Meta’s defence of its new commercial terms had come down to one sentence from Andy Stone, its vice president of communications: There is “no change” to how Meta uses Web crawlers, and the update is “not about scraping news sites.”

The crawler is not what changed on Friday. What appeared to have changed is Meta’s claim about what it has permission to scrape on Web sites owned by business users of Meta properties.

What changed

Meta announced that from October 30, any company that uses Meta’s products “for business or commercial purposes” — managing a Facebook page, running a group for a business, placing an advertisement, sending marketing messages, using Ads Manager or Meta Business Suite — authorises Meta “to use automated software (such as Web scrapers, bots, robots, spiders, crawlers and user agents)” and to “create test credentials to access, index, cache, analyse, collect, retrieve or crawl the domains, Web pages or URLs (and any content available through them) that are associated with your products or services.”

Meta “may use any content from such domains, Web pages or URLs to provide, develop, improve, research, train and maintain” its products. The terms list “AI at Meta” among those products.

There is no opt-out box. The remedy offered to anyone who disagrees is to “stop all access or use of our products for business or commercial purposes.”

I don’t know of a news brand of any size without a business account at Meta. That is the point of the clause. It doesn’t need to name journalism. It names everyone, and journalism is inside everyone.

The timing is not subtle, either. On Thursday, Meta released Muse Spark Open, an 816-billion-parameter model it calls the most capable open-weight model in the United States. Models that size are hungry.

What has already been said

The reaction has been fast and, in my view, correct.

Danielle Coffey of the News/Media Alliance called the terms “outrageous and probably illegal” and said they “cannot be allowed to be implemented on October 30.” Angela Mills Wade of the European Publishers Council wrote that “burying such far-reaching claims in platform terms is no substitute for consent,” that “a user sharing a news article cannot confer rights they do not hold,” and that the EPC is writing to Executive Vice-Presidents Teresa Ribera and Henna Virkkunen in Brussels.

Nieman Lab’s Andrew Deck and Hanaa’ Tameez, who broke the story, assembled the legal scholars:

• James Grimmelmann of Cornell says the terms clearly permit training on news content.
• Benjamin Sobel of the University of Wisconsin reaches for the antitrust word “tying.”
• Matt Topic of Loevy & Loevy advises publishers to “assume it would be enforceable.”

So let me add what INMA can add: context from our own record and a plan for the next two weeks.

Did Meta really blink?

In this evolving story over the weekend, there is a possibility that this is all a gigantic error on Meta’s part. 

A Meta spokesperson told PC Mag late Friday: “We’ve heard you and want to clarify: The crawling terms we consolidated only applied to our ads and business products, and we’ve clarified that in the terms themselves. It does not change how Meta uses Web crawlers. As always, content owners can use robots.txt to signal their preferences. This is not about scraping news sites, and these updates don’t give us new rights that supersede those preferences.”

The word “train” has since been removed from the section that alarmed publishers. The revised section also mentions applying the Web scraping specifically to “advertising and business products.”

Late Saturday, NiemanLab did an update to their story after revisiting their lawyers — who are still concerned for publishers. Matt Topic told NiemanLab Saturday: “The revised version of the new terms still allows Meta to use that scraped content to develop any ‘business product.’ That doesn’t seem to be a defined term, and presumably Meta is developing AI business products. I think ‘develop’ is broad enough to include ‘train.’”

One possibility is that, by specifying “no change” to existing conditions regarding publishers’ own Web sites, Meta is signaling it believes its current terms already permit scraping content for its AI products.

If not a blink then certainly a blunder. Until media’s lawyers can evaluate the ramifications of the revised wording many in news media are considering closing their publishing presence on Facebook and Instagram to protect their content. 

An expectation of fair value exchange is not an agreement

For 25 years, the open Web ran on expectations. You let the crawler in; the search engine sent you readers. Nobody signed anything. A U.S. court has now told publishers the bargain they believed in never existed in law.

Meta’s lawyers appear to have drawn the opposite lesson from the same ruling. If an expectation is not an agreement, write an agreement. Friday’s clause is exactly that: explicit, written, accepted by continued use.

Whether an agreement imposed as a condition of access makes Meta safer or more exposed — that is the “tying” question — is for lawyers to argue. The strategic lesson doesn’t wait for them: Consent has moved from robots.txt to the fine print, and the platform wrote the fine print first.

The two-tier market

In April, Robert Whitehead, who heads up the INMA Digital Platforms Initiative, argued Meta’s return to paying publishers — US$50 million a year to News Corp, and deals with CNN, Fox News, People Inc., USA Today Co., Le Monde, Grupo Prisa, Süddeutsche Zeitung, Le Figaro, and Reuters — amounted to “an embrace of copyright law that will have wider ramifications.”

It was. We can now see their shape.

Meta has licensed journalism from roughly a dozen companies with the lawyers and the leverage to make it pay. For everybody else, it has written a clause. Ars Technica frames this as Meta avoiding “millions in licensing fees.”

I would put it more structurally. This is a content market being built with a ceiling. The six largest AI companies struck 109 licensing and data-access deals between July 2023 and March 2026 at an average of about US$24 million, according to Pradeep Gairola’s new INMA report. Meta’s terms are a bid for the rest of the world’s journalism at a price of zero, bundled into the cost of a Facebook page.

And here is the irony INMA is old enough to remember. Meta has told us three times since 2022 — most recently when it closed the News tab in 2024 — that news is “less than 3% of what people around the world see in their Facebook feed.” It is now writing that 3% into a contract.

Meanwhile, Meta is testing a cap of two outbound links a month on Facebook pages unless the page pays for its new Meta One subscription — US$499.99 a month for unlimited links. News publishers appear to be exempt for now. Meta has not confirmed it, and some small publishers say they have hit the cap.

Read the two moves together: a link to your journalism carries a price. The journalism itself is free.

The Canadian control group

The threat implicit in the clause — leave if you don’t like it — is one INMA members have already lived through.

Canadian publishers did not leave Meta; Meta left them, in August 2023 and has not come back. The damage was real. McGill University’s Media Ecosystem Observatory found almost half of news engagement vanished and almost one-third of local outlets went inactive.

But the same case file holds the other half of the story. The Walrus doubled its traffic across every online channel by accelerating newsletters, events, and games. Le Devoir lifted direct traffic from 22% to 26% of its total and cut monthly churn from 2.3% to 1.4% with a nine-e-mail onboarding journey built in anticipation of the ban.

Marianne Gélinas’s line for INMA bears repeating: “In the age of generative AI, direct traffic has become our most reliable safeguard.”

Meta’s value to publishers today is also smaller than memory suggests. Parse.ly recorded Facebook’s share of referrals falling from 14% in the spring of 2023 to about 6% by the summer of 2024; Chartbeat has since logged modest, uneven lifts.

Facebook’s performance-based monetisation programme has paid an average of about US$75,000 per page across some 700 pages tracked by Echobox. Those are the numbers to put beside the clause.

The question is not whether Meta is worth something. It is whether what Meta sends you is worth a perpetual licence to everything you have ever published.

Before October 30

Everything you already post on Facebook and Instagram (as well as YouTube and LinkedIn for that matter) can already be scraped by the platform for its use in AI training. So what about the adjacencies to that content?

Pradeep’s first instruction in his report is the right one: “You cannot negotiate the value of an asset if you do not know what you own.”

Here is what that means in the next two weeks.

Audit the footprint.

List every page, Instagram account, WhatsApp channel, and ad account in your company, and every domain they link to. The clause reaches whatever is “associated with your products or services” — your archive, your paywalled pages, your sister titles. Most publishers will find more surface area than they expected.

Re-evaluate your crawler policy and double down on it. Meta documents five user agents:

• Meta-ExternalAgent is the one Meta itself describes as its crawler for training foundation models.
• Meta-WebIndexer feeds Meta AI search.
• Meta-ExternalAds serves advertising. FacebookExternalHit builds link previews.
• Meta-ExternalFetcher fetches pages at a user’s request and, by Meta’s own documentation, may bypass robots.txt.
• If you don’t want to train Muse, disallow Meta-ExternalAgent and leave FacebookExternalHit alone. Then hold Meta to its public assurance: It told Nieman Lab it does not intend to use the terms to override robots.txt. Put that sentence in an e-mail to your Meta representative and ask for it in writing.

INMA’s Jodie Hopperton has said for more than a year that robots.txt is “a gentleman’s agreement at best.” This is the fortnight to find out whether the gentleman keeps his word — so measure the bot traffic with the tools INMA has pointed members to since 2025: Cloudflare, TollBit, Dark Visitors.

Check your gates. The “test credentials” language puts your registration wall and paywall in scope. Make sure your own terms prohibit automated account creation, and make sure you would know if it happened.

Price the relationship. Put the referral number, the monetisation payment, and your dependence on Meta’s ad tooling on one page. Beside them, put the value of your archive. Most boards have never seen those figures together. The clause has just made them a single decision.

Negotiate against the default. INMA’s Robert Whitehead reports Meta is holding fresh talks with publishers in several countries. If you are in one of those conversations, Friday’s terms are now Meta’s opening offer. Any agreement you sign should carve you out of the Commercial Terms explicitly.

Organise locally. In Sweden, publishers acted together and Meta lost its place at IAB Sweden; the Swedish parliament then passed Europe’s first one-hour takedown rule for scam advertising.

Alexander Lydecker of Bonnier News put it this way at INMA’s Berlin roundtable, in Gabe Dorosz’s account: “Local pressure works. Global pressure would work more.”

In Australia in 2021, Meta switched off news and backed down after three days. Meta responds to publishers acting in concert in a way it never responds to publishers acting alone.

The ramification that outlasts Meta

Strip away the company name, and this is what Friday told us: Consent for the use of journalism is moving from the crawler protocol to the contract. A court has said implied bargains don’t exist. The platforms are now writing explicit ones in their own favour. I would expect other platforms to read Meta’s clause very closely.

That leaves the news industry a question that matters more than Meta. If consent is going to be contractual, who writes the contract?

The movement INMA has tracked all year — SPUR in the United Kingdom, the RSL standard, IAB Tech Lab’s tokenised access, Jodie’s case for authentication over licensing — is the news industry’s attempt to answer with its own paper: machine-readable terms a publisher sets so the publisher’s licence, not the platform’s, is the default a crawler meets.

INMA’s Basic Principles for News Media, approved by our Board of Directors in September 2025, state the standard plainly: Copyright holders “have the right to decide whether or not to give another party access to their content under a licence framework and agree terms of use with that party including payment.”

Consent, credit, compensation. Meta’s clause fails all three, and we should say so without apology — and then get on with the audit.

Earlier this month, I asked members to write down their three assets — original knowledge, super users, trust — and beside each, who sets its price. Meta has just filled in the first row on your behalf.

So: What is your Web site worth to Meta? What is Meta worth to you? And who, on October 31, will have to change what they have written that down?

Header photo: Adobe stock image by Piter2121.

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .