Nobody reads what AI writes

In 1971, the economist Herbert Simon observed that "a wealth of information creates a poverty of attention." Half a century later, we've tested that theory at scale. AI can now produce blog posts, newsletters, white papers, and marketing copy at near-zero marginal cost. The bottleneck was never production. It was always attention. When everyone can publish, nobody can be heard.

The supply side broke

Content creation used to be expensive. You needed time, skill, maybe an editor, maybe a designer. Those constraints acted as a natural filter. Not everything got published, and what did had at least some effort behind it. That filter is gone. Large language models can generate a passable 2,000-word article in seconds. SEO agencies are scaling from dozens of posts per month to thousands. A 2025 Semrush analysis found that Google's AI Overviews appeared in over 13% of U.S. desktop queries by March 2025, double the rate from January, while organic click-through rates dropped from 1.41% to 0.64% over the same period. The content is flooding in, and the clicks are drying up. The supply of words on the internet just went functionally infinite. But human attention stayed fixed at roughly 16 waking hours per day.

Writing is free, reading is expensive

The economics of content have quietly flipped. For most of the internet's history, the expensive part was creation. You paid writers, editors, and designers. Distribution was relatively cheap, just hit publish and let search engines do the work. Now creation costs almost nothing. A prompt and a few seconds of compute time can produce what used to take a writer a full day. But reading? Reading still costs the same: a person's time, focus, and cognitive energy. Those are finite resources that haven't scaled with the output of language models. This means the real cost in the content economy has shifted from the supply side to the demand side. The scarce resource isn't the ability to produce, it's the willingness to consume.

The machines are talking to each other

Here's the part that doesn't get discussed enough: a growing share of AI-generated content isn't being read by humans at all. It's being indexed by search crawlers, scraped by other AI models, summarized by aggregators, and fed back into training data. Pew Research found in 2025 that when Google's AI summaries appear on a results page, only about 8% of users click any external link, compared with 15% when no summary is shown. So who's reading all this content? Mostly other machines. We're approaching a strange loop where AI writes articles optimized for search engines, which are increasingly powered by AI that summarizes those articles so humans never have to read them. The content exists, technically. But it passes through the internet like background radiation, present everywhere, absorbed by almost no one.

The Spotify problem

There's a useful parallel in music. Streaming platforms gave us access to essentially every song ever recorded. The catalog is functionally infinite. And yet, most people listen to the same 50 songs on repeat. Infinite supply didn't create infinite exploration. It created decision fatigue, which pushed people toward the familiar, the curated, and the recommended. Spotify's Discover Weekly works not because it gives you more music, but because it gives you less, filtered through an algorithm that knows your taste. Content is heading the same way. When everything is available, the value shifts from creation to curation. The question isn't "can you make this?" It's "should anyone read this?"

Curation is the new creation

If content production is essentially free, then taste, judgment, and curation become the scarce resources. Someone still has to decide what's worth reading, sharing, and trusting. This is already playing out. Newsletters with strong editorial voices are thriving while generic SEO blogs fade into irrelevance. Readers are gravitating toward sources they trust, not sources that simply rank. As one analysis put it, "original thinking that hasn't been trained into a model stands out." The role of the writer is shifting. It's less about producing volume and more about filtering signal from noise. Less about having something to say and more about knowing what's worth saying.

The human signal

What cuts through an ocean of generated text? The things a model can't fake: lived experience, specific opinions, genuine uncertainty, and the willingness to be wrong in public. When you read something and think "a person actually wrote this," it's usually because of a detail that's too specific, too weird, or too honest to be generated. A story about a particular failure. An opinion that goes against the consensus. A reference that only someone who lived through something would make. Voice isn't a style setting you can toggle. It's the residue of a life lived and a perspective earned. That's what makes it hard to replicate and valuable to find.

"Just add AI content" is the new bad advice

Remember when the default startup advice was "just start a blog"? Every company launched a content marketing operation, most of it forgettable, and the internet filled up with listicles and SEO-optimized filler. The advice wasn't wrong exactly, but it was incomplete. It confused the act of publishing with the act of communicating. "Just add AI content" is the 2025 version of that advice. It sounds practical. It's easy to execute. And it misses the point entirely. The bottleneck isn't production capacity. It's whether anyone on the other end cares enough to read what you wrote. The companies and writers who will do well in this environment aren't the ones producing the most content. They're the ones producing content that's worth the reader's increasingly scarce attention.

What actually matters now

The attention economy hasn't changed its fundamental rules. Attention is still finite. Trust is still earned slowly. And the best way to get someone to read your work is still to write something that rewards the time they spend on it. What's changed is the baseline. When mediocre content is free and abundant, "good enough" disappears into the noise. The bar for earning attention has gone up, not because readers got pickier, but because the volume of content competing for their eyes went through the roof. The path forward isn't to write more. It's to write things that only you can write, shaped by what you've seen, what you believe, and what you're willing to stake a claim on. In a world of infinite content, the scarcest thing is a point of view.

References