Nothing Left But the Stump

Disclosure: This reflects my personal experience and interpretation of publicly available information. It represents my views alone—not any employer or organization—and is not professional advice.

The content industry stripped the Giving Tree of search until it had nothing left to give.

People didn’t stop reading. We removed the need to. And now everyone—from Gen Z to retirees—is converging on the same behavior: they don’t want content. They want answers.
Everyone keeps saying people don’t read anymore. That’s true. What’s not true is the explanation.

It wasn’t a cultural failure. It wasn’t a collapse in attention span. And it definitely wasn’t because kids these days lack intellectual seriousness—a comforting story that lets everyone over forty feel superior while ignoring what actually happened. It was engineered. Deliberately, incrementally, and profitably. And the people wringing their hands about the death of reading are, in many cases, the same ones who helped build the system that killed it.

For the last twenty years, we’ve been redesigning how people interact with information, one small optimization at a time. Faster search. Better ranking. Cleaner interfaces. Featured snippets. Autocomplete. Summaries at the top of the page before you even scroll. Each one removed a little friction. And each one removed a reason to read.

At first, you still had to click through to a page. Then you only had to skim. Now you don’t even have to leave the search bar.

But the interface was only half the problem. The other half was what happened to the content itself.

If you know Shel Silverstein’s The Giving Tree, you know how this goes. The tree gives the boy apples. Then branches. Then the trunk. Each time, the boy takes what’s offered and comes back for more—not out of malice, but because the system allows it. The story ends with a stump.

Search was the giving tree. It gave the content industry traffic—enormous, free, reliable traffic. And the industry took it. First, the easy fruit: useful articles that answered real questions and earned their rankings. Then the branches: optimized content engineered to rank whether or not it was worth reading. Then the trunk itself: entire publishing operations strip-mined for search volume, producing nothing of lasting value but generating just enough clicks to keep the ad revenue flowing.

Once ranking became the primary measure of visibility, quality became irrelevant. What mattered was optimization—keyword density, backlink volume, domain authority, publication cadence. Merit didn’t drive placement. Gaming the system did.

The internet is filled with competent noise. Professionally produced, carefully optimized content that existed not to inform but to rank. Articles that buried a one-paragraph answer inside fifteen paragraphs of filler because longer content ranked better. Entire editorial operations were built not around what was worth saying but around what had search volume. And the editors running those operations knew exactly what they were doing. They chose the traffic. They chose the rankings. They made the calculation that volume mattered more than value, and they were right—right up until the system they optimized for decided it didn’t need them anymore.

And the noise was homogeneous. When everyone optimizes for the same keywords, everyone writes the same story. A major product launches, and within hours, a dozen outlets publish functionally identical analyses hitting the same talking points in the same order. The system doesn’t reward differentiation. It rewards speed and volume on topics with proven demand. So that’s what it produces—the same article, fifteen times, in slightly different fonts.

When I was an active freelance columnist at ZDNet, I never chased the same story everyone else was racing to publish first. I’d wait days and publish a fully synthesized analysis that pursued different angles—because I understood my viewpoints were competing not just with writers at other outlets but with other columnists within ZDNet itself. Differentiation was the only viable strategy if you wanted your work to matter rather than simply exist. But the system didn’t select for that approach. It selected for being first, being optimized, and being indistinguishable from the next result.

People didn’t just stop reading because the interface made it optional. They stopped because two decades of SEO incentives made reading feel like a waste of their time.
Gen Z and Alpha never had to develop the instinct to browse, compare, and synthesize—the system does that for them. Reading is the fallback. Meanwhile, the people who can read deeply have been burned too many times by optimized garbage to bother. Both groups have converged on the same behavior for different reasons. The result looks identical: nobody reads.

And right as that convergence happens, the interface changes again.

What’s happening with AI-powered search—Google’s AI Overviews, Perplexity, ChatGPT answering questions directly—isn’t a betrayal. It’s an evolution that was always coming. The search interface was never going to stay frozen in 2005 forever, just because an entire content industry built its revenue model on top of it. The fact that AI can now synthesize an answer from multiple sources and deliver it without requiring a click isn’t a flaw. It’s the obvious next step. The flaw was an entire industry betting its future on being the middleman between a question and an answer, and then acting surprised when the middleman got cut out.

The traditional search engine outlived its usefulness—and so did the content ecosystem that grew up around gaming it.

And this isn’t theoretical. A Growtika analysis of ten major tech publications found their combined US organic search traffic fell 58%—from 112 million monthly visits to 47 million—in under two years. Digital Trends lost 97%. ZDNet, where I spent almost seventeen years as a freelance columnist, dropped 90%. The Verge fell 85%. HowToGeek, a site built almost entirely on step-by-step guides like “how to take a screenshot on Windows,” lost 85%—because those are precisely the queries AI Overviews now answer without requiring a click.

The pattern extends beyond tech. NerdWallet, a publicly traded company whose business depends on converting search visitors into financial product referrals, lost 73%. Healthline lost 55 million monthly visits—more traffic shed by a single health site than the entire ten-publication tech media sample currently receives combined. The steepest declines started mid-2025, coinciding with the expansion of AI Overviews. If your entire value proposition was being the page between the question and the answer, you were already dead. You just hadn’t checked the analytics yet.

The next step is already here. It’s just not evenly distributed yet.

Agentic systems—software that doesn’t just retrieve information but interprets, synthesizes, and acts on it—are the logical conclusion of everything above. Instead of searching, you ask. Instead of reading, the system interprets. Instead of comparing sources, the system does it and gives you the result. Content isn’t something you consume. It’s something the system uses as raw material.

Projects like OpenClaw are early signals of this shift. They’re frameworks for turning domain knowledge into something executable—systems that can take a question, apply structured reasoning, and produce an outcome without requiring a human to read anything along the way. A 2,000-word article, in that world, isn’t the product. It’s the ore. The value is in whatever gets extracted from it and acted upon.

That’s a fundamentally different relationship between knowledge and interface. And it breaks the economics that paid for the content in the first place.

The entire infrastructure of digital media was built on a simple transaction: a human visits a page, sees an ad, and the publisher gets paid. When the reader is a system instead of a person, that transaction disappears. An agent that synthesizes your article into an answer doesn’t load your ads, register a pageview, or convert on your affiliate links. It extracts the value and delivers it somewhere else.

What’s left is either gating everything behind subscriptions—which only works for publications with brand loyalty, not the vast middle of the content ecosystem that never had subscribers, only search rankings—or negotiating licensing deals with agent makers and aggregators. That market barely exists, and the few deals struck so far are opaque and tilted toward the platforms.

The technology for replacing the reader is here. The business model for compensating the writer is not. The content industry had twenty years to build something more durable than ad-supported search dependency. It didn’t. It kept taking from the tree—the apples, the branches, the trunk—and never planted anything of its own.

Now there’s nothing left but the stump.
That doesn’t mean longform disappears entirely. It means it moves out of the critical path. The audience becomes self-selecting, not captive. And the economics of writing for a self-selecting audience are very different from writing for one that had no alternative.

The people adapting have already figured this out. They’re not just writing about what they know. They’re building systems that do something with it—encoding expertise into tools, workflows, and frameworks that operate whether or not anyone reads a word.

Everyone else is still producing content for an interface that’s quietly being retired. And if that describes you, the uncomfortable question isn’t whether you see it coming. It’s what you’re going to do about it now that you can’t pretend you don’t.

People didn’t stop reading. We just made it unnecessary.

The tree gave everything it had. And now the boy is sitting on the stump, wondering why nobody’s reading his content anymore.