How I Built a Faceless YouTube Channel to $41k a Month With Claude

How I Built a Faceless YouTube Channel to $41k a Month With Claude

Eight months ago I had a laptop, a $20 Claude subscription, and no idea what I was doing.

No camera, no editor, no team. Last month the channel did $9,400 and I've never shown my face once.

The whole thing is six steps. Claude does the heavy lifting on four of them. Here's the exact build, with the prompts I actually paste.

First, the money

Everyone asks what YouTube pays per view. Wrong question. The niche decides everything:

Finance and investing: $15-50 RPM Technology and AI: $12-30 Health and longevity: $10-25 General and entertainment: $3-8

My channel is finance. Around 480,000 views last month at an $18 RPM is about $8,600 from AdSense alone, before one affiliate link. A cooking channel at $3 RPM needs five million views for the same check.

Views are the engine. The niche is the fuel.

Step 1: Pick the niche by RPM, not by what you like

I picked a cooking idea first because I liked it. That mistake cost me three weeks. Now I make Claude run the math before I commit:

1Act as a YouTube channel strategist. I want to start a faceless channel.
2
3Analyze these 5 niches: [list yours]
4
5For each give me:
6
71. Estimated RPM range
8
92. Competition level (low / medium / high)
10
113. Content repeatability score (1-10)
12
134. Audience size potential
14
155. Monetization options beyond AdSense
16
176. One untapped content angle nobody is doing
18
19Then build a 52-video title calendar for the winner. If you can't fill 52, the niche is too narrow.
20Final recommendation: which niche to start, and why.

It handed me "debt payoff for people on a normal salary" inside finance, plus a year of titles. That's a $300 consultant call for free.

Step 2: Write the script. This is the actual business

My first 5 videos flopped. Clean voice, fine footage, 200 views each. The problem was the writing.

YouTube only really measures one thing, do people keep watching. A script that holds 70% gets pushed to new viewers, one that holds 30% gets buried. That gap is the script, not the production.

This is the prompt I paste for every video:

1You are a senior YouTube scriptwriter for a faceless finance channel.
2
3Write a full script for: "[title]"
4
5Length: 10 minutes (~1,500 words). Audience: [who]
6
7Structure:
8
91. Hook (0:00-0:30): open on a shocking number or counterintuitive claim. No intro, no "hey guys".
10
112. Problem amplification (0:30-1:30): make them feel it with specific numbers, use "you" constantly.
12
133. Credibility bridge (1:30-2:00): one reason to trust this, under 30 seconds.
14
154. Main body (2:00-9:00): 3-5 sections. Each one a bold claim, a real example with numbers, one usable takeaway, a bridge to the next. A pattern interrupt every 90 seconds.
16
175. Close (9:00-10:00): call back to the hook, one action, earn the subscribe.
18
19Rules: write for speaking, short sentences, active voice, no filler. Retention target 70%.
20Output: full script, estimated runtime, 5 thumbnail concepts, 3 titles (curiosity / SEO / emotional).

Once I fixed the script, the same channel that did 200 views started doing 40,000. Nothing else has changed.

Step 3: Voice it once, keep that voice forever

ElevenLabs reads the script. I picked one voice and never change it, so the channel keeps an identity viewers recognize. Set the speed to 0.9x for finance, it reads with more authority. Drop [pause] markers into the script where it should breathe, that's what makes it stop sounding like a robot

$22 a month. This step is not where you win or lose, so don't overthink it.

Step 4: Assemble it in CapCut

No camera means no editing, just matching footage to the voiceover. The rule: the visual changes every 3 to 5 seconds. Pexels and Pixabay cover most of the B-roll for free.

Turn on CapCut's auto-captions. They lifted my watch time about 15% because most people scroll with the sound off.

Step 5: Let Claude write the metadata

I had one video that deserved 80,000 views sitting at 800. Great script, clean voice, decent thumbnail. The title was terrible.

Now Claude writes the whole metadata layer in one shot:

1Act as a YouTube SEO specialist.
2
3Video topic: [topic]. Niche: finance. Script summary: [2-3 sentences]
4
5Output:
6
71. Title under 60 characters, main keyword, triggers curiosity
8
92. Description, 200 words, SEO optimized, with timestamps and a CTA
10
113. 15 tags, broad and specific mixed
12
134. 5 chapter titles with timestamps
14
155. 3 pinned-comment options to boost engagement
16
17Optimize for click-through rate and watch time.

That one prompt is worth $50 a video if you hire it out. I run it in 20 seconds.

Step 6: Post 20 to 30 videos before you judge anything

I almost quit at video 8. Nothing was moving. The channel didn't start climbing until video 24.

The algorithm needs 20 to 30 videos before it learns who to show you to. Most people post 8, see flat numbers, and walk away one month before the curve turns. Same day, same time, every week, no exceptions for the first six months.

What it costs

Claude: $20/mo ElevenLabs: $22/mo CapCut, Canva, Pexels: free Total: ~$57/mo

Against what this replaces, a scriptwriter at $150-400 a video, a voice actor, an editor, an SEO person. I'm running all of it on the price of two coffees a month.

The part nobody tells you

For fifteen years the barrier to YouTube was the camera. You needed a face, a voice, presence, gear, years of reps on camera.

That barrier is gone, and losing it showed me something that was always true. The camera was a distraction from the actual job, which is writing things people don't want to stop watching.

The faceless channel didn't remove the hard part. It removed everything that was hiding it. I stopped competing on being watchable and started competing on being un-skippable.

That one's a writing problem, and Claude already solved it.