AI Overviews drained publisher traffic. Newsrooms are scrambling for revenue replacements that stick.
Two years into Google’s expansion of AI-generated answers across search, news publishers say referral traffic has fallen further than anything they prepared for. Licensing deals, paid-crawl marketplaces, and lawsuits are now competing to fill the gap.
When Google began surfacing AI Overviews above its search results in May 2024, the company framed the feature as an enhancement: faster answers for readers, with links beneath the summary to the publishers whose work the system relied on. Two years on — after an aggressive expansion in 2025 and the launch of a fully conversational “AI Mode” earlier this year — the same publishers are reporting a very different outcome.
Referral traffic from Google search is down sharply at most newsrooms BookerPost spoke with for this piece. Independent publishers in particular describe the past 18 months as the worst search-driven traffic decline of their existence.
What the data shows
Public studies tell a consistent story. A Pew Research Center analysis released last year found that when an AI Overview appears on a search results page, users click through to source links roughly 8 percent of the time — about half the rate of a traditional results page. Similarweb’s running tracker on news traffic shows global referrals from Google to news domains down by double-digit percentages year-on-year, with the steepest losses concentrated in how-to, health, and explainer content — the same categories AI Overviews most aggressively summarize.
Independent publishers describe even sharper falls. Several mid-size US news sites told BookerPost on background that organic search now contributes 30 to 50 percent less traffic than it did in early 2024, with no equivalent recovery from other channels.
What publishers are saying publicly
The News/Media Alliance, the US trade body representing more than 2,000 newspapers and magazines, has spent the past two years lobbying Congress and the Federal Trade Commission with a single message: AI products are cannibalizing the publishers they depend on for training data and live citations. Its chief executive, Danielle Coffey, has called the dynamic an existential threat to local journalism in particular, where margins are thinnest.
Across the Atlantic, the UK’s News Media Association has filed evidence with the Competition and Markets Authority. Penske Media, Condé Nast, and the BBC have all, in different forums, made the same basic argument: that AI Overviews and competing chat products extract value from journalism without compensating the journalists who produce it.
A handful of publishers have gone further. The New York Times’s 2023 lawsuit against OpenAI and Microsoft is now in discovery and remains the highest-profile test of whether large-scale ingestion of news archives constitutes infringement. Chegg’s 2025 antitrust suit against Google over AI Overviews, the first of its kind from a content publisher, is being closely watched by media lawyers as a possible template.
The licensing lane
For the largest publishers, the practical answer so far has been to sign with the platforms instead of fighting them. OpenAI has now inked content-and-product deals with News Corp, Axel Springer, Vox Media, The Atlantic, the Financial Times, the Associated Press, Le Monde, Condé Nast, Time, Hearst, and Dotdash Meredith, among others. Reported values range from a few million dollars to as much as a quarter of a billion over multi-year terms, though most agreements remain confidential.
Google, which previously kept publishers at arm’s length on AI deals, has also begun cutting cheques — most visibly through its Reddit content arrangement and a string of smaller deals with regional publishers. Perplexity, the AI-search start-up, launched a publisher revenue-share program in 2024 and has expanded it to dozens of partners.
Pay-per-crawl and the new gatekeepers
For the rest of the industry, the most-talked-about development is a layer of new infrastructure designed to make AI crawling itself a commercial transaction. Cloudflare, which fronts a substantial share of the open web, last year launched a “Pay Per Crawl” framework that lets site owners charge AI bots per request, with the network operator collecting and remitting payment. The product expanded out of beta in early 2026 and now claims tens of thousands of publishers, including most independent outlets BookerPost surveyed.
Specialist start-ups including TollBit and ScalePost have built similar marketplaces, layered with audit logs that let publishers see which AI companies have ingested which pages. Several large AI labs have quietly begun routing crawls through these intermediaries to head off litigation risk.
The simplest defensive tool remains the oldest. Updates to robots.txt blocking the major AI crawlers — GPTBot, Google-Extended, ClaudeBot, PerplexityBot — are now standard at most major newsrooms. The catch, publishers concede, is that blocking the crawler does not block the citation: a model that has already trained on a snapshot of the open web continues to summarize that snapshot until the next training run.
Owning the audience
The strategic conversation inside newsrooms has shifted, accordingly, toward channels publishers actually control. Subscription growth, which plateaued at most general-interest titles in 2024, has become the headline metric again. Newsletters — particularly those that are paid, or that drive subscription funnels — have moved from a marketing function to a core editorial product.
Several publishers told BookerPost that they have explicitly restructured their newsrooms around what one editor called “the post-Google reality”: writing for an audience that arrives in their inbox, not via a query. Live events, branded commerce, and proprietary data products — long the domain of B2B publishers like Bloomberg and the Financial Times — are now being prototyped at consumer outlets that previously dismissed them.
What’s left to fix
Legislative answers are moving more slowly. The Journalism Competition and Preservation Act has stalled in successive sessions of the US Congress. Australia’s news bargaining code, which once produced a windfall for publishers from Meta and Google, has weakened as both platforms have downgraded news content in their products. The EU’s Digital Services Act and AI Act provide a framework but no immediate revenue.
For the publishers BookerPost interviewed, the most honest answer to “what now” is that no single fix replaces what AI Overviews displaced. Licensing helps the largest. Pay-per-crawl helps the technically equipped. Subscription pivots help the brands strong enough to ask their readers to pay. Lawsuits, at best, help redraw the rules for the next cycle.
What none of them does, yet, is restore the assumption that built the modern web — that publishing on the open internet would, by default, reach a search-driven audience.
“We’re past that,” the digital publisher’s chief executive said. “The question is what gets built on the other side of it.”