I spent 30 days on AI videos and got 71 views. Here's what I did wrong

0:0071 views, seven videos, 30 days, two subscribers, and one of them's me. I'm going to spoil the ending because the ending isn't the interesting part. This channel, the one you're watching, where I talk about building faceless channels, did 8,755 views in the same 30 days. Same person, same free tools. The channel about the thing beat the thing by about 120 times. So, the question isn't whether it worked, 'cause it didn't. The question is why the wrong one won. In the end, that information turned out to be worth more than anything. Here's what I built. A faceless children's storybook channel.

0:42Forest animals, soft lighting, cozy bedtime vibes, fully AI assisted, start to finish. Why this niche? Two reasons. First, I wrote a children's book that's currently breaking records for lowest sales. Second, AI is actually really good at forest animals. Now, you give it a prompt about a rabbit in the morning light and boom, it actually looks like a rabbit in the morning light. What I thought was going to happen was me casually pumping out Pixar level animations with a few clicks. What actually happened is animation moves at the speed of sunbathing turtles. That was my whole entire reasoning. I picked the niche because it was easy to produce and I was already there. Hold on to that because it's the whole mistake and it took me 30 days and 71 views to see it. Full stack and I want to be specific because all tools can get expensive. Gemini for scripting, Flow for visuals, Adobe for assembly and captions. Claude for the parts I kept getting wrong, pacing and later reading my own analytics back to me when I couldn't tell what was happening. Total spend across 30 days, $72. 20 for Gemini, 20 for Claude, and 32 for the old Adobe subscription I was locked into. I was already using both Gemini and Claude for work. So, I had the subscriptions for what I was trying to accomplish. Not once did I think, if only I had the free versions. Every failure was mine, which is inconvenient because buying software is much easier than fixing your judgment.

2:10So, here's the judgment. Week one, I published four videos. I got six views. Six. That was with a Facebook page with 194 followers. And I posted on the page I was going to put some videos up on YouTube. Four videos, three different formats, a narrated storyboard story, an advertisement, and two Facebook reels. Four signals for the algorithm about who this channel was for. YouTube had no idea because I hadn't decided. Second mistake, and this one has a receipt. I put hashtags in my titles. One view, 490 views. Same topic, same week, same me. The only difference was hashtags in the titles. It was obvious I was trying to figure out what to do on YouTube. I want to be honest about the part nobody puts in these videos. Six tells you nothing. There's no signal in it. You can't test anything against six views. You can't read retention off it. You just sit there refreshing a page that doesn't change. And I kept changing things. New thumbnails, new titles, new format, new upload time every couple days because doing something felt better than doing nothing, which destroyed the only data I had. Every change reset the experiment. I spent 30 days running 30 different experiments with a sample size of zero. I wasn't optimizing. I was panicking with a nicer interface. 71 views. Let's do the money because it's shorter than you think. Forest Friends has earned $0.00. There's no RPM because there's no revenue because it isn't monetized and isn't close. The partner program needs 1,000 subscribers and 4,000 watch hours. I have two subscribers. And if every one of those 71 views had watched a full 5-minute video start to finish, which absolutely did not happen, that's 5.9 hours. 6 hours out of 4,000. Even monetized at a generous kids content rate, 71 views is about 14 cents. It's not a discouraging number. That's just the number. Every honest version of this math looks like this. Which is why nobody puts this slide in, of course. So, is faceless automation worth it in 2026? The pitch is that AI removes the hard part. It doesn't. It removes the slow part. Those are different problems. I can make a finished decent looking video in 90 minutes that would have taken two days 3 years ago. That's real. But the hard part was never production. The hard part is knowing what to make and for whom. AI didn't touch that. And it made it worse because now everyone can produce. So judgment is the only thing left that separates anybody. And here's the actual answer to why the wrong channel won. This channel makes things people type into a search bar. That's demand sitting there waiting. Forest Friends makes bedtime stories and nobody searches for bedtime stories. Kids content gets found through browse and suggested, which means you need an audience to get an audience. I built one channel with a front door and the other one with no door, then acted surprised when only one had visitors. Here's what I didn't expect. I started this channel to document the other one. It was supposed to be the footnote. It outgrew the thing it was about by 120 times. And it did that because it had a person in it, someone who got it wrong out loud. 240 people subscribed in a month to find out what happened to a channel with two subscribers. That's worth more right now than 14 cents. I'd rather have the 240 than the money. And I didn't know that 30 days ago. Day 60 numbers are coming. If Forest Friends is still sitting at 71, I'll show you that, too.