Paid Social

AI Is Manufacturing an Unlimited Supply of Cheap Ad Impressions

How advertising machines broke the attention economy
How advertising machines broke the attention economydjedj / pixabay

ChatGPT is on track for $1 billion in annual ad revenue with 100,000+ advertisers — the latest sign that AI is flooding a finite attention economy with cheap impressions.

By Marcus Bennett

3 min read

At a glance

  • ChatGPT says it's on track to generate $1 billion in annual ad revenue, with 100,000+ advertisers on the platform
  • BuzzFeed's 2016 exploding watermelon live stream drew 3 million views; three-second autoplay views counted as impressions
  • Fox Networks Group tested six-second spots and single-sponsor streams; buyers refused to pay more for less
  • The authors argue AI-generated content is creating a 'subprime market for human attention'

ChatGPT says it's on track to generate $1 billion in annual ad revenue, with more than 100,000 advertisers already running campaigns on the platform. That number, from a book excerpt by Joe Marchese and Jonathan Bing, Human Dot AI, is the clearest signal yet that advertising is entering its third act: after TV's decline and the mobile feed's glut, AI now generates unlimited content against a fixed supply of human attention.

The authors trace the cycle back to 2016, when cable cancellations ran into the millions, TV ratings slipped, and advertisers shifted billions into digital platforms promising bigger audiences at lower cost. That April, BuzzFeed live-streamed two staffers wrapping rubber bands around a watermelon until it burst. More than 3 million people viewed the stream; CNBC called it the future of television.

Why three seconds counted as much as a World Series game

It wasn't. Nobody could tell how much of that audience actually watched, and autoplay views counted anyway. Every three-second "view" registered as an impression — the industry's basic unit of value — and scrolling past an exploding watermelon was graded identically to watching every minute of a prime-time drama or a World Series game.

The mechanics matter for anyone buying media today. Algorithms were given one incentive: maximize ad revenue. So they produced whatever generated the most impressions, and when real attention ran short, they intercepted demand that already existed and claimed credit for purchases people would have made anyway. Supply exploded, prices fell for every seller, and the result was what the authors call "a subprime market for human attention."

What did TV try to do about it?

Plenty, and it mostly failed. Jonathan Bing, then in charge of advertising at Fox Networks Group, says the network tested six-second spots, premium slots at the start and end of commercial breaks, and single-sponsor streams with minimal interruptions. Other networks ran their own versions of fewer, better ads. The formats offered a better value exchange for viewers' attention, but buyers balked. "Buyers asked why six seconds should cost more than 30," the authors write. Brands wanted volume, and most refused to let their agencies take risks.

Meanwhile the vicious cycle compounded: networks ran more ad minutes, the viewing experience degraded, and viewers fled faster to streaming and ad-free platforms. The less attention was worth, the more impressions the industry had to manufacture to close the gap.

What does this change for practitioners?

The core tension hasn't moved, but its scale has. AI now automates every step of advertising — writing the ad, targeting the audience, placing the buy — and feeds are filling with AI-generated video, with ads now appearing inside AI assistants too. An endless supply of content chasing 24 hours a day of human attention means each impression gets cheaper and less meaningful, and the authors argue the subprime market will keep growing until advertising resembles "a toll on global commerce" rather than a market for attention.

That has concrete budget implications. If AI keeps optimizing everything measurable while neglectging what it can't measure — culture, storytelling, trust, human connection — the measurable stuff keeps deflating in value. The authors' counterargument: AI could instead be built to help people manage time, filter information and reclaim attention, freeing it for the stories and connections that actually matter. That's a design choice, not an inevitability, and it's the reset the industry rarely gets to attempt.

Source: fastcompany.com

Marcus Bennett

Show full bio

Market editor covering money, funding and market shifts around Social platform news and policy at Reach Brief.

More from Marcus Bennett →

Related articles

« Previous articleNext article »