I was pitched a $9,800 coaching program to grow a YouTube channel. Editing was $5,000 more. I said I’d think about it, went home, and asked Claude what was actually in the curriculum.
The short version. The program: $9,800 for six months. The build: one Claude project, a one-page channel brief and four prompts, on the Claude plan I already pay for ($20 a month, so $120 over the same six months). Verdict: Both. Build the method first, and buy coaching if what you’re missing is a person keeping you on track. The finished build is a $49 kit; everything it does is shown below.
What was in the program
Here’s what the sales call laid out. Six months. An online course you keep. A method for finding video ideas people already want, a system for titles, thumbnails and the first seconds of a video, and lessons on writing a script. Then the people: a private coach twice a month, group calls every week, and a community of other students. Expect four to six hours a week, I was told. I was also told the price would go up 15% the next month.
That’s a real curriculum. Finding small channels whose videos beat their subscriber count, making the title, thumbnail and opening promise the same thing, getting to the point fast: none of it is a gimmick. It’s how a lot of small channels actually grow. The price isn’t unusual for the category either. Public price pages put YouTube accelerators at $3,000 to $9,000 and up, and a year of 1:1 coaching at $1,188 to $2,388 (checked September 29, 2026).
So the program has two halves: the method and the people. I rebuilt the method. I did not rebuild the people. Keep that in mind until the end.
What I built instead
One Claude project. A project is a folder where the AI keeps your documents and your standing instructions. Mine holds two things:
- A one-page channel brief. Who my videos are for, what the channel is for, and how I make videos. Every answer reads it, so the advice fits my channel instead of being generic.
- Project instructions that tell Claude how to coach. They rest on four ideas: start from topics people already look for, make one promise per video, deliver it fast, and change one thing a week.
Then a step before the prompts, and four prompts:
- Step 0, find outliers. Small channels with a video that got far more views than they have subscribers. The gap means people who don’t follow that channel went looking for the topic.
- Prompt 1, pick a topic. Turns the outlier list into 20 topics, ranked by the evidence.
- Prompt 2, score the title. Checks that the title, the thumbnail text and the first 15 seconds make one promise, before you record.
- Prompt 3, write the script. Questions first, then an outline you approve, then the script.
- Prompt 4, review the numbers. Once a week, against your goal, and it picks one thing to change.
What it cost: $0 on top of what I already pay. I was paying for Claude Pro anyway: $20 a month billed monthly (claude.com/pricing, checked October 8, 2026). It does need a paid plan. I tried it on the free one and ran out of usage partway through Step 0.
What it took: building this first version took me longer than it should have, because I was also building the system I’ll use for every episode after it. I didn’t track the hours. Next time I expect it to take a fraction of the time. Using it is a different number: setting it up from scratch and running every step once took me about an hour and 15 minutes, and about 30 minutes of that was checking the outlier list by hand.
What broke, and how I fixed it
I didn’t write these prompts. I described what I wanted, Claude drafted them, and I ran them the way a beginner would. Here’s where they fell over.
- Prompt 1 buried the answer. The recommendation was right, but it sat at the bottom, under a table of ratios and medians that mean nothing to a beginner. I asked for the answer first, in plain English. Every prompt now opens with a short version.
- Prompt 2 gave a score and no way to use it. It graded five qualities out of 5 but never said which piece lost the points, what a 3 means, or what total is good enough to publish. I asked for all three. Now every lost point is tied to the title, the thumbnail text or the opening, with a fix, and there’s a bar: 20 of 25, nothing below 3. That bar is my rule of thumb, not YouTube’s.
- Prompt 4 had nothing to say to a new channel. No videos means no numbers. Now it says so and tells you what to write down instead.
- My own rules leaked in. The script prompt told every script to include a “what it cannot do” section, because that’s how I make videos. That’s my channel, not yours. I moved it into my own channel brief, and the brief now has a line anyone can use for what every one of their videos includes. The test I use now: every line in the kit has to make sense for a dog trainer’s channel and an accountant’s.
What it produced
I ran the whole chain on my own channel, which had zero videos at the time. The quotes below are Claude’s answers, word for word. […] marks a cut. The kit’s worked example is excerpts from this same run, so you can see what a good answer looks like before you paste anything.
Step 0 and prompt 1. My outlier list was short: three videos from two channels. The best was a channel with 144 subscribers whose video got 8,200 views, 57 times its subscriber count. Prompt 1 turned the list into 20 ranked topics, and the first thing it told me was the part I’d rather not hear:
I sorted your 3 videos into 3 viewer questions and turned them into 20 topics, ranked by how many times more views than subscribers each question earned. All the evidence is thin: 3 videos from 2 channels.
Then it checked the ranking against my goal, which I hadn’t asked it to do:
The strongest evidence comes from a life coach video, and people wanting a life coach are not necessarily solo operators weighing an expensive pitch.
Three rows from the table:
| # | Working title | Evidence | Flag |
|---|---|---|---|
| 1 | I rebuilt a $9,800 coaching program in a weekend with Claude | 57×, 1 channel | Describe the program, never name it |
| 2 | Build your own business coach in Claude Projects before you hire one | 57×, 1 channel | None |
| 9 | I replaced a [PRICE] VA’s inbox work with AI | 32×, 1 channel | Needs a real quote |
Topic 1 is this episode. It says “in a weekend” because that’s what my notes said at the start. It took longer.
Prompt 2. My first title package scored 20 out of 25:
20 out of 25, ready to publish, but only just. […] The fix that matters most is the opening: its first sentence is missing the word “costs”, and it never says who the video is for.
It rewrote the opening two ways, told me which one to ship, and left the title and thumbnail alone on purpose: change one thing at a time, or you won’t know which change helped. The new score was 22, and it said plainly that it was scoring its own rewrite, so 22 is an estimate.
Prompt 3. It didn’t start writing. It asked five questions first, each with an example answer, then proposed an outline and waited for me to approve it. Then it suggested a line I hadn’t asked for:
Your channel has 0 videos, so the build has produced a plan, not results. I would add one line […]: “This is video one, so I can’t yet tell you it works.”
I agreed. In the video it became “This is video one. Judge me by the next eight.” The script I recorded started as Claude’s. Then I put it in my own words.
Prompt 4. Week 0: zero views, zero impressions, one subscriber. It didn’t invent a pattern:
There is nothing to analyze yet: no video is published, so every number is zero. I skipped the analysis and set up what to record instead.
It listed six numbers to write down seven days after each video and where each one lives in YouTube Studio. It named impressions as the first number to watch. And it gave me one thing to do this week: publish episode 1, and check that the description link shows up in my site’s analytics. One thing, not ten.
What it cannot do
First, plainly: this will not grow your channel or make you money. It gives you the method and a routine. The program sold the same method plus people to keep you on it, and neither comes with a guarantee.
What a paid coach still does that this can’t:
- Check your work. No person looks at your numbers or your video.
- Give you a group. There’s no community. You’re doing this alone.
- Watch your video. Claude reads text. It can’t tell you your delivery is flat.
- Know from experience. It has never run a channel, and it doesn’t talk to students every day. A good coach has seen things it hasn’t.
Three parts of the program I didn’t rebuild at all: choosing keywords so a video gets found, making money from the channel, and editing, which was the $5,000 add-on.
And the method is only as good as your outlier list. My corner of YouTube is young, and three videos from two channels is thin. Prompt 1 ranks what you give it. It can’t find demand that isn’t there yet.
The curriculum was the cheap part. The $9,800 pays for someone who notices when you skip a week, and a folder of prompts doesn’t notice anything. Here’s what I’m doing about that: a partner, a Tuesday deadline, and prompt 4 on the same day every week. This is video one. Judge me by the next eight.
Verdict: Both
Build this first. Run it for a few weeks. Then, if what you’re missing is a person keeping you on track, buy coaching for that. It’s a good use of the money.
Buy instead of building if you already have a channel that works, cash to spare, and a time problem rather than a knowledge problem. That wasn’t me.
If you’d rather not build it yourself, the finished build is the Channel Strategy Kit: $49, once. Get the Channel Strategy Kit, $49 →
What’s in it:
- Setup for Claude, ChatGPT or Gemini, on a paid plan
- The project instructions, and a channel brief you fill in once (about 15 minutes)
- Step 0 and prompts 1 to 4, each with why it works, what to give it, and what a good answer looks like
- A worked example: excerpts from the run above
- A routine: what to run each month, each week before you record, and each week on the same day
- Free updates, email support, and a 7-day refund, no reason needed
Plan on about 30 minutes to set up and about an hour for your first full run. It’s for you if you want the method and will run it yourself. If what you’re missing is someone keeping you on track, the kit won’t fix that.