AI Slop on YouTube: How Respondemesta Makes It Human

You're not going to beat automated channels by uploading faster. They don't sleep, overthink thumbnails, or suddenly decide the entire video needs a new intro.

But one Colombian creator has found a useful weakness: AI can invent the spectacle. Humans can make people care about it.

A human copied the machines

Respondemesta has been recreating the impossible food seen in AI-generated videos. Strange watermelons. Questionable kitchen experiments. Objects that should absolutely not be edible.

The important bit isn't the food. It's the conversion of a synthetic idea into a real challenge with mistakes, reactions and an actual payoff. Translation tools then help the longer videos travel beyond Spanish-speaking viewers.

That formula added roughly 310,000 subscribers during one chart week and placed Respondemesta 13th globally for subscriber growth. The climb kept going. A third-party channel tracker¹ recorded 4.57 million subscribers on September 21, up around 1.39 million over 30 days.

Meanwhile, 38 of that week's 50 most-watched channels were primarily publishing Shorts. Yes, the vertical conveyor belt is still running at full speed.

The slop is winning attention

This isn't a small corner of YouTube anymore. Kapwing's research² found that 21% of the first 500 Shorts shown to a fresh YouTube account were identifiable AI slop. Its equivalent TikTok test produced an even uglier number: 59%.

That volume changes viewer behaviour. People become suspicious, scroll faster and stop rewarding polish for its own sake. A perfectly generated clip is no longer impressive. Sometimes it's barely noticed.

Respondemesta adds what the machine removed: consequences. Will this ridiculous thing work? What does it look like in real life? How badly will the attempt go sideways?

Don't compete with infinite output. Compete with having a pulse.

There's also a money angle. YouTube's monetization rules³ now explicitly call out repetitive and mass-produced material. AI itself isn't banned, and an AI label alone doesn't kill distribution or earnings. Low-effort sameness is the risk.

Build around the human gap

First, study synthetic trends as raw material, not as your finished product. Find the visual idea already stopping thumbs, then ask what only a person could test, explain, ruin or improve.

Second, capture proof. Show your hands, process, failed attempts and decisions. These aren't messy leftovers. They're increasingly the valuable part.

Third, use AI backstage. Translate, research, subtitle and prototype with it. Keep your face, judgement and lived experience out front.

And slow down before copying the latest automated format. If anyone can generate it in three prompts, your audience can replace you in three swipes.