Still Arriving
A Norwegian court stopped Meta's new terms on Wednesday.
0:00 / 14:31
Meta's new commercial terms for business accounts were due to take effect on 20 November, three weeks after the outcry began. On Wednesday a Norwegian court stopped them until a judge has assessed whether they are lawful, and Meta said the same day that it would not change policy without a full consultation with publishers. The rest of this week's material is about where the readers and the money still are, which is more places than the mood suggests.
The injunction: a finding you can cite. Category: AI and platform. Assessment: a trend to watch. The ruling is a precedent a publisher can point at, not a protection anyone can rely on.
The Norwegian Media Businesses' Association filed its request for a temporary injunction on Tuesday, and the Oslo District Court granted it on Wednesday. The court held that the legality of the terms must be reviewed before they can take effect, and ordered Meta to pay one million Norwegian kroner in costs. 1
The association represents Schibsted, Amedia, Polaris, Mentor Media, the business daily Dagens Næringsliv, and the broadcasters NRK and TV2. Its filing argued that every article, picture and video, including the ones behind paywalls, could be scraped, and that the only way to refuse would be to delete the account and lose contact with the audience. Lene Eltvik Vindfeld, the association's board chair, said the court had agreed with the publishers that Meta holds a dominant position in the relevant market and that the new terms are unreasonable. 2
Meta's answer was not a change to the terms. Andy Stone, the company's vice president of communications, said Meta had heard publishers' feedback, does not plan to make policy changes without a full consultation with industry groups and leaders, and expects that consultation could extend beyond 20 November, in which case implementation would be delayed again. Until then the existing terms remain in place. Meta did not answer Nieman Lab's questions about which groups and leaders it intends to consult. 2
Angela Mills Wade, executive director of the European Publishers Council, said the organisation welcomed the offer to consult and the willingness to delay, and then set the condition: the consultation must lead to binding contractual safeguards, because using Meta's services to reach an audience should not amount to granting wider rights to exploit the journalism. US senator Adam Schiff wrote to Mark Zuckerberg the same week asking for guarantees on paywalls and robots.txt, with a reply requested by 2 November.
Two small papers have decided they can do without the platform. A Norwegian local title, Fjell-Ljom in Røros, will delete its official Facebook page and leave Meta from 1 January, and McKinley Park News, a community outlet in Chicago, is deleting its accounts. What the large organisations said is more instructive. Olav Nyhus, legal director at NRK, explained that the broadcaster cannot do the same, because it has an obligation to reach people and the platforms have become integral to distribution. The publishers who can walk away are the ones with the least to lose; the ones who cannot are the ones the terms were written for.
The traffic: thirty-seven against forty-one. Category: Search and discovery. Assessment: a trend to watch, and a measurement task that is due now rather than later.
Ask people whether they would click on a news link and the answers are closer than the argument suggests. When Google's AI Mode was in front of them, 37 percent said they would click at least one news or media link. With ChatGPT, 26 percent. With conventional Google search results and no AI summary at the top, 41 percent. The figures come from an Ipsos study for the Association of Online Publishers, reported in a new DJB Strategies report, From Search Risk to Agentic Risk, which Press Gazette covered on Thursday. 3
These are intentions, not clicks, and the gap between the two is the whole argument. What makes the comparison worth keeping is the reason David Buttle of DJB Strategies gives for it: Google and OpenAI are running different businesses. ChatGPT answers inside its own application, so it has no structural reason to send anyone anywhere. Google built its business on publishers allowing it to crawl for free in exchange for traffic, and abandoning that exchange would be a larger break with its own model. It still returns readers, less generously than before.
The figure that ought to change a planning meeting is not in the study at all. Chartbeat's latest numbers show Google Discover delivering more than twice as much traffic to its publisher network as conventional Google search. The bigger of the two channels is the feed, and the feed is where Google has been pushing AI summaries hardest.
That is the trap. After the UK Competition and Markets Authority required Google to give publishers effective controls over AI use without penalising them in conventional search, Google bundled AI Overviews, AI Mode and the AI features in Discover into a single opt-out. A publisher who decides to keep its work out of AI summaries has probably also switched off the channel sending it the most traffic. Most newsrooms cannot answer that question today, because their analytics report Discover and search as one number.
The agent: a visitor you can't count. Category: Product. Assessment: a trend to watch, with one decision that is available this week.
Agents do not announce themselves. Robert Whitehead, who leads INMA's Digital Platform Initiative, told the association's Media Tech and AI Week in San Francisco that agents working in computer-use mode click and type the way a person does, and that on most publishers' analytics they currently appear as human readers. "We haven't solved that yet," he said. 4
James Henderson, founder and chief executive of MonetizationOS, asked the room to treat each reader's agent as a decision engine working for its owner. It does not care about a homepage or a tier structure; it picks whichever site does the job. He calls agents the best thing to happen to the industry in thirty years, and his reasoning is uncomfortable rather than reassuring: an agent can defeat a five-screen cancellation flow in seconds, so retention built on friction will not hold, but the same friction has kept out more people than it ever kept in — the reader who wants one article, the reader who would pay for three months and not a year.
His prescription is to price the decision rather than the access. Two agents ask about building applications on the same street. One works for a curious neighbour; the other works for a buyer about to spend 600,000 dollars on a house there. Today both get the same thing, which is probably nothing. The neighbour wants the facts; the buyer's agent needs every application, current and sourced.
The unresolved case is the agent sent by someone who already pays. One publisher in the room had built an MCP connection — the interface between a publisher and an AI model — that admits an authenticated subscriber's agent, the first INMA had come across. Google is building a comparable route: Gemini Notebook's Expert Intelligence framework will let readers link their subscriptions so a research agent can retrieve articles they pay for, with publishers opting in, and the news product is not built yet. No publisher in the cohort said it was letting subscriber agents in. Julie Young, founder of Laurel, pays for the Financial Times and runs her own bot that turns her subscription into a personal feed; she said she would cancel if the paper blocked it, and considers herself the ideal customer. The paper cannot currently tell her bot from a crawler using her login, which is why blocking is the default.
The product: selling the answer. Category: Revenue. Assessment: an immediate opportunity, at a scale a regional publisher can actually run.
The Daily Mail is running between 12 and 15 generative engine optimisation tests across the US, UK and Australia, and intends to turn the work into an official product early next year. Carly Steven, director of SEO and editorial commerce, described the method at a Digiday event: identify the prompts a customer would use when researching a brand's product, track those prompts across AI platforms over weeks because the models are probabilistic, record which sources are cited, and where the publisher's own content already supplies a large share of citations and the brand is invisible, write branded content and measure again. 5
In one early test for an unnamed personal finance brand, visibility across the tracked prompts rose from about 50 percent to nearly 75 percent. That is one brand and the publisher's own figure. The measurement problem is the harder one, and Steven was direct about it: AI recommendations do not necessarily produce clicks, so referral analytics understate the effect. She cited a brand that found Claude accounting for 0.1 percent of referrals but about 11 percent of conversions once it added a form asking customers how they had heard about it.
Something similar is already a business elsewhere. Forbes sells AI-search visibility as a consultative service to brand partners and says it has improved affiliate, sponsorship and native deals; Future's GEO product has more than 30 clients with renewals; Ziff Davis reports multiple clients. A regional publisher will not run a national product. What it does have is a set of queries no national title will chase — the best plumber in a town of 40,000, which school district to buy in, where to eat on a Friday — and an afternoon with those prompts in its own market shows whose reporting is assembling the answer.
The package: selling the list. Category: Subscriber conversion. Assessment: an immediate opportunity to test, with returns that are unproven.
The Times of London has roughly 680,000 digital subscribers behind a hard paywall and now wants the readers who would pay for one thing and never for everything. Abs Ahmed, director of subscriptions, calls the approach macro payments, and is explicit that it is not micropayment: charging a little for each article has been examined for a decade, he said, and never landed, because readers come to feel they are being metered and start moderating themselves. Macro payments sell one specific product at the moment someone wants it. For The Times the opening is its annual guides, the rich list and the best places. 6
The research behind it found a significant occasional audience, more than three quarters of whom said flexible access would appeal more than a subscription, and Ahmed's position is that these readers are not lost subscribers, so the base is not being eaten. Testing begins in the next couple of months, in the UK first, and the unit of flexible access is still undecided between a day pass and a week. Abi Watson of Enders stated the risk in one line: the publisher has to get money from people who would never subscribe without letting the people who would subscribe take the cheaper route instead.
For a local newsroom the equivalent asset is whatever it already produces that people want once a year — the high school football preview, the restaurant guide, the places to live — sold to the person who arrives for it, rather than answered with a month of everything.
The week's material points at three actions, none of which requires approval from outside the building.
- Write the agent policy. One page stating what an agent acting for a paying subscriber may read and what it must present before it is admitted. Most publishers are currently blocking those agents by accident rather than by decision, and the decision costs nothing to make.
- Separate Discover from AI Mode in the analytics, and look at what each sends and whether those readers return. A decision about a single AI opt-out cannot be made without knowing which of the two channels pays for the reporting.
- Pick one product the audience wants once a year and one set of questions people in the market are putting to an AI. Test the guide as a package, and test the answer as something one advertiser will pay for.
The lawyers and regulators will settle the platform question on their own calendar. These three get settled by whoever in the building writes them down first.
References
- 1Norwegian media companies in court victory over Meta AI scraping terms
pressgazette.co.uk
- 2
- 3
- 4
- 5
- 6
This story was produced automatically by a channel. One sentence is all it takes for Neodrop to keep producing for you.