Media Trendlines — September 29, 2026
📰 Key Themes
- The M&A market has stopped paying for advertising-dependent media, with one buyer telling advisers display revenue would get no multiple at all because zero-click search has killed the traffic it relies on.
- Media deals fell 46 percent in the first half of 2026 while information deals rose 88 percent and events climbed 31 percent, as buyers reprice every asset around one question: can AI copy it.
- The Atlantic now earns more from readers than advertisers, and two of its top executives spent the day describing a business built on the one thing AI cannot easily disintermediate — a direct relationship with a paying audience.
- Publishers are lobbying to legislate their own defensibility, doubling their presence on Capitol Hill to reframe journalism as critical infrastructure while accusing the Justice Department of siding with AI companies.
- OpenAI’s advertising business, the machine meant to monetize the attention search once sent to publishers, is running on projected numbers and starved for inventory.
Jump to: 📺 Big Media Moves · 💡 Business Model Innovation · 🎙️ From the Pods · 📎 Also Noted · 🧭 Takeaways
📺 Big Media Moves
Buyers Priced Display Advertising at Zero, and the Repricing Is a Map
Source: A Media Operator (Christiana Sciaudone), on the latest Collingwood Market Report.
The clearest signal of where media is headed this year came not from a newsroom but from the people who buy and sell media companies for a living. The M&A advisory firm Collingwood surveyed roughly 100 buyers and investors and found the market has split violently along a single fault line. Media deals fell 46 percent in the first half of 2026, from 54 transactions to 29. Over the same period, deals for information businesses rose 88 percent, from 51 to 96, and events deals climbed 31 percent to 47.
The quote that should be pinned above every publisher’s desk came from a private-equity-backed strategic buyer describing how it now values advertising income: “We wouldn’t apply any kind of multiple to that. Zero-click search is killing the traffic that model depends on.” Display advertising revenue, in other words, gets essentially no credit in an acquisition price. Content that AI can replicate is treated the same way. What buyers will pay for is proprietary data embedded in a customer’s workflow, contracted revenue, and a direct audience relationship that a chatbot cannot stand between.
Collingwood describes AI as an “investment gate.” Before a buyer gets to the ordinary questions of growth, margins and retention, it now asks whether the business can survive AI disruption at all — and a bad answer kills the process before it begins. Respondents said 60 percent of the deals they entered never closed. Events cleared that gate more easily than most: the private-equity buyers behind the Emerald, CloserStill and Hyve deals all cited AI-resistance as part of the thesis, at multiples estimated around 11 times earnings for Emerald and up to 14.5 times for Hyve. As one buyer put it, “You can’t send a bot to meet your customers.”
Strip away the deal mechanics and the report is a valuation of the open web’s dominant business model, delivered by the only judges whose opinion carries a price tag. The verdict is that ad-supported, search-fed, commodity content is now close to worthless as an asset, while owning a relationship AI can’t disintermediate is the whole game. Collingwood expects more media companies to end up in fire sales than in managed auctions next year. The advertising-and-traffic model didn’t just lose readers; it lost the confidence of the people who would have to write the check.
Publishers Try to Legislate the Defensibility Buyers Won’t Pay For
Source: Axios Media Trends (Sara Fischer, Kerry Flynn) — ⚠️ Paywalled; summary based on the available preview.
If the acquisition market won’t reward a defensible position, the industry is trying to build one in Washington. The News/Media Alliance has roughly doubled its lobbying force, sending about 140 executives to Capitol Hill, and reframed its pitch entirely. Journalism is now being sold to lawmakers as “critical infrastructure” for American competitiveness against China in AI — a national-interest argument rather than a plea for a dying trade. The immediate target is a “Bad Bots” bill to give publishers stronger anti-scraping protections, with the expectation that content licensing becomes a durable revenue line the way it did for the music industry.
The lobbying push arrives because the alternative — the courts — is looking less friendly. Axios reports the Justice Department filed a statement supporting AI firms in publisher copyright suits without consulting the Patent and Trademark Office or the Copyright Office, prompting accusations of political favoritism. Even a Microsoft director warned of a “doom loop” in which training AI on unpaid content eventually destroys both the models and the web that feeds them. Meanwhile the agencies that might police the AI giants are hollowing out: FTC competition staff have fallen from 744 to 600, the DOJ’s antitrust ranks from 910 to 776, and merger-review times sit at ten-year lows.
Put the two stories together and the shape of the moment is clear. Buyers have decided the market won’t protect publishers, so publishers are asking the government to — at exactly the moment the government’s appetite for restraining large technology companies is at its weakest in a decade. Licensing may well become a real revenue line, but an industry betting its future on legislation is an industry that has run out of leverage in the market.
💡 Business Model Innovation
The Atlantic Is the Working Model of the Asset Buyers Actually Want
Source: A Media Operator (Christiana Sciaudone) and The Rebooting (Brian Morrissey).
On the same day the deal market declared reader relationships the only asset worth buying, two of The Atlantic’s leaders described a company built on exactly that. Editor Jeffrey Goldberg told a journalism festival in São Paulo that the magazine now has 1.6 million subscribers, up from one million two years ago, and that subscriptions have overtaken advertising, which supplied 80 percent of revenue eight years ago. His diagnosis of what went wrong across the industry was blunt: publishers stopped believing their work was worth paying for. “Imagine you opened a restaurant and didn’t think your food was good enough to charge for. That’s what we were doing.”
The Atlantic’s CEO, Nick Thompson, supplied the forward-looking half in a conversation with Brian Morrissey. He sketched three scenarios for 2031 — AI plateaus, AI shatters the information economy, or AI proves as disruptive as the internet while a recognizable journalism business survives — and is betting on the third while hedging the other two. The hedges are instructive: expanding print from 10 to 12 issues a year because postal delivery is a distribution channel no platform can throttle, striking a micropayments deal with Parallel to charge AI agents that ingest its work, and hiring aggressively on the theory that AI will empower individual talent, so the institution’s job is to keep the best individuals inside it.
Thompson’s framing of the risk is the sharpest line in either interview: “The businesses that get disrupted are the ones that are closest to the way it was trained.” Generic, search-optimized, ad-supported content sits closest to what the models already do for free — which is precisely why Collingwood’s buyers won’t pay for it. The Atlantic’s answer is to move in the opposite direction: reporting a model can’t replicate, a reader who pays on purpose, and distribution the company controls end to end. It is not a magic formula. It is the unglamorous work of being worth the money, which Goldberg called “no secret” at all.
The Smartest AI Play Isn’t Faster Articles — It’s Turning the Archive Into a Product
Source: The Media Copilot (Pete Pachal).
Nearly every AI pitch to a newsroom is a promise to do the same work faster: quicker drafts, quicker headlines, quicker social copy. Pete Pachal argues the more valuable move is the opposite — using AI to offer readers something they could never get before. The clearest example is Nursing Times, a trade title whose answer engine draws only on its own clinical and news archive. It had fielded more than 200,000 questions by mid-2025, with roughly 100 subscriptions directly attributable to the feature.
The detail that matters is how people used it: 80 to 90 percent of the queries came from suggested questions embedded inside articles, not from an open chat box. “Readers don’t want to interrogate an archive,” Pachal writes. “They want the next question answered right where they already are.” That reframes a publisher’s back catalogue from a cost center into inventory — the same proprietary, hard-to-replicate asset the deal market is now willing to pay a premium for, expressed at the level of a single page.
The WordPress angle: An archive only becomes a product if the platform can surface it inside the reading experience — suggested questions rendered in the article template, answers drawn from a site’s own content rather than a generic model, all served fast from infrastructure the publisher controls. The lesson of the Nursing Times numbers is that the feature lives or dies at the point of the page, which makes it a content-management problem as much as an AI one.
🎙️ From the Pods
OpenAI’s Ad Business Is Running on Hopeful Math and Empty Inventory
Source: 🎙️ The Digiday Podcast (Tim Peterson and Kamika McCoy with Krystal Scanlon), September 29.
If AI is supposed to inherit the advertising money that search once funneled to publishers, its front-runner is not ready to receive it. Digiday’s reporters picked apart OpenAI’s claim that ChatGPT ads hit a $1 billion “annualized run rate,” noting the phrase describes a projection, not booked revenue — take one strong month or quarter and multiply. “My run rate looks a lot better on payday weeks versus non-payday weeks,” one host observed. An internal target of $2.5 billion in actual ad revenue for 2026 has reportedly been missed, and independent forecasters put the entire U.S. chatbot ad market at only about $5.4 billion by 2030 — against OpenAI’s stated ambition of $100 billion in ad revenue by then.
The binding constraint isn’t demand; it’s inventory. Buyers described clients spending $10 million a month on Google who cannot place more than $100,000 a month on ChatGPT, simply because there aren’t enough conversations with commercial intent to carry the ads. Measurement is thin, legal teams balk at the data terms, and the most promising fix — a Shopify integration to pull in a long tail of small merchants — is still early. The takeaway for publishers is bracing: the model that supposedly makes their ad business obsolete is itself a test-budget line item propped up by projected numbers. The old ad economy is being dismantled faster than any replacement is being built.
📎 Also Noted
🔹 Malicious bot traffic grew 124 percent between July 2025 and June 2026 — more than nine times faster than human traffic — with scraping making up 70.9 percent of it, and AI agents logging 605.6 million requests to login, cart and account pages in the first half of the year, according to DataDome’s latest report. (The Media Copilot)
🔹 About one in five visits to News UK’s sites over the past six months came from bots rather than people, CTO Tom Jackson said, with roughly 6 percent of total traffic blocked as malicious; he also noted Google impressions skew toward evergreen stories while ChatGPT impressions skew toward live, timely ones. (The Media Copilot)
🔹 NewsGuard has taken its vetted-source chatbot into France, Italy, Germany, Austria and the UK, sharing revenue with the roughly 12,000 publishers it cites and letting partners keep half the €8-a-month subscription revenue — but has not disclosed what a single citation actually pays, the number that decides whether this is a business or a press release. (The Media Copilot)
🔹 Adobe has moved Acrobat and its creative tools inside Claude, exposing more than 80 tools in the chat window, with a Gemini integration to follow — another sign that the document and design work publishers do all day is migrating into the assistant. (The Media Copilot)
🔹 OpenAI issued an open letter apologizing for unauthorized access to Australian government websites during training earlier this year — a rare public mea culpa that says more about mounting scrutiny of how models are fed than about any change in practice. (Mumbrella)
🧭 Takeaways
- Advertising revenue is now a discount, not a premium. When sophisticated buyers refuse to attach any multiple to display income, a publisher still organized around ad-supported traffic is running a business the capital markets have already written off.
- Defensibility is the whole valuation. Proprietary data, contracted revenue and a direct reader relationship are what command a price, because they are the parts of the business a model cannot reproduce or stand between. Everything else is being repriced toward zero.
- The Atlantic is a template, not a miracle. Reader revenue over ad revenue, owned distribution over borrowed reach, and talent over volume are available to any publisher willing to make something worth paying for — the hard part Goldberg admits is “no secret.”
- Don’t count on inheriting AI’s ad money. The largest AI advertising business is still running on projected figures and can’t absorb real budgets for lack of inventory; the old ad economy is being torn down long before a new one is ready to replace it.
- Turn the archive into a product before someone turns it into training data. The most durable AI use in media isn’t faster output — it’s making a proprietary back catalogue answer the reader’s next question, right on the page.