What These “Make Money With AI” Videos Don't Tell You — Part 2: Platform Models

Part 1 covered the content models — books, affiliate sites, ad-supported tools — where the hidden problem was always distribution: getting someone to actually find the thing you made. This half is different. These three models — YouTube, Etsy, and app stores — don't have a findability problem so much as a landlord problem. You're not just competing for attention. You're building your entire income on land you don't own, under rules you don't write, that can change without your input and without warning.

YouTube: What About AI Demonetization in 2026?

This is the one I want to spend the most time on, because it's the clearest, most current example of the whole danger in this half of the post.

If you've watched any of these videos recently, you've probably seen the “faceless AI channel” pitch — AI-written scripts, AI voiceover, stock or AI-generated footage, no host, no face, scaled across dozens of videos a week. Here's what a lot of these videos either don't know or conveniently leave out: that exact format is currently in the highest-risk category on the platform. In January 2026, days after YouTube CEO Neal Mohan's annual letter named “AI slop” a top priority for the year, the platform permanently terminated 16 channels in a single enforcement wave — not just demonetised, deleted entirely — wiping out a combined 35 million subscribers, roughly 4.7 billion lifetime views, and an estimated $10 million in annual ad revenue. A second sweep, a few months later, expanded further — catching legitimate creators too, including faceless exam-prep channels and documentary-style videos with no visible host, channels that thought they were doing nothing wrong.

Think about what that actually means for the pitch you're being sold. The strategy and the crackdown are pointed directly at each other. You're not just building on someone else's platform — you're building the exact format that platform has explicitly decided it wants less of.

A few other things worth knowing:

• You don't earn a cent until you clear a threshold. YouTube's Partner Program requires 1,000 subscribers and 4,000 watch hours before monetization even switches on. Plenty of people quit, or run out of budget, well before reaching it.
• Copyright claims can wipe out revenue after the fact. Reused stock footage, AI voices trained on existing material, or unlicensed background music can trigger claims that retroactively strip earnings from a video that was already live and performing.
• It's genuinely, heavily dependent on one company's whim. Algorithm changes, policy updates, monetization rule changes — none of it requires your input or agreement, and all of it can happen overnight.

Etsy / Print-on-Demand: Is It Still Worth It?

This one has a lot in common with the Amazon KDP problem from Part 1, and for the same underlying reason: a pitch that feels replicable at scale runs into the fact that thousands of other people are running the identical playbook simultaneously.

• The niches are getting tapped just as fast as KDP's are. Generate 50 designs in a trending style, list them, repeat — and so is everyone else who watched the same video. Differentiation evaporates fast when the entire catalogue is generated the same way, from the same prompts, in the same style.
• Etsy has its own AI-disclosure crackdown, similar in spirit to Amazon's. Flooding a niche with obviously AI-generated, undisclosed designs is a policy risk on top of a competition problem.
• The margins are thinner than the demo suggests. Listing fees, transaction fees, and payment processing fees stack up per sale, and rarely make it into the “profit per item” number shown in the pitch.
• Copyright exposure is real and mostly unaddressed. Image generators trained on existing artists' styles create genuine legal ambiguity around what you actually have the right to sell — something a “pick a trending niche and generate away” tutorial has no interest in mentioning.

Apple/Android Apps: What Happens to the “AI Wrapper” Business Model?

This is probably the sneakiest one, because it looks the most “technical” and therefore the most legitimate — surely if you can actually build and ship an app, that's real skill, real work, a real asset? Sometimes. But a lot of what's being pitched here isn't really a product. It's a thin interface sitting on top of someone else's AI model, usually accessed through an API.

• The wrapper problem is a moat problem. If the underlying AI provider ships the same feature natively — which happens constantly — the entire reason your app exists can disappear overnight. You're not just dependent on one platform (the app store); you're dependent on a second one underneath it (whichever AI company's API you're calling), and you control neither.
• There's a real, ongoing cost most demos don't show. A developer account fee, plus whatever you're paying per API call at actual usage volume — not the free tier the demo was recorded on. Costs that looked negligible in testing can become the entire problem once you have real users.
• Discovery is pay-to-play now. Organic downloads with zero user-acquisition budget are extremely rare. “Build it and they'll find it” essentially doesn't happen anymore in a store with millions of listings.
• Just existing isn't the same as anyone wanting it. Having a working app in the store is the easy part. Getting a stranger to notice it, trust it, download it, and actually pay for it is the entire business — and it's the part the tutorial glossed over in about thirty seconds.
• Maintenance never stops. OS updates, API changes, deprecated libraries — an app needs continual work just to keep functioning. It's not the “build once, collect forever” asset it's often pitched as.

The pattern, again

Look at what's actually shared across all three of these, and it's a different flavour of the same lesson from Part 1. These aren't findability problems — traffic isn't really the bottleneck for a viral video or a good app idea. The real risk is that you don't own the ground you're standing on. YouTube can reclassify your entire format as a violation. Etsy can flag your listings. Apple or Google can reject an update, change a policy, or simply get outcompeted by the very API your app depends on. None of that requires your permission, and none of it comes with much warning.

None of this means these models are scams, or that nobody succeeds at them — plenty of people do, genuinely. What it means is that the version being pitched in a 12-minute video has quietly edited out the actual risk, the actual cost, and the actual time it takes, because “here's the hard, boring, slow part” doesn't get clicks the way “I made $10k this month” does.

If you're going to try one of these anyway — and there's nothing wrong with trying — go in with the real picture, not the highlight reel. Know what you don't control, know what it actually costs at real scale, and know that the video you watched showed you the one that worked, never the forty that didn't.

References

YouTube Wiped 35M Subscribers Over AI Slop: Now It's Judging Your Taste — Tech Times, July 2026 (techtimes.com/articles/320629).

YouTube Partner Program: Requirements & Payouts (2026) — 1of10.com/blog/youtube-partner-program — current monetization eligibility thresholds.