Editorial illustration for: Bots Now Outnumber Readers, and Publishers Must Choose Between Charging the Machines and Building What They Can’t Copy

Media

› Industry report

Bots Now Outnumber Readers, and Publishers Must Choose Between Charging the Machines and Building What They Can’t Copy

Media Trendlines — June 16, 2026

📰 Key Themes

  1. Automated crawlers passed human visitors as the majority of web traffic for the first time, turning the audience publishers optimize for from people into machines.
  2. The fight over how to respond hardened at WAN-IFRA’s Marseille congress, between publishers licensing their archives to AI companies and those betting on reporting a model cannot reproduce.
  3. Fresh figures put hard numbers on the referral collapse: rival AI engines send a fraction of the traffic a Google search once did, and the largest news sites have shed nearly half their visitors in four years.
  4. With acquisition traffic drying up, publishers are pouring money into events and membership, where People Inc. and Country Living are finding the growth their websites no longer deliver.
  5. A UK regulator handed publishers their first real lever over AI Overviews, ordering Google to let sites opt out of AI summaries without punishing their search ranking.
  6. The registration-wall playbook is spreading, with new data showing that a sliver of logged-in newsletter readers drives the bulk of a publisher’s subscriptions.

Jump to: 💡 Business Model Innovation · 📺 Big Media Moves · 📎 Also Noted · 🧭 Takeaways

💡 Business Model Innovation

The Audience Is Now Mostly Machines, and That Turns Content Into a Data Product

Source: The Media Copilot, Pete Pachal.

Cloudflare CEO Matthew Prince says bot traffic has crossed the threshold that the industry has been bracing for: 57.4% of requests on the web are now automated, against 42.6% from humans — a crossover that arrived eighteen months ahead of his own forecast, with agent traffic growing eight times faster than human activity. The economics behind that number are brutal. TollBit has logged scrape-to-referral ratios of 179-to-1 for OpenAI, 369-to-1 for Perplexity, and 8,692-to-1 for Anthropic. Digital Trends counted 4.1 million bot scrapes against 4,200 human referrals in a single week.

For twenty years the bargain with search was that Google indexed a site and sent visitors back through a headline and a link. AI search broke it: answer engines ingest the work, summarize it, and resolve the query without the click. Pachal’s argument is that what survives is the citation — a named source inside the answer — and that this makes the relationship look less like distribution and more like syndication. The same week, the UK’s Competition and Markets Authority ordered Google to let publishers opt out of AI Overviews without tanking their search ranking, the first time a regulator has split AI crawling from search indexing.

The reframing is the important part. If the readership that matters is a fleet of crawlers, then a page stops being a destination and becomes inventory — something to be priced, licensed, and metered rather than merely published and optimized for clicks. That is a harder business than the one publishers spent two decades learning, because it strips away the comforting fiction that traffic was ever the product. It wasn’t. The information was. The machines just made that impossible to keep ignoring.


At Marseille, the Industry Split Into Sellers and Builders

Source: The Audiencers, Marion Wyss. ⚠️ Paywalled — summary based on the available preview.

WAN-IFRA’s World News Media Congress opened with AG Sulzberger of The New York Times calling generative AI’s training data the industry’s “original sin” — talent, compute, and energy all get paid for, he argued, while the fourth input, the journalism, is simply taken. He brought the numbers: the Times publishes roughly 500,000 original works a year at a cost north of $2 billion, the Google click is ten times harder than a decade ago, rival AI models send 96% less referral traffic than a Google search, and the largest news sites tracked by comScore have lost more than 45% of their traffic in four years.

The two responses sat on the same stage. Louis Dreyfus of Groupe Le Monde defended licensing: ChatGPT, he said, converts 17 times better than Facebook and 173 times better than Google Discover, with no cannibalization, and Le Monde routes 25% of the deal revenue to its newsroom. David Buttle of the SPUR coalition delivered the counter-warning — “most of today’s AI deals are defensive agreements, not product partnerships. You sign, and you still get scraped.” The builders’ camp had its own metric, from VG’s Gard Steiro: “the share of our content that AI cannot copy,” now tracked every morning alongside traffic, conversion, and churn.

Underneath the philosophy was plumbing. Mizal AI’s Florent Daudens laid out a three-layer stack — rights, access, and payment — meant to let publishers charge agents directly, with per-query micro-payments that failed for humans because of mental friction but should work for machines because there is none. The split is real and it is strategic, not temperamental: selling to the machines books revenue now and concedes control of the product, while building what they can’t copy forfeits the near-term check to defend a long-term moat. Most publishers will end up doing both, badly, before they learn to do either well.

The WordPress angle: A rights-access-payment stack only works if a publisher’s platform can describe its content to a machine — structured metadata, provenance, licensing terms, and a payment hook exposed at the layer agents actually read. Sites whose CMS still treats an article as a blob of rendered HTML will struggle to meter and license at the granularity these proposals assume; the ones already running structured, API-first or headless setups are closer to charging the machines than they realize. Charging the agent is, in the end, a content-modeling problem before it is a commercial one.


News Is Becoming a Feature, So It Has to Become a Community

Source: The Rebooting, Brian Morrissey.

Morrissey reads the new Reuters Institute Digital News Report as a set of warning lights, not a debate about trust. For the first time, social and video networks (54%) outrank publishers’ own websites and apps (51%) as a source of news, and owned digital properties have lost twelve points of reach since 2020. Paying for online news has flatlined at 17% across twenty markets. AI chatbot news use climbed from 7% to 10%, and only 42% of those users say they regularly click through to a source. Speaking on Morrissey’s podcast, the report’s Jim Egan dispensed with the hope that young audiences will age into the habit: “older people are becoming more like younger audiences rather than the other way around.”

His conclusion is that news is sliding from a product to a feature — a personalized update folded into an agent’s daily briefing, “good enough” for the majority who were never news junkies. The defensible move, he argues, is to stop chasing the mass mandate alone and build genuine community models for the diehards: bundles, in-house creators, the thousand-true-fans logic that news organizations have always been too distant to attempt.

This is the constructive half of the same story the bot numbers tell. If the open web no longer delivers an audience and AI no longer delivers the click, the only traffic a publisher fully owns is the relationship it builds directly. Community isn’t a soft consolation prize here; it is the one asset a model can’t scrape and a platform can’t disintermediate.


Country Living Is Growing Because It Sells Belonging, Not Pageviews

Source: A Media Operator, Shannon Thaler Cherry.

A nearly 50-year-old Hearst shelter brand grew its All Access membership 57% year over year, and its audience is getting younger as it does — Gen Z and Millennials now make up 51% of readers, up from 36% a year ago, against a median age of 52. The growth is built on the unfashionable stuff: a $35 membership bundling print, a members-only antiquing newsletter, and live events, from a sold-out Dollywood weekend at $1,400 a ticket to the revived Country Living Fair, where the VIP tier stands for “Very Important Picker.”

Editor-in-chief Rachel Hardage Barrett frames the antiquing revival as “the opposite of SEO” — surprise and delight rather than scale and ubiquity — and leans on events explicitly to future-proof against AI. It is a tidy demonstration that the post-traffic playbook isn’t only for hard news. A lifestyle brand with a passionate niche can convert that passion into recurring revenue precisely because none of it depends on ranking, referral, or a crawler’s goodwill.


Broadsheet’s Registration Wall Shows Where Subscriptions Actually Come From

Source: The Audiencers, Madeleine White.

Australia’s Broadsheet put a registration wall in front of its paywall and shared the data. The cohort of registered, newsletter-engaged readers is just 0.7% of its total audience — and drives 43% of all digital subscription conversions, an over-index of 61 times. Those same readers hit the paywall 3.5 times each, against 0.1 for an anonymous visitor, and direct offer emails unlocked by registration account for another 18% of conversions.

The lesson is that conversion is a frequency game, and the only way to win it is to know who the reader is. Anonymous scale has never been worth less; a logged-in, emailable relationship has never been worth more. Registration is how a publisher manufactures the repeat exposure that paywalls need to work — the same instinct driving the community and membership pushes everywhere else in today’s coverage.

📺 Big Media Moves

People Inc. Buys a Barbecue Festival, Because Events Are the New Front Page

Source: A Media Operator, Christiana Sciaudone (also Axios, Sara Fischer).

People Inc. acquired Hot Luck, the Austin food-and-music festival co-founded by pitmaster Aaron Franklin, its first pure food-festival purchase as it builds out an events business that already includes the Food & Wine Classics in Aspen and Charleston. The publisher has more than 60 events on the 2026 calendar and plans to add a second Hot Luck date and a third city by 2027.

The quote that matters comes from EVP Eric Handelsman: loyalty built at an event “trickles down into all other parts.” That is the whole thesis of the post-traffic publisher in one sentence — the live experience is no longer marketing for the website; the website is increasingly marketing for everything that pays, and events sit at the top of that list.


Hightouch Bids Up to $1.2 Billion for LiveRamp’s Identity Business

Source: Axios Media Trends, Sara Fischer.

Hightouch has put an $800 million to $1.2 billion offer on the table for LiveRamp’s identity assets — RampID and LiveRamp Connect — the infrastructure that stitches together audience data across the ad ecosystem. The move follows Publicis’s $2.2 billion agreement to buy LiveRamp, and the industry’s discomfort with a holding company owning that much connective tissue.

Identity is the quiet battleground of the post-cookie, AI-mediated web. Whoever controls the graph that links a reader to an audience segment controls how publisher inventory gets valued, and a tug-of-war over LiveRamp’s plumbing is really a fight over who sets the terms underneath every programmatic dollar publishers still earn.

📎 Also Noted

🔹 The global ad market hit its highest share of GDP on record, with the top three platforms — Alphabet, Meta, and Amazon — now controlling 57.6% of spending outside China, up from 43.8% five years ago (Axios, Sara Fischer).

🔹 The UK plans to bar under-16s from Snapchat, TikTok, YouTube, Instagram, Facebook, and X by spring 2027, joining Australia, Brazil, and others in age-gating the platforms publishers depend on for reach (Axios Media Trends, Sara Fischer).

🔹 The Bulwark is forecasting $30 million in revenue and 50% growth this year off one million email subscribers and a 14% paid-conversion rate — outrunning The Dispatch, which launched from a similar starting point, largely by developing its own talent (The Rebooting, Brian Morrissey).

🔹 Substack has finally rolled out a program to matchmake sponsors with its publishers, a belated move into advertising that signals the platform is “setting up the supply side” of a rebundling media economy (The Rebooting, Brian Morrissey).

🔹 USA Today added 1,000 Marvel comics and a new exclusive series, “Spider-Man Today,” to its games and puzzles app, leaning on licensed entertainment to deepen engagement in a product that lives outside the search funnel (Axios, Sara Fischer).

🔹 ARN Media settled with former breakfast host Kyle Sandilands for $12 million, converting an $80 million contract liability into a roughly $14 million one; the stock jumped almost a third on the news (Capital Brief, Jemeema Hanson; Mumbrella, Hal Crawford).

🔹 A sharp essay on “fashwave” argues that far-right design systems now recruit through aesthetics before ideology — and that generative image tools have automated the coding wholesale, because the models inherited it from the same centuries of training data (Splice, Rishad Patel).

🧭 Takeaways

  • Stop measuring the audience you’ve already lost. When bots are 57% of traffic and the human click keeps shrinking, vanity pageviews describe a business that no longer exists. The metrics worth tracking are the ones VG and Broadsheet watch: what a model can’t copy, and who is logged in and emailable.
  • Licensing and differentiation are not a choice of one. Le Monde’s revenue and VG’s moat are both rational; the mistake is treating either as a complete strategy. Take the AI money to fund the reporting that makes you worth citing — and never confuse a defensive deal with a product partnership.
  • Events and membership are the growth engine, not the gift shop. People Inc. and Country Living are expanding because they sell belonging directly. Any publisher still treating live experiences and communities as side projects is under-investing in the only audience it fully owns.
  • Registration is the cheapest subscription lever there is. Broadsheet’s 0.7% of readers driving 43% of conversions is not a quirk; it’s the mechanics of frequency. Knowing the reader is now the precondition for monetizing one.
  • Get machine-readable before the standards harden. The rights, access, and payment layers being drafted now will decide who can charge an agent later. Publishers whose content is structured and licensable at the article level will be ready; those shipping HTML blobs will be negotiating from behind.