Wonderwall
Publishers spent 15 years believing paywalls would save them. The strategy was real. The desperation behind it was too. Now AI is rendering the question moot
There’s a lyric from the Oasis song “Wonderwall” that’s been living rent-free in my head every time I read another paywall story: “I said maybe, you’re gonna be the one that saves me.” That’s the emotion publishers brought to paywalls when the model went mainstream in the early 2010s. Not strategy. Emotion. Hope masquerading as a business plan.
I don’t say that to be cruel. I say it because I was in the room. I watched publishers, smart operators with real businesses, latch onto the paywall narrative the way drowning people grab floating debris. The economics of digital media were already brutal. Print was collapsing. CPMs were a fraction of what the industry needed. And here, finally, was something that felt like agency. We’ll charge for content. We’ll own our audience. We’ll escape the platform trap.
Fifteen years later, we have enough data to render a verdict. And the verdict is complicated, more complicated than the paywall evangelists or skeptics will admit. A handful of publishers built real businesses behind paywalls. Most didn’t. And now, just as the model was finally maturing into something defensible, agentic AI has arrived to stress-test the entire premise.
This is the story of what actually happened, why it happened, and what publishers who are still standing need to understand about what comes next.
Act I: The Promise (2010–2018)
Salvation Narrative
The Wall Street Journal had been running a hard paywall since 1996, a fact that sounds prescient in retrospect but was largely irrelevant to the broader industry at the time. The WSJ served a professional audience whose employers often footed the bill. It was a B2B subscription product wearing a B2C costume. Everyone acknowledged it worked. Everyone assumed it wouldn’t work for general-interest news.
Then the print cliff accelerated faster than anyone anticipated. Between 2008 and 2012, newspaper print ad revenue dropped by roughly half. Digital advertising was growing, but it was growing for Google and Facebook, not publishers. The CPM math was unforgiving: a publisher might earn $25–$40 per thousand impressions in print and $1–$3 online. No volume of page views could close that gap.
I watched this from the inside. When I was publisher of InformationWeek, we had 400,000 controlled-circulation subscribers. These readers received the magazine free because they had the right title, the right company size, the right buying authority. That verified, qualified readership built a rate base: the audited circulation number that determined what we could charge advertisers. The currency from the readers was their time, engagement, and loyalty, which was the entire benefit to advertisers. At its peak, InformationWeek generated over $175 million in annual advertising revenue on that model. The paywall era inverted that logic entirely. It assumed the reader's currency should be cash. For most of publishing history, their attention was worth far more.
The New York Times launched its metered paywall in March 2011, and the industry held its breath. The conventional wisdom was that readers would simply navigate around it, that free content was too available online, that the paywall was a noble but doomed gesture. Instead, the Times started signing up digital subscribers. Not immediately transformative numbers, but real ones. Proof of concept.
Between 2017 and 2020, the number of news outlets implementing paywalls nearly doubled each year. The trade press treated it as a rescue mission.
By the mid-2010s, the paywall had become the dominant narrative of publisher survival. Industry conferences devoted entire tracks to paywall strategy. Consulting firms built practices around it. Publishers who hadn’t implemented one felt pressure to explain why. The enthusiasm was genuine, and in hindsight, genuinely disproportionate to the results most publishers would actually achieve.
Act II: The Scorecard (2018–2023)
Who Won, Who Lost, and Why
Let me give you the honest operator’s scorecard, because the industry press tends to celebrate the wins and bury the losses.
The New York Times won. Decisively. By February 2022, three years ahead of schedule, it hit 10 million digital subscribers and reset its target to 15 million by 2027. But the Times didn’t win because it built a better paywall. It won because it built a better product. Wordle. The Athletic. NYT Cooking. NYT Games. The paywall became a bundle gateway, not a content gate. Subscribers were paying for a suite of daily habits, many of which had nothing to do with news.
The Wall Street Journal won. For the same structural reason I described earlier; it had always been a professional subscription product, and first-party subscriber data made its advertising inventory genuinely more valuable to marketers. WSJ reportedly sells 90% of its inventory direct, and advertisers using its first-party data renew at significantly higher rates. The paywall isn’t a barrier; it’s a data-collection mechanism that funds a premium ad business.
The Financial Times won. Same playbook: a professional audience with employer-funded subscriptions and irreplaceable market-specific content.
Everyone else? Mixed at best. The economics became clear pretty quickly: paywalls work when they protect something a specific reader cannot get anywhere else. General-interest news is available everywhere. Local news is undervalued by local readers until it disappears. Niche B2B content; if it’s genuinely proprietary, or delivered with enough utility that it functions as a workflow tool rather than a reading experience, can command real subscription prices. Bloomberg didn’t build a $6 billion terminal business by having better journalism. It built it by embedding that journalism into a platform traders couldn’t do their jobs without. Some B2B publishers followed a version of that playbook: combining content with data, analytics, or workflow integration until the product became operationally indispensable to its audience. When a reader can’t do their job without you, price sensitivity largely disappears. But that bar is high, and most publisher content — if we’re being honest — never cleared it.
Most households will pay for one or two news subscriptions. That budget overwhelmingly goes to the same three or four scaled national brands. Everyone else is fighting over the remainder.
The Sun tried a hard paywall and abandoned it. The Toronto Star did the same. Countless regional newspapers experimented, retreated, and experimented again. Local news operators found that the readers most willing to pay were often those most engaged. The very people who would have supported the paper anyway. The marginal subscriber who needed to be convinced often just didn’t show up.
Meanwhile, a more structural problem was hardening. Publishers had signed up most of the readers who were ever going to subscribe. Growth was slowing not because the paywall model was wrong, but because the addressable market was smaller than assumed. As one Reuters Institute report bluntly noted, publishers had already captured many of those prepared to pay, and in a tight economic climate it had been hard to persuade others to do the same. Subscription growth was plateauing industry-wide by 2022–2023.
Act III: The Innovation Layer (2022–2025)
Smarter Walls, Shrinking Returns
To their credit, publishers didn’t give up. They got more sophisticated. The blunt instrument of the early paywall — a fixed meter, same for every reader — gave way to dynamic, AI-driven systems that personalize the conversion moment. Instead of offering every reader the same number of free articles, smart meters identify which readers are close to converting and apply pressure accordingly; they identify casual browsers and leave them in the open funnel to generate ad impressions.
The results have been real. Business Insider reported a 75% increase in subscriptions after implementing a dynamic paywall system. The Philadelphia Inquirer achieved a 35% lift in subscriber growth using a similar approach. By late 2024, 38% of news publishers were already using or planning to transition to dynamic paywall models. The internal debate between the subscriber revenue team (lock it down) and the advertising team (keep it open) had found a technology-mediated resolution: optimize for both simultaneously using behavioral data.
This was genuinely encouraging progress. Publishers were finally applying the kind of data discipline to subscription conversion that e-commerce had been using for years. It felt like the industry was growing up.
And then the floor started to move.
Act IV: The Existential Threat (2025 and Now)
The Wall That AI Walks Right Through
The paywall was designed to stop humans. It was never designed to stop agents.
In October 2025, researchers at the Columbia Journalism Review documented something that should have sent shockwaves through every publisher’s boardroom: AI browsers like OpenAI’s Atlas and Perplexity’s Comet were able to retrieve and summarize a 9,000-word subscriber-exclusive article from MIT Technology Review without triggering any of the barriers designed to block automated access. These systems aren’t scrapers in the traditional sense. They behave like humans. They load the full page, render JavaScript, pass bot-detection systems, because to the website’s infrastructure, they are indistinguishable from a person using Chrome.
The scale is staggering. AI scraping attacks on streaming and media properties jumped 56% year over year. From Q2 through Q4 2025, the rate of AI content scraping grew at an average of 24.4% per quarter. In March 2025 alone, 26 million scraping attempts ignored standard robots.txt directives. The traditional defenses weren’t built for this.
The paywall was designed to stop humans. It was never designed to stop agents. And the agents have arrived.
But the scraping problem, as serious as it is, may actually be the second-order threat. The first-order threat is more subtle and more devastating: if AI systems can synthesize answers from publisher content without sending users to publisher pages, readers never hit the paywall at all. They never become subscribers. The loss is invisible.
Zero-click search rates: queries that get answered without any click to a publisher’s page, rose from 56% in May 2024 to 69% by May 2025. DMG Media reported an 89% drop in click-through rates in September 2025, attributing it directly to AI Overviews in Google search results. The 500 most-visited publishers saw an average traffic decline of 27% year-over-year. Industry analysis estimates AI-powered search summaries reduce publisher traffic by 20% to 60% on average, with niche publications experiencing losses approaching 90%. In an industry where reach equals revenue, this is devastating.
The paywall assumes a reader arrives at your door. Increasingly, they don’t. They get what they needed from the AI and never make the trip.
The Framework: What Actually Works Now
Post-Paywall Revenue Architecture
The lesson isn’t that paywalls failed. It’s that paywalls were always a distribution fix applied to a value problem. Publishers put a gate on content that, in many cases, hadn’t earned the gate. The ones who built sustainable subscription businesses did so because they built something genuinely irreplaceable, then used the paywall to capture value from it.
That logic doesn’t change in the AI era. It intensifies. Here’s what the evidence suggests actually works:
Proprietary intelligence over commodity content. If an AI can synthesize a reasonable version of your article from publicly available information, your content has a commodity problem, not a paywall problem. The publishers who will survive the AI transition are those with original reporting, exclusive access, proprietary data, or a unique analytical perspective that cannot be reconstructed from digital breadcrumbs. This is what the WSJ has. It’s what the FT has. It’s what most publishers don’t.
Memberships over subscriptions. The semantics matter. A subscription is a transaction: pay for access. A membership is a relationship: belong to something. The publishers building durable reader revenue are increasingly framing their asks around community, identity, and shared purpose. This is not AI-proof, but it is significantly more human. People don’t cancel memberships the way they cancel subscriptions.
Experiences over content. Live events, conferences, workshops, dinners — anything that requires human presence cannot be scraped, summarized, or synthesized by an AI agent. This is where B2B media has always had an advantage: the event, the roundtable, the peer-to-peer connection, is genuinely irreplaceable. Publishers who haven’t built an events business need to start building one.
First-party data as a B2B product. The WSJ model: using subscriber data to make advertising dramatically more valuable, is replicable in principle by any publisher with a real subscriber base and the discipline to develop it. And for B2B publishers who took the utility path, embedding content into workflows, combining editorial with data and analytics, the first-party data advantage is even greater. A reader who uses your product to do their job generates behavioral signals that dwarf anything a casual visitor produces. The challenge is that most publishers let this data sit underutilized, treated as a byproduct rather than a product. In an environment where third-party cookies are dead and AI is eroding traffic-based ad models, the publisher who owns rich, consented, first-party audience data — especially data generated through active, functional use — has a genuine and defensible asset.
AI licensing deals. The New York Times is suing OpenAI; a case that has escalated into one of the most consequential AI copyright battles in history, now centered on evidence that AI models memorize and reproduce copyrighted content verbatim. The FT signed a licensing deal with OpenAI. Both strategies may prove correct for their respective situations. But the larger point stands: the question every publisher should be asking is not whether to resist AI, but how to monetize their content's role in the AI ecosystem.
The Song Ends the Same Way
The Oasis lyric has one more line worth sitting with: “After all, you’re my wonderwall.” There’s something heartbreaking in it. The conviction that this thing, whatever it is, is the answer. Publishers felt that about paywalls. The conviction was sincere. The timing was real. The economics were genuinely dire enough to justify almost any hopeful narrative.
But a paywall is not a content strategy. It is not a value proposition. It is not a relationship with a reader. It is a gate. And a gate only has value if what’s behind it is worth the price of entry, and if readers can’t get what’s behind it somewhere else for free.
For fifteen years, the publishing industry argued about the gate. The conversation we needed to have was always about what was behind it.
The AI era is forcing that conversation at last. Not because publishers wanted to have it, but because the old evasions have run out. The scrapers are inside the walls. The readers never arrive. The CPMs are gone.
What’s left is the only thing that was ever going to matter: content so good, so specific, so irreplaceable, that a reader would genuinely feel its absence.
Build that. Then put a wall around it.
The views expressed in Uphoff on Media are entirely my own. They don’t represent the opinions of any company I’ve led, any board I’ve sat on, or any investor who’s had the pleasure of debating strategy with me over the years. If something I write here sounds brilliant, I’ll take full credit. If it turns out to be wrong, I was clearly misquoted by myself.
Wow fascinating story! And great title. My first thought before I clicked was “like Oasis?” 😁