for agents/llms.txtv0.6.20 · 29 Sep 2026

Home / Articles / The reader was always the product: a corrected history of how news got into this mess

The reader was always the product: a corrected history of how news got into this mess

By · 2026-09-29 · v0.6.20 · newspublishinghistoryadvertisingprovenancemicropaymentseconomicsarticle

Abstract: This started as a voice memo setting out my understanding of how the publishing and news industry got to where it is, in four eras, print, web, platforms and AI, and an instruction to check it and correct it. The research corrected it in four places, and the corrections are the article. The reader did not become the product when the web arrived; the reader has been the product since the penny press of 1833, and by 2005 advertising was 82% of American newspaper revenue. What the web took was not the business model but the monopoly underneath it, the local toll bridge that let a paper charge what it liked and fund reporting with margins of 20 to 30 per cent; classifieds alone fell from $19.6 billion to about $6 billion in nine years. The platforms then made the reader a measurable product and the publisher a tenant: Google and Meta took over half of American digital advertising by 2017, Facebook referrals fell 58% in six years, false news travelled 70% further than true, and newspaper newsrooms lost 57% of their staff. AI removed the traffic itself, and the industry's answer has been to go back to selling to readers, by subscription, so that circulation revenue now exceeds advertising for the first time in living memory. The road not taken was there from the start, a payment code reserved in the web's own protocol in 1997 and never used, and the evidence that people pay when paying is easy, from a million songs in a week in 2003 to five million paid newsletter subscriptions in 2025, is what the story vault work on this site is built on.

Four eras, and what changed in each. The model, selling readers to advertisers, did not change until the last one. The monopoly under it went in the second, the reader became measurable in the third, and the traffic went in the fourth. Every figure is sourced below.

This piece started as a voice memo. I set out my understanding of how the news and publishing industry got into the state it is in, era by era, and asked for it to be researched and corrected where it was wrong. It was wrong in four places, and each of the four turned out to matter for the argument I have been making on this site about what the industry should sell instead. So this article keeps the shape of the memo, four eras, and puts the record next to my reading in each one. Where I was right, the numbers are here. Where I was wrong, the correction is the point.

In short

Era one: print, and what I got wrong about it

What I thought. That the industry began on printed advertising and single-copy sales, that subscriptions were minor, that the customer was the reader, and that this model built powerful institutions with a business case for investigative journalism.

What the record says. The order is backwards, and the correction is the most important one in this article. Before 1833, American newspapers were subscription products: most cost six cents and "were distributed mostly through subscriptions," paid annually, to elite readers, and were controlled by political parties or merchants. The penny press, beginning with the New York Sun in September 1833, inverted that. It sold on the street for a cent, to anyone, and "the main revenue for the penny press was advertising while other newspapers relied heavily on high-priced subscriptions." The reader stopped being the customer in 1833. The reader became the audience that was sold to the customer, and the cover price became a way of proving the audience existed. Richard Serra said it of television in 1973, "You are the product of TV. You are delivered to the advertiser who is the customer," but television inherited the arrangement from the newspaper.

By the time the web arrived, the arrangement was total. In 2005, the best year American newspapers ever had, advertising brought in $49.4 billion and circulation $10.7 billion: 82 cents of every dollar came from the advertiser, and the reader who bought the paper was paying for the right to be advertised to. The historian Nicholas Lemann puts it precisely: newspapers "increasingly sold not just news to readers, but also readers to advertisers."

So where did the journalism come from? Not from the model. From the monopoly on top of it. Through the twentieth century, hundreds of afternoon papers failed or merged into the morning paper, until most American cities had one. Lemann: by the mid-1960s "the nontabloid, subscription-based morning daily newspaper had become a lucrative monopoly, generating profits of 20 or even 30 percent a year," and the papers "adopted a journalistic version of the ethic of 'corporate social responsibility'... which in their case meant an economically irrational overinvestment in reporting." Warren Buffett, who bought newspapers for exactly this reason, likened "owning a monopoly newspaper or dominant market newspaper to owning an unregulated toll bridge. You have relative freedom to increase rates when and as much as you want." Every local advertiser had to cross the bridge, because there was no other way to reach local households. The foreign bureaus, the investigations, the Watergate era, were paid for by the toll.

That is the corrected version of era one, and it changes the story. The industry did not lose a model that rewarded journalism. It never had one. It had a monopoly whose owners chose, for a few decades, to spend some of the toll on journalism. What the next era took was not the model. It was the bridge.

Era two: the web takes the bridge

What I thought. That the web was a shock to an industry that never treated technology as core; that it went online kicking and screaming; that this began the race to the bottom, and that the customer became the product.

What the record says. Right about the shock and the lag, wrong about the mechanism, and the product part had already happened. The web did not turn readers into a product. It built a second bridge next to the toll bridge, and then a third, and the toll went to zero.

The classified pages were the purest toll. Robert Seamans and Feng Zhu found that newspapers historically "received around 40 percent of its revenue from classified," that Craigslist alone cost them at least $5 billion between 2000 and 2007, and that "when Craigslist enters a market, the effect on a newspaper's classified ads is almost immediate." The Newspaper Association of America's own figures, reported by Poynter, have classifieds falling from $19.6 billion in 2000 to roughly $6 billion in 2009, "a stunning 70 percent" in a decade. Nobody had to out-report the newspaper. They had to let people list a sofa for free.

Display advertising went the same way for the same reason. The first banner ad ran on HotWired on 27 October 1994, and the first buyers paid for scarcity that did not exist: within a few years anyone could publish a page and sell the space beside it. What the newspaper had sold was not attention but exclusive attention, and the web has no exclusive.

On the lag, I was right, and so was almost everyone who has looked at it: the industry put its content online for free, in the ad model it knew, and discovered that the ad model without the bridge is a different business. But I would now put it less harshly. This was not a failure to understand technology. It was a failure to understand that the technology had been the moat. Every industry that owned its distribution, music, travel agents, classifieds, encyclopaedias, had the same decade. News was not special. It was just the one whose product we need to be true.

Era three: the platforms make the reader measurable

What I thought. That publishers came to depend on Google and then on social media for traffic, a Faustian deal with the companies that would kill them; that the platforms rewarded extreme content, so truth stopped selling; and that provenance was lost as journalists were cut.

What the record says. Right on all three, and the numbers are worse than the memo.

The reader had been the product since 1833, but the newspaper could only sell the reader in bulk, by circulation. DoubleClick's cookies in 1996 made the reader a product that could be followed from page to page; AdSense in June 2003 let any page sell that reader through Google; and the party that could see the reader everywhere took the market from the parties who could see the reader only on their own pages. By 2017 Google and Meta together took 54.7% of American digital advertising. The publisher's advertising business became a tenancy on somebody else's land, and the rent was traffic.

Then the landlord changed the terms. In January 2018 Facebook reweighted its feed towards friends and away from publishers. Chartbeat's data across 792 news sites, reported by Press Gazette in May 2024, shows Facebook referrals down 58% in six years, from 1.3 billion a month to 561 million, and down 50% in the final twelve; as a share of publishers' external traffic, Facebook fell from 30% to 7%. When Canada's Online News Act required platforms to pay for links in 2023, Meta blocked news in Canada rather than pay, and it has stayed blocked. Australia's bargaining code of 2021 produced deals, for the publishers large enough to be designated. A tenant does not negotiate with a landlord who can turn off the road.

On truth not selling, the record is exact. Soroush Vosoughi, Deb Roy and Sinan Aral studied 126,000 stories tweeted 4.5 million times between 2006 and 2017 and found that "false news stories are 70 percent more likely to be retweeted than true stories are," that it took true stories "about six times as long to reach 1,500 people," and that removing every bot from the data made no difference: "falsehood diffuses significantly farther, faster, deeper, and more broadly than the truth, in all categories of information." Their explanation is novelty, not malice, which is worse, because novelty is what a newsroom under pressure to be shared will manufacture. Facebook's own ranking formula, revealed in its internal documents in 2021, weighted every emoji reaction at five times a like from 2017; the anger reaction was one of them, and the company's own data scientists found in 2019 that posts drawing anger were disproportionately likely to contain misinformation and low-quality news. The memo's phrase was that a scandalous statement sells better than a factual one. The platforms measured it and priced it in.

And on provenance, the memo said that checking a source is expensive, that journalists were cut, and that the checking went with them. The staffing numbers are these. American newspaper newsrooms fell from roughly 71,000 jobs in 2008 to about 31,000 in 2020, a 57% loss. In Britain, at least 293 local newspapers have closed since 2005, and the three largest regional publishers employ about 3,000 journalists against 9,000 in 2007. Provenance is labour. Following a claim to its source, freezing what the source said, recording who said it and when, is hours of a person's time, and under both of the models the industry ran, it was the hours that did not move the click and so were the first to go. I said in the memo that most of that work was done in notebooks and spreadsheets, and I stand by it as an observation rather than a finding: I have not found a study that counts it, and the tooling to do it cheaply, the story as a graph with every claim tied to hashed evidence, did not exist until recently.

Era four: AI removes the traffic

What I thought. That AI aggregators now consume the content and give the answer, so the traffic that remained is going; that it is more expensive for the AI companies to process badly formatted sites than it need be, with no financial model to fix it; and that the whole thing is a highly inefficient loop.

What the record says. Right, and documented on this site over the past week, so briefly. Organic search traffic to publishers fell 33% in the year to November 2025, 38% in the United States, as the search engine began answering the question itself. The crawlers that replaced the traffic fetched, in Cloudflare's July 2025 data, 38,000 pages for every visitor Anthropic's sent back, 1,100 for OpenAI's, 195 for Perplexity's. One small publisher's week, posted this weekend, was 16,000 fetches and ten clicks. And the fetching is done in the most expensive way available, HTML parsed and tokenised on every visit, more than half of them for pages that have not changed, which is the argument of the token bill nobody is sending: the inefficiency the memo describes is real, measurable, and paid for twice, by the publisher in bandwidth and by the AI company in tokens, with nobody paid by the other.

One more number belongs in this era, and it is the one that ties the four together. In 2021, for the first time in the modern record, American newspapers' circulation revenue overtook advertising: $11.5 billion against $10.3 billion, and in 2022 $11.6 billion against $9.8 billion, advertising having fallen 80% from its peak. After a century and a half of selling the reader, the industry is selling to the reader again. But it is selling the subscription, unlimited access to everything, priced for the reader who forgets to cancel, which is the model the story vault article argues is the wrong product for the wrong buyer, and which the law is now taking apart on its own timetable. The industry did not go back to 1832. It went to rent.

The road not taken was in the protocol

What I thought. That in the early days of the internet there was a big idea that the creator should be paid, that somewhere along the way, through decisions nobody quite made, the business model became advertising and surveillance, and that the platforms took the value while the creators fought for peanuts.

What the record says. Right about the idea and right about the outcome, but the decisions were made, and they have names and dates.

The idea was not vague. It was in the plumbing. The HTTP/1.1 specification of 1997 reserved a status code, 402 Payment Required, which "was created to enable digital cash or (micro) payment systems and would indicate that requested content is not available until the client makes a payment." The current specification still describes it as reserved for future use, and no browser has ever implemented it. The web was built with a slot for the creator to be paid, and the slot was left empty for nearly thirty years.

It was left empty because the first attempts to fill it failed, and then because two well-argued essays persuaded everyone not to try again. David Chaum's DigiCash, the first serious digital cash, filed for bankruptcy in November 1998; Digital Equipment's Millicent, built for payments of a tenth of a cent, piloted with 7,000 customers and 24 merchants in 1997 and went no further. Nick Szabo's Micropayments and Mental Transaction Costs of 1999 and Clay Shirky's The Case Against Micropayments of 2000 made the same argument from two sides: the cost of deciding whether an article is worth two cents exceeds two cents, so people will always choose the flat fee or the free page. Both were right about a person clicking a link in 2000. Neither imagined a buyer with no mental transaction cost at all, which is what an agent with a wallet is, and which is why that argument no longer decides the question.

Meanwhile the other slot filled itself, one small decision at a time. The banner ad in 1994. The cookie that could follow a reader across sites in 1996. The pop-up ad, invented at Tripod in the late 1990s by Ethan Zuckerman, who apologised for it in 2014 and drew the conclusion this section rests on: "the fallen state of our Internet is a direct, if unintentional, consequence of choosing advertising as the default model to support online content and services." AdSense in 2003, which made that default available to every page on the web. None of these was the decision. Together they were.

And the evidence that the other road was passable was there the whole time. On 28 April 2003 Apple opened a store that sold songs for 99 cents, at a moment when the music industry's position was that its customers were thieves. Steve Jobs's launch statement was that "consumers don't want to be treated like criminals and artists don't want their valuable work stolen," and a million songs sold in the first week. Make the legal way the easy way and people take it. Twenty-two years later, Substack passed five million paid subscriptions, four months after passing four million, "real people rewarding the work of writers and creators with real money." That is people paying writers directly, at scale, in an industry that spent two decades insisting they never would. It is also, still, the subscription: rent on a writer rather than on a paper. The market for paying creators exists. What has not existed is a way to pay for the thing itself, the fact, the evidence, the confirmation, the projection, rather than for a season ticket to the person who produced it.

What the corrected history says to build

Where the money went, era by era. In print the advertiser paid the publisher and the publisher's monopoly margin paid for reporting. On the web and the platforms the advertiser paid the platform, and the publisher received traffic, which is not money and can be switched off. In the AI era the reader pays the answer engine and the publisher sells rent. The fourth panel is the proposal: charge for the cargo, not the road.

Put the four corrections in a row and they point somewhere.

The reader was always the product, so returning to "the reader as customer" is not a return; it is new. The journalism was paid for by a monopoly on distribution, not by the model, so no model that depends on distribution will pay for it again; distribution now belongs to search engines, platforms and answer engines, permanently. The platforms showed that when the reader is measured, novelty beats truth, so any model that pays for attention will pay for falsehood. Provenance was cut because it was labour that did not move the click, so it will only come back if it is what is being paid for. And the AI era has removed even the traffic, while the payment slot the web reserved in 1997 has, in the last eighteen months, finally acquired rails that work.

That is the case for the thing this site has been building towards: stop selling the reader, stop renting the reader, and sell what the reporting was made from, the story as a graph, every claim walked to frozen and hashed evidence, sold in pence and on demand, to the reader who wants one thing, the firm that must forward it, and the agent that needs to know whether it is true, with the money walking back to whoever made the fact. Investigative journalism becomes the most valuable thing a newsroom produces not because more people read it, but because it is the thing with the most evidence attached, and the evidence is the product.

We have never needed an accurate, provenance-first press more than now, and we have never consumed more of what it produces, in more forms, through more intermediaries. The consumption is not the problem. The problem is that for a hundred and ninety years the money has flowed to whoever owned the road, and the road has changed hands three times. The fix is to stop charging for the road and start charging for the cargo. That was the road not taken in 1997, and it is open now.

Sources

© 2026 Dinis Cruz. This article's own text is licensed under Creative Commons Attribution 4.0 International (CC BY 4.0). You're free to share and adapt it, as long as you give credit. Quoted material and linked sources keep their own licences.

← All articles