publishing. Claude speeds up the drafting. You’re still the one deciding what the article should actually argue.

How long does the setup take?

About an afternoon for the first pass. Building the Project, writing one Skill, and connecting a research source usually takes a few focused hours, not a project plan. You can refine all three later. The first version just needs to be good enough to test on a real article.

Is the draft actually publish-ready?

Usually close, but not automatically finished. Expect a strong draft that follows your structure and holds up on research, with room for a real edit pass before it goes live. Publish-ready means you’re editing a good draft, not rebuilding a weak one from scratch.

What’s the difference between a Project and a Skill?

A Project holds who you are. A Skill holds how you want something specific done. Your Project carries brand voice and context across everything you write. A Skill applies to one kind of task, like structuring an article a certain way, every time that task comes up.

PS... If your content is coming back generic no matter how carefully you prompt it, the fix usually isn’t the prompt. Let’s look at the actual setup together → Let’s chat.

PPS... If you’d rather build it yourself first, the Copy That Compounds dashboard walks you through setting up the same kind of Project, Skill, and Connector system this article just described → Content Playbook.

Over the past ten years, I have written more than 1,000 articles for local businesses, founders, and my own projects.

That work has helped produce results such as:

→ More than 1,400 store visits from one article about a celebrity product launch

→ More than 2,000 monthly visitors for a local business that started at zero

→ Even earned a trusted citation from Wikipedia

The writing mattered, but the results came from giving each article a clear job. The topic matched what people were searching for, the angle made sense for the business, and the page gave readers enough proof to trust what they found.

Over the past three years, Claude has helped me bring more of that work into one place and complete it faster without lowering the standard.

The AI SEO Playbook for Founders Who Use Claude

I have always used software to help me publish articles. For most of those years, that meant moving between research tools, documents, spreadsheets, editors, and image programs.

It worked, but the process was held together with more pieces than it needed.

Today, most of that work lives inside one Claude system.

Here is the work it brings together:

The researcher, who finds the topics people are searching for and gathers the sources, examples, and details the article will need.

The strategist, who chooses the right topic for the business and sets the angle and structure before the first draft is written.

The copywriter, who turns the research and strategy into an article people can understand and enjoy reading.

The editor, who checks the facts, protects the founder’s voice, and removes anything that does not earn its place.

The SEO lead, who prepares the title, description, headings, and internal links for Google and AI search.

The image creator, who builds the visuals the article needs.

The distributor, who publishes the article in the right place and turns it into smaller posts that lead people back to it.

The Copy That Compounds AI SEO Playbook takes the decisions behind those results and turns them into a process you can run with Claude.

It covers the work from early research to a publish-ready article while keeping your experience and judgment at the center.

The playbook helps you avoid:

→ Losing days researching before you can begin writing

→ Spending hours repairing Claude drafts that are generic

→ Building every brief, prompt, and workflow again from the beginning

→ Publishing finished articles with no clear job after they go live

These problems do not start with the writing. They start with the system around it.

A Finished Draft Is Only Part of the Job

Claude can produce a clean article in minutes, but a clean article can still disappear after it is published.

The page needs a topic people are actually searching for, useful information, real proof, and a clear connection to the rest of your website. Its structure should also help readers, Google, and AI search understand what the page covers and why the source should be trusted.

Most Claude workflows stop when the draft is finished. That solves the blank page without solving what happens after the article goes live.

This playbook helps you build articles with a job beyond filling a content calendar.

You will learn how to:

→ Research and choose an article topic before asking Claude to write

→ Build briefs and prompts around the way your business works

→ Improve weak drafts without spending hours rewriting them

→ Add experience, proof, and examples that make the article worth trusting

→ Prepare revenue-driven articles for Google and AI search

Ready to Outsource More of Your AI SEO Workflow to Claude?

And Stop Burning Weeks on Just One Article

Gone are the days where an article takes three or four days to research, another week or two to write, and several long nights to edit.

A CMO recently told me this system saved her company $225,000 a year.

That kind of result comes from giving Claude a full process for research, planning, writing, editing, and preparing articles for search.

For founders, it means turning more of their ideas into revenue-driven articles without hiring a full marketing team or starting from scratch every time.

A system like this usually comes with a $9,000 price tag. I am not doing that here.

This first edition is $0.99. I kept it that low on purpose for founders who want to get in early and are willing to tell me what works, what does not, and where the instructions need to be clearer.

The founders who learn this process first get a real head start. They can build a library of articles for Google and AI search while everyone else is still writing one article at a time by hand.

It is $0.99 right now, but that will not be the price for long.

No subscription, no upsell, just the system.

Get the AI SEO Playbook

This is the start of my AI SEO With Claude series on Copy That Compounds.

I’ll be sharing how I use Claude to research, plan, write, and improve evergreen articles built for Google and AI search.

If you are a founder who wants your ideas to be easier to find, this series is for you.

Here is what I’ll cover:

→ How I find topics people are already searching for

→ How I turn a founder’s experience into a brief Claude can follow

→ How I improve Claude’s drafts before they are published

→ How I structure articles for Google and AI search

→ How I update articles so they remain useful over time

These lessons come from more than ten years of writing hundreds of articles, including work that has ranked on Google and earned citations from sources like Wikipedia.

Subscribe below to follow the series.

Each article will stand on its own, so you can start wherever the subject fits your business.

Talk soon!
Jeromie Rosa
Website: jeromierosa.com