For a few hours this week, the contract that governs how millions of businesses use Meta’s advertising and commerce tools contained a sentence that the lawyers who read it found hard to interpret any other way. Meta, it said, could “use any content from” a business account’s “domains, web pages, or URLs to provide, develop, improve, research, train, and maintain the integrity of” its products and services.
By the weekend the sentence was gone. Meta revised the commercial terms, which take effect on October 30, after Nieman Lab quoted legal experts calling the provision remarkable, and after The Verge and TechCrunch reported on the reaction. The removal was confirmed in a statement from the company, which said the language had been consolidated from existing policies and had been misread.
“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,” Meta said. “It does not change how Meta uses web crawlers.”
The clarification is accurate and also beside the point, according to the lawyers who flagged the clause in the first place. What a company is permitted to fetch from a website is settled by the file that sits at the root of the domain, by the behavior of the crawler and by the law of copyright, not by a sentence in a business contract. The terms matter for a different reason: they allocate risk in advance, and they convert an argument that would otherwise be had in court into one that has already been settled by the act of opening an account.
That distinction explains why the deletion satisfies almost nobody. A publisher who objected to the clause was not objecting to what Meta’s crawler does today, which it can observe. The objection was to the record Meta was building for what it might argue later, if a licensing negotiation or a lawsuit turned on whether a site owner had granted permission.
The population covered by the terms is large. Anyone who runs a page, buys an advertisement or opens a shop on Facebook or Instagram accepts them, and that population includes most news organizations, which maintain business accounts for distribution whether or not they sell anything through them. A clause written for merchants therefore reaches sites that exist to publish journalism, which is what made the reading plausible enough to require a response.
Meta’s legal position on training data has been tested before. A federal judge in San Francisco ruled in its favor in June 2025 in a suit brought by authors over books used to train its models, and the company has continued to face claims from publishers filed this year. A contract clause does not decide those cases. It shapes the facts that surround them, which is cheaper than litigating.
The commercial backdrop is a market for content licenses that has been forming for three years. Large publishers have signed agreements with model developers, smaller ones have not, and the price of any license depends on what the alternative costs the buyer. If a platform can point to terms that a site accepted when it created a business account, the alternative looks cheaper than it would otherwise.
Removing the sentence changes that calculation at the margin without changing the underlying capability. The crawler that visits a page to render a link preview is the same crawler that could collect text, and the operator decides what it collects. Terms describe intent. Code describes behavior.
Executives at publishing companies have said as much in private, according to people familiar with the discussions, and their attention has moved to the technical layer: server logs, robots directives and the licensing deals that a handful of large outlets have signed. Those are the instruments that produce evidence.
There is also a timing question. Terms that take effect on October 30 were published weeks earlier, and the revision arrived days before the effective date, which means the version that binds users is the one nobody has had time to read carefully. That is normal for platform contracts, which are revised on the vendor’s schedule and accepted on the customer’s, and it is why the archived copy of a clause can matter more than the clause itself.
For the businesses that use Meta’s tools, the practical remedy has not changed. A company that dislikes the terms can stop advertising, which for a small retailer means giving up the channel through which a large share of its customers find it. The asymmetry in bargaining power is the reason these documents are written the way they are.
What the episode leaves behind is a cleaner text and an unchanged arrangement. Meta will keep crawling, publishers will keep watching their logs, and the licensing conversations will continue on the terms the market sets rather than the ones a legal team drafts.
The clause is gone. The crawler is not, and the next version of the terms will arrive on a Friday too.
