51,500 views on Instagram. 1,971 on YouTube. Last 30 days, same videos, posted the same days.
Bohemia Chronicles is my AI-run history channel: short animated reels about Czech history, rendered and published on autopilot from my own GPU. Both accounts started the same day. Both get every reel the pipeline produces.
Instagram distributes them. YouTube showed them 1,612 times in a whole month. That's a 26x gap with the content held constant, which makes it a rare thing in creator analytics: a controlled experiment I didn't have to design.
The 26x experiment is now a video.
Identical reels, 30 days, Instagram vs YouTube: 51,500 views against 1,971.
And the number that hurt more: 2 link clicks.
Five minutes, every number on screen, plus a look inside the studio that renders the reels.
There is also a video version of this essay:
Last 30 days, from each platform's own analytics:
Views: Instagram 51,500 · YouTube 1,971 — 26x
People reached: Instagram 25,562 viewers · YouTube 634 unique viewers — 40x
New followers: Instagram +334 · YouTube +2 — 167x
The gap widens at every step of the funnel. Instagram shows the videos to more people, and those people follow at a higher rate.
If you'd rather watch than read: I walked through the same numbers on video, including the studio that renders the reels.
My first guess was that YouTube viewers see the videos and don't like them.
YouTube reports 1,612 impressions for the whole month. That's the total number of times it offered my content to anyone. Instagram put the same videos in front of 25,562 people.
The audience that does arrive behaves fine: 97% new viewers and a small core of 14 people who keep coming back. The whole month adds up to about five hours of watch time.
The bottleneck is distribution. 1,612 impressions all month.
My read: Instagram is a reels platform. The feed is built around exactly this format, while Shorts is a feature bolted onto a platform whose users came for something else.
Fewer people live in the Shorts feed, so there's less distribution to hand out, especially to a two-month-old channel with no history. It's a hypothesis. The sample is two months of one channel.
What I can say: the popular explanation "your content just isn't good enough for YouTube" falls apart at 1,612 impressions. Nobody got the chance to reject anything.
Here's the part I'd skip if this were a growth thread on X.
Those 51,500 Instagram views produced 435 profile visits. And those 435 profile visits produced 2 clicks on the link in my bio.
The path is video, then profile, then link, then destination. Mine: 51,500, then 435, then 2.
I wrote a whole essay recently about a consultant whose pitch didn't match her archive, and my conclusion was that a funnel is the thing I was missing.
25,000 people a month see these videos. That converts to almost nothing unless every step between the video and the destination is built. A history-reel viewer has no reason to visit a bio, and a bio link has no reason to interest a history fan.
I have reach. The audience and the funnel are still missing.
I keep re-learning this with my own numbers.
Instagram gets the attention. It's the only channel where growth feeds itself: more reels bring more reach, and follows come with it. The pipeline publishes daily either way, so attention here means tuning topics and formats toward what IG rewards, not more hours.
YouTube changes role. I'm not deleting it. The marginal cost of publishing there is zero; my renderer doesn't charge per upload.
But I stop expecting Shorts distribution and start treating the channel as an archive that will matter later, when I publish long-form there. That part has already started: the video version of this experiment is the first long-form upload. Search traffic works on a different clock than feed traffic.
The bio link gets a reason to exist. Two clicks means the link was decoration. If the destination isn't something a history viewer wants, reach won't fix it.
A restack is two clicks. I now know exactly what two clicks are worth.
Every video in this experiment came out of Reel Pipeline Studio, the tool I'm building to run faceless channels end to end on your own GPU. It still takes real work — picking angles, approving shots, re-rendering the ones that come out wrong. What it doesn't take is a timeline, a subscription, or a per-video bill. It launches in October, and the waitlist is open if you want the build and the numbers as they happen.
Hi, I’m Daniel — a software engineer and solo builder behind Digital Craft Workshop.
Ten years of TypeScript, .NET, and React for a living. In the evenings I build small digital properties: AI video tools like the Reel Pipeline Kit, and a history reel channel that publishes without me touching it. Some shipped, some I killed, and I write about both.
The long game is funding what I’ve wanted since I was a kid with a borrowed GameBoy: my own game worlds.
Each essay is something you can read in one sitting and use the same week. Everything here I actually ran, including the parts that broke.
No posts
Comments
Nothing yet. Say the first thing.
Sign in to join the conversation.