Meta has paused the business terms that would let it crawl publishers' websites after courts in Norway and Brazil, UK and EU regulators and two US senators closed in, but the crawl clause remains its stated position.
Meta has paused planned updates to its Commercial Terms, Product Catalog Terms, Self-Serve Ad Terms and Ad Creative Generative AI Terms that were due to take effect on 20 November, The Next Web reports. Existing terms remain in place. A Meta spokesperson said the company is "pausing the changes" while it continues to engage with stakeholders, and Meta has not said when the changes return or whether the crawling rules will be revised.
The pause follows pressure from several directions. A Norwegian court temporarily blocked the terms, the UK Competition and Markets Authority is continuing to examine them, the European Commission is assessing a publishers' complaint, and two US senators asked Meta's chief executive to withdraw them. PPC Land adds that Meta's notice is dated 5 November.
A court in Sao Paulo added a further order on 3 November, according to PPC Land. It suspended the crawling clause for members of three Brazilian media associations until final judgment and set a penalty of 1 million reais a day, capped at 100 million reais, if Meta scrapes their sites or uses their content to train or feed Meta AI. The judge treated AI scraping as unauthorized reproduction under Brazilian copyright law and held the take-it-or-leave-it terms abusive.
A pause is not a withdrawal: the crawl clause is still Meta's stated position, and nothing in the notice stops it returning in a revised form.
Publishers should use the pause to put their own records in place before the next version of the terms is published. I learned this building a local media network around federal compliance reporting: a commitment counts only if the reporting that proves it was designed in from the start.
Reuters booked about $19 million of transactional generative AI licensing revenue in the third quarter of 2026, Digiday reports. Organic revenue grew 12%, or 7% excluding that AI revenue.
Reuters' chief financial officer expects fourth-quarter growth to be "flattish" because the quarter laps $5 million of AI licensing revenue booked a year earlier. Reuters has three disclosed deals, with Meta, Microsoft and Amazon, and others it has not disclosed. The briefing adds that People Inc. expects licensing growth to slow to roughly 20% in the fourth quarter as it laps its December 2025 Meta deal, and it gives no AI licensing figure for the New York Times or News Corp.
The Operator’s Take: Deal revenue lifts a growth rate once, then becomes the comparison the next quarter has to beat. Reuters shows an AI licensing line can be reported on its own, and also that it arrives in lumps.
Commercial · Economic Model Report AI licensing as its own line in your forecast, and model the quarter after each deal, when you lap it, as carefully as the quarter in which it is booked.
On its fiscal first-quarter call on 4 November, News Corp said it expects the first payment from Anthropic's $1.5 billion books settlement "in the near term," without giving an amount. News Corp owns HarperCollins. Its chief executive named OpenAI and Meta as AI licensing partners, said talks with other AI platforms are progressing, and said publishers "have serious concerns about Anthropic's unlicensed use of copyrighted content" as Anthropic signals an IPO.
Dow Jones revenue rose 8% to $630 million and segment EBITDA rose 11% to $160 million, while group revenue rose 6% to $2.28 billion. TheWrap adds that the chief executive said completed deals would show up in results and that the company will both "woo" and "sue," and that News Corp has countersued Brave.
The Operator’s Take: "Woo and sue" is now a stated two-track policy, and the settlement shows what the sue track can pay. The woo track is the one whose value is still hidden: PPC Land reports that News Corp's filings assign no value to its OpenAI and Meta licenses.
Legal · Deal Execution Pair every licensing negotiation with a dated record of any unlicensed use by the same counterparty, so a settlement path stays open if the talks fail.
Ziff Davis told analysts on 6 November that organic Google referrals to CNET, IGN and PCMag fell about 37% in the third quarter, after about 23% a year earlier, and that the company is "not planning on a search recovery." Its content appeared in Google's AI Overviews and AI Mode about 2.7 billion times, nearly ten times its Google clicks.
Its chief executive said Ziff Davis is open to AI licensing deals but is waiting for clarity from its copyright lawsuit against OpenAI. Third-quarter revenue fell 5.6% to $289.3 million. PPC Land adds that USA Today Co. announced $100 million of annual cost cuts, with third-quarter revenue down 7.4% to $519.2 million, and that its $901 million suit against Google (before trebling) goes to trial on 5 April 2027.
The Operator’s Take: Exposure is now measurable. About 2.7 billion appearances against roughly a tenth of that in clicks is the gap between what Google takes and what it returns, and it is a number a licensing price can be built on, provided the count is yours.
Build · Detection & Telemetry Count how often your pages appear in AI answers and how many visits those appearances send back, using your own measurement, so the ratio is a figure you can put in a price negotiation.
Cloudflare reported third-quarter revenue up 39% to $782.5 million on 5 November, and said it recorded its first revenue from products that sit between AI agents and websites and charge for or meter that traffic. It did not disclose the amount.
Its chief executive said "OpenAI is paying for" Cloudflare's page-change signals, which stop crawlers refetching unchanged pages; no pricing was disclosed. A "leading European media company" signed a five-year, $8.1 million contract that he said gives Bot Management "a direct on-ramp for Pay Per Crawl." Cloudflare also dropped a planned Googlebot block in favor of a "Disallow AI Training" robots.txt setting.
The Operator’s Take: The infrastructure layer is now collecting from both sides of the crawl: publishers pay to block and meter, and an AI company pays to crawl less. With both amounts undisclosed there is no benchmark price yet, but the product category now exists.
Build · Detection & Telemetry Compare what a metering service reports against your own server records for one month before you sign any multi-year contract that relies on its numbers.
Anthropic submitted a robots.txt "policy token" to the Internet Engineering Task Force on 30 October, and Anthropic, Google, Microsoft and OpenAI have agreed to support it, Digiday reports. Anthropic is rolling it out now and OpenAI "in the coming months." Google is targeting 2027, and Microsoft complies and offers a grounding opt-out from 1 February 2027. Meta has voiced support but has not committed.
The token covers training only, not inference or grounding. ScalePost data shows that only 0.15% of about 1 billion requests from the four companies' crawlers over 30 days were for robots.txt. Critics at an industry coalition and a publisher ad network call it a de facto standard imposed before review, or "window dressing."
The Operator’s Take: An opt-out that covers training but not retrieval protects the smaller half of the value, because the retrieval that answers a reader's question is the use that replaces the visit. Only 0.15% of those crawler requests read the file at all.
Legal · Industry Positioning Add the token to your robots.txt file once it is published, but pair it with contract terms that cover retrieval and grounding, since the token does not.
On 29 October the Paris Court of Appeal dismissed both of Meta's appeals against interim measures adopted by France's competition authority on 8 July. Meta must negotiate in good faith on neighboring rights across all of its services, provide revenue and usage information within 15 days, and not degrade how press content is displayed. The court said Meta's payment-setting practices "appear capable of constituting an abuse of a dominant position."
The two French publisher groups involved, one representing more than 300 newspapers and the other about 920 publications and more than 60 news agencies, have received no payment from Meta since their agreements expired on 31 December 2024 and 31 January 2025.
The Operator’s Take: Regulators are starting with data: the 15-day order makes Meta show its revenue and usage before any price is set. For publishers outside France, the order is a template worth copying.
Legal · Industry Positioning Ask your trade body whether a data-disclosure request like France's can be put to your own regulator, and prepare the list of figures you would want Meta to produce.
On 30 October OpenAI published "Fair use benefits everyone," excerpting 20 court filings, including from the US government, that back fair use for AI training. It called the New York Times suit a pursuit of "an undeserved payday."
The Wikimedia Foundation asked on 29 October to file a brief supporting neither party, saying training is likely fair use but ChatGPT's retrieval of web pages likely is not. The case is before a federal judge in the Southern District of New York; summary judgment motions were filed on 4 September, the plaintiffs assert more than 10.8 million works, and no fair-use ruling has been issued.
The Operator’s Take: The split Wikimedia draws, with training likely lawful and retrieval likely not, is a line a licensing contract can follow today without waiting for the court. Training and retrieval licenses should be priced and drafted separately.
Commercial · Economic Model Draft separate price lines for training use and for live retrieval in every term sheet, so a ruling on one does not reprice the other.