AI Killed The Internet Star
From radio → video → internet → AI — each wave doesn’t just add a new format. It rewrites what “a creator” even is.
There’s a cultural breadcrumb trail hidden in three titles: “Video Killed the Radio Star” (a pop prophecy from the end of the 1970s), “Internet Killed the Video Star” (a 2006 tongue-in-cheek update from the early social era), and my new rendition of “AI Killed the Internet Star” the version we’re living through right now.
This article is about that third title. Not because the internet is disappearing, but because the internet is becoming reinforced training data — and that changes what gets created, what gets discovered, what gets trusted, and what gets paid.
1) Video Killed The Radio Star
Radio was pure audio. It was intimate. A voice in your car. A song in your kitchen. An invisible stage.
But when MTV launched; "Video Killed the Radio Star" by The Buggles was the very first music video played on MTV, airing at 12:01 a.m. on August 1, 1981
Video arrived with a brutal advantage: it didn’t just deliver the music — it delivered the face. The look. The style. The identity. That shift turned “artist” into “audiovisual brand.” In the new regime, the winners weren’t only the best musicians. They were the most watchable — iconic enough to be remembered in a single frame.
2) Internet Killed The Video Star
By the mid 2000’s there was a new default behavior: search replaced scheduling, on-demand replaced broadcast, and the feed replaced the channel.
The internet didn’t kill video. It killed the gatekeepers of video. You no longer needed to be selected. You just needed to be uploaded.
The Limousines was released Internet Killed The Video Star on on June 8, 2010.
Social Media turned the world into a massive open audition, and it created a new kind of celebrity: the internet-native creator who wins by speed, consistency, and community. In the internet era, distribution belonged to the platform — but identity belonged to the individual.
At least… for a while.
3) AI Killed The Internet Star
AI Generated Lyrics by ChatGPT and AI Generated Music by Suno
AI didn’t show up as a new platform. It showed up as a new layer — a system that can read, summarize, rewrite, remix, and generate content at the speed of electricity.
Trained on the entire history of media content and past 30 years of internet content. The uncomfortable truth is simple.
The text-based internet was always a dataset.
For roughly thirty years, the web turned human thought into searchable text: blogs and forums, Wikipedia and reviews, academic papers and tutorials, tweets and Reddit threads, support tickets and comment sections.
We wrote it for each other.
But large models can ingest it at a scale no human ever could. So the internet became a fossil record of the human mind — and AI became the creature that learned to animate it.
The Internet Starts to Echo
Here’s the twist: once generative tools become mainstream, the web begins filling with machine-written text (like this post). Not because people want to deceive, but because it’s faster, cheaper, and “works well enough,” and because algorithms reward volume.
The result is a self-referential loop: models trained on human text produce new text that gets published online and can end up feeding future training sets. Over time, synthetic output can crowd out human originality.
Some researchers describe the risk as model collapse — the idea that training too heavily on generated data can degrade future models and distort what they learn.
Even if you ignore the extreme versions, the weaker version is already visible. The average quality of searchable text feels lower. More pages feel written “for rankings” instead of humans. More accounts feel automated.
When the majority of the internet’s activity is non-human, the experience of the internet changes — even if the infrastructure stays alive.
