In March 2016, Emily Bell wrote what turned out to be a call to action. “Facebook is eating the world,” she declared in the Columbia Journalism Review, and she meant it structurally, not metaphorically. News publishers had ceded control over distribution to platforms.
The four horsemen — Google, Facebook, Apple, Amazon — were at war over whose infrastructure would win, and journalism was caught in the middle, dependent on systems it did not own and could not influence.
Also on Mediabistro
Nearly a decade later, Bell’s diagnosis looks prescient — but so does the industry’s response. Each wave of platform disruption has pushed publishers to build more direct relationships with their readers. The current wave, driven by AI, may be the most significant yet. And it is producing the most fundamental rethinking of the news business model in a generation.
Phase One: Facebook and the Referral Era
For most of the 2010s, Facebook referral traffic was a primary growth engine for digital publishing. Outlets built editorial strategies, audience teams, and revenue models around the assumption that social sharing would keep readers flowing to their websites. Then, gradually and then all at once, Meta decided to move on from news.
The shift accelerated in 2023. CNBC documented the scale of the change in January 2024: the top 100 global news publishers saw Facebook referral traffic drop sharply year over year, continuing a decline that had been building for years.
Mother Jones reported a 99% drop from its Facebook referral peak. Meta’s rationale was direct: “We know that people don’t come to Facebook for news and political content” — and the company deprecated its News Tab in the US and multiple European markets.
What made the contraction so severe was how unevenly it fell. National publishers with established subscription bases — The Atlantic, The New York Times — absorbed the shock better than local and regional outlets that had never built an alternative revenue floor. The warning signs had been visible for years, but the economic logic of chasing platform scale was difficult to resist when the traffic was still flowing.
The fundamental lesson — that distribution built on someone else’s platform is always conditional — had to be relearned at painful cost. Publishers who had diversified into newsletters, direct subscriptions, and owned audiences were better positioned than those who hadn’t.
Phase Two: Search Changes the Deal
Search had always felt like the more stable relationship. Unlike social, where virality was unpredictable, Google Search was transactional: someone types a query, Google returns links, and readers arrive. That model sustained publishing economics for more than two decades.
It is worth noting that the zero-click problem predates AI. Google’s featured snippets, introduced in 2014, began appearing on the results page to answer simple queries. Knowledge panels, shopping carousels, and local packs all eroded click-through rates over time. AI Overviews are not a new phenomenon so much as an acceleration of a trend that was already underway — which is precisely what makes them so consequential.
Reuters Institute’s January 2026 Trends and Predictions report, surveying 280 senior newsroom executives across 51 countries, found that Google search traffic to publishers fell by roughly a third globally in the prior year. Traffic from Google Discover was down 21%. News organizations now forecast a 40% decline in search referrals over the next three years.
The mechanism is familiar: AI synthesizes answers from publisher content at the top of the results page, resolving many queries before a click is needed. Publishers supply the reporting. Google surfaces the answer.
PPC Land’s December 2025 analysis found that 51% of the Google Discover feed in the US now consists of AI Summaries, with the remainder increasingly shifting toward YouTube. Google’s Q2 2025 earnings reflected the same dynamic: Network advertising (which includes publisher partners) declined 1%, while YouTube advertising grew 13% to $9.8 billion.
Jim Albrecht, who spent six years as Senior Director of News Ecosystem Products at Google, argued in a February 2024 Washington Post opinion piece that generative AI represents something categorically different from earlier disruptions. Radio, television, and the internet each changed how news was distributed. AI changes whether a news visit happens at all.
The implication for publishers is not despair but urgency: the value has to live somewhere other than the pageview.
Some publishers have responded not through litigation or advocacy alone, but by cutting deals directly. The Associated Press, Axel Springer, and News Corp have each signed licensing agreements with major AI developers — trading structured content access for a revenue stream that doesn’t depend on clicks.
The terms of these deals remain largely private, which makes it difficult to assess whether they represent a meaningful alternative or a modest concession. But they establish a precedent: AI companies have acknowledged, at least commercially, that publisher content has value beyond what the pageview economy has ever priced it at.
Phase Three: Agentic Browsers and the Next Frontier
The next shift is already visible on the horizon. Digiday’s December 2025 report on agentic browsers describes tools like Perplexity’s Comet, OpenAI’s Atlas, and Google’s Gemini-in-Chrome features that are designed to handle browsing tasks on behalf of users. Rather than a reader navigating to a news site, an AI agent retrieves, reads, and summarizes the content, returning an answer without a visit occurring.
Publishers are paying close attention. The concern is not simply traffic loss — it is that the advertising, subscription, and relationship infrastructure that supports journalism is built around the visit. No visit means no ad impression, no subscription prompt, no chance to convert a reader into a supporter.
The Digiday report notes that publishers want platforms to clearly distinguish human traffic from agentic traffic so that measurement and monetization remain meaningful.
The subscription model faces a particular tension here that has not yet been fully addressed. Agentic browsers do not subscribe. If a user’s AI assistant can pull the text of a paywalled article — through cached versions, through previews, through inference from available fragments — the subscription wall that sustains direct-relationship publishing is effectively bypassed without anyone making a deliberate choice to circumvent it.
This is the core challenge of what some analysts are calling the relationship economy: the relationship itself has to be desirable enough that readers seek it out directly, rather than delegating it to an agent.
The JournalismAI Festival 2025 in London framed the shift precisely: news media is “becoming part of AI systems.” That is both a challenge and a commercial opportunity. Publishers whose content is valuable enough to be ingested into AI systems are in a position to negotiate for compensation — if they act collectively.
What Publishers Have Tried, and What Comes Next
The most concrete attempt to restructure the platform relationship came from Australia. As CJR documented in 2022, Australia’s News Media Bargaining Code extracted more than $200 million from tech companies in its first year, funding dozens of new journalism positions. The model required a concentrated media market and a determined competition regulator, but it demonstrated that collective action could shift the economics.
California’s attempt in 2024 produced a more modest result. Google agreed to $172 million over five years, administered through UC Berkeley, plus an AI accelerator program. Critics called it inadequate. But it established a precedent that AI companies have obligations to the publishers whose content trains and informs their systems.
Madhav Chinnappa, the former Google and BBC News executive, has proposed a more ambitious framework: a “NATO for News” collective licensing model in which publishers negotiate AI data access terms together rather than individually. He calls it “the least worst option” — pragmatic about the power imbalance, but clear that collective leverage is the only kind that works at scale.
What these negotiations have clarified, perhaps more than anything else, is that publishers are only beginning to understand what their content is actually worth in an AI context.
The pageview economy priced journalism by the click. The AI economy will price it by something closer to knowledge value — a metric that favors deep, authoritative, well-sourced reporting over volume-optimized content. That is a structural shift that changes who the winners are.
The Business Model Hiding in Plain Sight
Every platform disruption has pointed toward the same answer: direct audience relationships. Newsletters. Subscriptions. Membership. Events. Products where the reader and the publisher have a connection that no algorithm can interrupt.
The New York Times offers the clearest illustration of what this looks like at scale. It has reduced its reliance on advertising and search traffic while growing subscribers, partly through journalism and partly through products — Wordle, Connections, the Cooking app — that create daily habit loops independent of the news cycle.
The strategy is not replicable for most publishers, but the underlying principle is: build something people want to come back to directly, not something they arrive at accidentally through search.
The Reuters Institute 2025 Digital News Report found that 65% of people now consume social video news, up from 52% in 2020 — but it also found that overall news trust has held at 40% for three consecutive years, and that audiences still want journalism they can rely on. The appetite is there, but the delivery mechanism is what’s in flux.
Nieman Lab’s 2026 Predictions, drawing on 210 expert forecasts, found broad consensus on what sustainable news organizations look like: ones that treat editorial independence and community connection as the core product, with distribution as a secondary consideration. Publishers who have diversified revenue, built owned audiences, and stopped measuring success primarily in pageviews are the ones facing this transition from a position of strength.
The news business has effectively split in two. On the one hand, organizations that built direct reader relationships before the AI transition have revenue that doesn’t depend on intermediaries and are positioned to negotiate with AI platforms from a position of something approaching parity. On the other: outlets still reliant on platform traffic, facing the prospect of a 40% decline in search referrals with no structural alternative in place.
It is worth noting what this means for the people working in these organizations. Publishers investing in owned technology and direct audience infrastructure need people who understand both journalism and product thinking — audience development, data analysis, engagement strategy, and reader experience design.
The skills that are gaining value in newsrooms are not the traditional traffic-driving skills but those that build durable relationships. That shift is already visible in where media organizations are hiring.
The arc from Facebook to Google to agentic browsers is a single story about what happens when journalism’s value is delivered through someone else’s infrastructure. The publishers rewriting that story are the ones worth watching.
Mediabistro is a leading career platform for media, journalism, content, and marketing. We publish media news, career advice, and featured interviews with practitioners at the forefront of media. Employers can post jobs to reach mid and senior level talent.
Topics:
media-news
