YouTube's second-biggest channel is full of AI babies

Youtube Is Not Looking So Hot Right Now

—by Adam Bumas

Last week, YouTube announced it was tightening its bans on monetizing “inauthentic,” “repetitive,” and “mass-produced” content. People were so confused they had to make a not-very-clear clarification. They emphasized they weren’t banning AI, just requiring it to be labeled, and limiting ad revenue on dozens of near-identical AI-generated videos from the same account.

The restrictions were slated to go into effect yesterday, but we aren’t seeing much of a change yet. According to data from Playboard, the second-most subscribed channel on Tuesday is still breaking all the rules YouTube is now supposed to be enforcing. In the past week, it’s uploaded nearly-identical variations of the exact same “AI baby in danger” video to YouTube Shorts seven times, including twice since the new regulations were supposed to start. And for the cherry on top, it’s not even an original disturbing AI short — it’s so obviously stolen from another account they left in the watermark!

This is the exact “mass-produced” stuff the platform is specifically trying to cut down on. The video doesn’t appear to be monetized, and it’s labeled “synthetic content,” but data shows it’s getting similar engagement numbers as before, to the tune of millions of views and hundreds of thousands of likes. YouTube has spent years going full speed on Shorts and AI-generated video, and now it’s gotten so bad that everyone from uploaders to viewers to advertisers are starting to notice something’s wrong. But the YouTubers — the ones who make a living off the platform — were the first to notice.

For more than a year, there’s been a slow exodus to Patreon as revenue from the platform is no longer enough on its own. The creators have all been clear that the problem has been the algorithm. YouTube continues to prioritize channels that upload as frequently as possible. This is why the “mass production” is a problem in the first place, but it has the side effect of taking attention away from creators who actually put effort and production value and, y’know, time into their videos.

For a while, it looked like the solution was Beastification. Billionaire and failed thumbnail hawker MrBeast really did systematically decode YouTube’s algorithm, and so it looked like getting attention for your own videos was as simple as copying everything he did. But what MrBeast does is constantly changing. Like how he famously talked about trying to minimize any personality, only to later say “more personality” improved engagement. And now, it looks like people following his lead should be moving focus away from YouTube entirely, since lately he’s put a lot more energy into his just-renewed game show and into (posting to and maybe buying) TikTok. 

But back on YouTube, not everyone has an escape hatch, and things have just been getting worse. Especially now there are Shorts to contend with. Just like all the bandwagoners trying to copy MrBeast, YouTube has been constantly changing the nature of Shorts to keep up with TikTok. YouTube’s team aren’t blind to the problems they’re creating, they just don’t seem to care. In 2023, the product lead for Shorts said in an interview that making Shorts longer than 60 seconds would “blur the line” with normal YouTube videos. A year later, they made Shorts three minutes long anyway. Similarly, their first comment on AI-generated videos said they would require them to be labeled, but this May, the most-subscribed channel on YouTube uploaded AI-generated music videos without any label.

And now ordinary viewers are noticing. Last Week Tonight doing an episode on the problem last month shows this is a mainstream talking point. Earlier this month, a post on X complaining about YouTube Shorts went so viral that YouTube’s staff account had to reply — just two days before they announced those monetizing changes. But even assuming those changes actually make a difference, they’ve still made someone else angry: Advertisers.

Earlier this month, a report by Adalytics covered how YouTube is very bad at restricting ad revenue from videos that violate their policies, and even worse about telling the advertisers when that happens and refunding them. Coverage by The New York Times has focused on how bootlegs of Hollywood blockbusters are getting ad money from the platform, but to me, that second part is an even bigger problem.

Because this isn’t even the first time this year YouTube has quietly made money off scammers capitalizing on movie studio prestige. This new report, which claims that YouTube will only provide specific information on videos to advertisers if they ask for it within a few days, shows the same intent is holding true. As a website and a business, it looks like YouTube has been relying on people not noticing the worst parts of the platform.

But it’s clear now that they’re too big to keep flying under the radar. According to Nielsen, YouTube has been the single most popular thing on TVs for months. The site celebrated its 20th anniversary this year. (When I was installing my AC unit, I lost count of how many contractors, service reps, and even product manuals foisted me off onto YouTube.) It’s not just a massively successful app, it’s probably the most important cultural institution of the century. And they seem to be intent on destroying all of that, clogging it all up with identical, disturbing videos, and it’ll take more than some vague updated policies to get back on track.

The following is NOT a paid ad, but a good old fashion promo4promo. If you’re interested in advertising, though, email me at [email protected] and let’s talk. Thanks!

The Fence – billed as The UK’s Only Magazine - is a quarterly print mag stuffed with facts, features and funnies brought straight from London's Soho. 

It also boasts the UK's only newsletter, Off The Fence, to which you can subscribe here, for a free, weekly dose of investigations, gossip and stupid gags.

The Fence contains no ads, no photos, just beautifully illustrated writing from the best writers around, be that a searing scoop about special needs schooling in the UK, the ins and outs of arranging kidnap and ransom payments from a corporate office in London, or an insider piece about what it was like to spend 9/11 working on a sexting hotline.

Subscribe today to get four bumper issues per year, all for a price so small it's practically insulting.

A Really Delightful Instagram Account

Derek Guy, The Menswear Guy, shared a few videos from this account on X this week. Twice Sold Tales is a bookstore in Seattle and it’s full of cats and the owner does a video series on Instagram that is super cute.

Republicans Can’t Untangle The Epstein Knot

Over 200 Republicans in the House of Representatives voted to block the release of the Epstein Files this week (assuming the Epstein Files actually exist, of course). Per The New Republic, not a single Republican voted for the measure, with nine Republicans abstaining. How curious!

President Donald Trump posted on Truth Social about the Epstein Files again, calling them “the Jeffrey Epstein Hoax,” concocted by Democrats. He also said that any of his supporters who continue bring up the files are “weaklings.” Truth Social users are not taking this well, of course. The top reply to Trump’s post this morning was from a user who asked, “Why is the Epstein case so important to the MAGA movement?”

Great question! Probably best not to think about why every single Republican is against the release of the client list of the world’s most prolific pedophile. I’m sure there’s a perfectly good explanation.

The wider MAGA internet is continuing to crash out over all of this. The world’s baldest man, Tim Pool, is on X trying to spin the push for the release of the list as Democrats voting “to have the US government publish child porn.” Which is not what’s happening at all, obviously. Joe Rogan is also flipping out about the government’s refusal to release the files. Matt Walsh is whining about it on X. And Rep. Nancy Mace posted on X, “All child r*pists should get the death penalty,” minutes after she voted against the release of the Epstein Files per a very funny Community Note on X. (In a followup post, she wrote that there was no vote, actually, and that it was all a Democrat trick. (There was a vote.))

Oh, also, according to WIRED, there are actually three whole minutes missing from the FBI’s Jeffrey Epstein security camera footage. Fascinating! Again, there’s probably a very good and normal explanation for that. Also, the Annoying Orange is demanding the Trump administration release the list. Sure why not.

Get Ready For A Lot Of Cringe Zohran Knockoffs

My favorite moment in an election is when a fresh, new candidate slides into the lead and totally changes the meta. It’s the moment when all the other candidates realize they’re probably cooked and have to face a hilariously awful choice: Do they copy the leader or do they stick to their campaign? I, personally, love when they decide to copy because it usually reveals how little the establishment understands why this fresh, new candidate is so fresh and new.

For instance, Zohran Mamdani is, most likely, going to be the next mayor of New York City. We’ve got a few more months until the election, but so far, he’s lightyears ahead of the competition. Former New York Govenor Andrew “Grandma Grim Reaper” Cuomo lost the primary but hasn’t dropped out of the race. And now his team is clearly forcing him to try and copy Mamdani’s social strategy, posting videos on X of him walking around New York City neighborhoods acting like he actually lives here. He’s also coopting Mamdani’s affordability platform. Also, Zohran ratioed Cuomo on X so bad his team briefly tried to hide the post lol.

There are going to be a lot of establishment Democrats LARPing as Mamdani very soon and I, personally, hope all of them are as embarrassing as Cuomo.

I Love Uncle Robot So Much

There’s a robot in China that’s dressed like Adam Sandler and runs like he shit himself and I love him. Users on X are calling him Uncle Robot and in a new video that hit Chinese social media this week he was posing for photos with his human fans (embedded above).

This looks like a Unitree H1 humanoid robot. There’s been a few videos featuring the H1 that have gone viral this summer. I’m sure this thing is very evil and a sign of a much colder, lonelier, less human future, but he’s also friend-shaped and I want to drink a beer with him.

People Are Gooning With Grok On The Timeline

OK so this is gross and funny, but it’s fascinating that this isn’t the first time something like this has happened. Back in May, when Meta released their new AI slop feed, I dug around on it and discovered it was full of older users who did not realize that their interactions with the AI were defaulting to public.

It seems like the next stage of evolution for social platforms is one where AI bots blur the line between what we used to do in private — search for information or, uh, in Grok’s case, sext — and what we typically do in public — socialize, share content, etc. And I haven’t really seen an AI-augmented platform figure out how to balance that yet. You can laugh at all the guys on X asking, “Grok is this real,” but it’s a fascinating emergent user behavior to keep an eye on.

A Good Post

Some Stray Links

• “How Social Media Is Fueling Gen Z’s Sex Recession”

• “The truth behind the endless ‘kids can’t read’ discourse”

• “I Tried Grok’s Built-In Anime Companion and It Called Me a Twat”

P.S. here’s a recipe for smack barm pey wet.

***Any typos in this email are on purpose actually***