The Death of the Internet

Somewhere right now someone is publishing a blog post they spent two days on. A repair guide. How to clear the exact fault on a particular car — the recurring error code the dealership wants $400 to plug a laptop in and read out. They took the photos themselves. They tested the fix on their own car. They wrote it plainly so a stranger wouldn’t have to hand over $400 to be told a $12 sensor and twenty minutes would have solved it.

When a driver searches for that fault, an answer engine will have read the page, kept the part that helps, and printed it at the top of the results in its own voice. The reader gets the fix. The person who wrote it gets nothing. No visit. No ad shown. No reason discovered to read anything else they’ve written. The work was useful. It just wasn’t useful to them.

That’s not a glitch. That’s the new arrangement. And it’s quietly killing the thing that made the internet worth having.

The web ran on a handshake nobody signed

For about twenty years the open web ran on a deal that no one ever wrote down.

You made the thing. Search engines sent you the readers. Ads on your page paid you enough to go and make the next thing. Round and round. A loop that funded an absurd amount of human effort — recipes, tutorials, reviews, forum answers, obituaries of obscure software, every “how do I” you’ve ever typed at 11pm.

Nobody loved every part of it. The ads were ugly. The SEO games were tiresome. But the loop held. Make, distribute, get paid, make again. That loop is what’s dying. Not the websites themselves — the incentive underneath them.

Pull the incentive out and the sites don’t vanish overnight. They just stop being replaced when they go stale. The decay is slow at first. Then it isn’t.

Blogging was the first time the little guy had a printing press

Go back to the start. The genuinely radical thing about the early web wasn’t the technology. It was the distribution.

Before it, if you knew something worth knowing, you had two options. Convince a publisher to print you, or keep it to yourself. There was a gate, and someone else held the key. Then blogging arrived — Blogger in 1999, WordPress a few years later — and the gate fell over.

Suddenly a nurse in Adelaide could write the definitive guide to a rare condition and a stranger in Glasgow could find it. A bloke who restored old motorbikes could document a rebuild step by step and become the reason a hundred other people kept their bikes on the road. None of these people asked permission. None of them went through a publisher. They just knew things, and now they had a press.

This is the part people forget when they get misty about the old internet. It wasn’t nostalgia for slow modems. It was that, for the first time, expertise didn’t need a gatekeeper’s blessing to reach the person who needed it. The little guy had a printing press. And for a while, the press paid for itself.

Ads were the deal that funded everything you read for free

The free web was never free. It was ad-funded.

When Google launched AdSense in 2003, it did something clever and slightly grubby at the same time. It let any website, no matter how small, slot advertising onto its pages and get paid per view and per click. The motorbike bloke didn’t need to sell ad space or chase sponsors. He pasted in a snippet of code and the machine matched ads to his readers and sent him a cheque.

It was a Faustian little bargain. You read the guide for free. In exchange your attention got sold to whoever wanted it. Most of us made that trade a thousand times a day without thinking about it.

Ugly as it was, it worked. It paid the writers. Not all of them well — but enough of them, enough to keep going. The recipe blog with the long rambling story before the ingredients? That story exists because the page needed enough room to carry the ads that paid for the recipe. You can hate the format and still understand the economics. The format was the economics.

Google was the toll road, and for years it let you through

None of it would have mattered without distribution. And distribution, overwhelmingly, meant Google.

Search became the toll road of the entire web. If you wanted readers, you needed to rank. So an entire craft grew up around being understood by the algorithm — SEO, half science and half superstition, an industry of people trying to guess what the machine wanted.

Resent it all you like, but for years the arrangement was genuinely symbiotic. Google needed the pages to exist, because a search engine with nothing to point at is worthless. The pages needed the clicks, because clicks were how they got paid. Two parties, each holding something the other couldn’t do without.

Google took its cut. It always took a cut. But it let the travellers through to the destination. The writer still got the visit. The loop still closed.

That’s the part that changed.

Then the toll road started keeping the travellers

It began with something that sounded helpful. Search results that answered your question right there on the page, so you didn’t have to click.

The weather. A sports score. A celebrity’s age. Then the “featured snippet” — a chunk lifted straight from someone’s page and displayed at the top, often answering the question completely. Convenient for you. Catastrophic for the person whose words were sitting in that box getting no visit for the privilege.

The people who track this stuff have a name for it: zero-click search. A search that ends without anyone going anywhere. By most credible measures it stopped being a fringe case years ago and became something close to half of all searches. The query happened. The answer was delivered. The toll road kept the traveller and the destination never saw them.

Then, in 2024, the real shift. AI Overviews — a generated summary stitched together from multiple sources, sitting above the actual results, answering in a confident synthetic voice. Not a snippet from one page now. A blended digest of many, owing nothing to any of them.

The toll road stopped being a road. It became the destination. Why drive to the town when the booth at the entrance hands you everything the town was going to give you?

Now the machine eats the meal and leaves no tip

This is the heart of it, so let me say it plainly.

An AI summary is built out of work that people did. It reads that car repair guide and a dozen others. It keeps what helps. It composes an answer. It shows that answer to the person who would, in the old loop, have clicked through and counted as a reader and triggered an ad and paid for the next guide.

The reader’s need gets met. The creator’s never enters the equation. The machine eats the meal and leaves no tip.

And the cruelty of it is the dependency runs one way and pretends it doesn’t. The model needs the guide to exist in order to have anything to summarise. It cannot tell you how to clear that fault unless someone, somewhere, first worked out how and wrote it down. But the act of summarising removes that someone’s reason to have bothered. It’s eating the seed corn and calling it a harvest.

Defenders say the model is just doing what a human researcher does — reading widely and synthesising. But a human researcher who reads your guide and passes it on at least leaves a trail back to you. They cite. They link. They send the next person your way. The answer engine, by design, is the trail’s end. It is built to be the last stop. Being the last stop is the entire product.

Kill the incentive and you kill the supply

Here’s what people miss when they wave this away. The complexity doesn’t disappear. It just moves.

The work of knowing how to clear that fault still has to be done by someone. The model can’t do it from first principles. It can only repackage what a human already worked out. So when you sever the payment from the creation, you don’t abolish the cost of creating. You just stop paying it. And the bill lands on the least powerful party in the system — the individual creator who used to scrape together a living, or at least a reason, from the loop.

Take the reason away and watch what happens to the supply.

Not immediately. The web today is still thick with two decades of accumulated knowledge, written back when the loop still paid. The models feed on that backlog and seem inexhaustible. But the backlog is finite. The car that throws a new fault next year, the software released next month, the regulation that changes in the spring — somebody has to sit down and work those out fresh, and then write them down, in public, for free.

Why would they? If the only certainty is that an answer engine will harvest the work and hand it to the world while you collect nothing — no visit, no ad, no audience, no reason discovered to come back — then the rational move is not to bother. Keep it in a paywalled course. Keep it in a private group. Keep it in your own head and charge $400 to come and plug the laptop in yourself.

That’s not creators being precious. That’s creators responding to incentives, the same way they always did. The loop made it worth publishing in the open. Break the loop and the open web stops being where the knowledge goes.

What we lose

So here’s where it lands, and it’s bigger than any one blogger’s traffic.

The open web was a commons. Not a charity and not a utopia — a commons, built on a grubby little ad-funded bargain that nonetheless produced something extraordinary. A vast, searchable, free record of what an enormous number of people had bothered to work out and write down. The nurse’s guide to the rare condition. The motorbike rebuild. The car fault and the $12 sensor. Knowledge that existed in public, findable by anyone, precisely because the loop paid the people who made it.

The answer engines are downstream of that commons. They are, in the most literal sense, drinking from a well they are also draining. They can only summarise what the loop already funded into existence. And the act of summarising is dismantling the loop that fills the well.

This is the quiet death. A slow withdrawal of the reason to contribute, until what’s left in public is the old stuff, the ad-farm sludge, and whatever the machines generate by recombining a past that nobody is bothering to extend.

The internet that mattered wasn’t the cables or the apps. It was the deal — that if you knew something and shared it openly, the system would carry it to whoever needed it and give you enough back to keep going. That deal is being cancelled. And the thing being built on top of its corpse is convinced it doesn’t need it.

It does.