Executive Summary

Key Figures

The first number is the one date Meta did not move. The other three come from opening the revised terms in three languages and counting.

Sources: Meta commercial terms preview (facebook.com/legal/commercial_terms_preview), US English, UK English and Korean editions read on October 19, 2026, and the verification PPC Land recorded on October 18.

Meta's commercial terms are not a document only advertisers see. They apply the moment a company runs a Facebook Page, operates a business group, or opens Ads Manager or Meta Business Suite. The previous version of this document was revised on March 6, 2024, and it contained no language about going out to crawl an external website. Section 1 dealt with content a business had itself put up inside Meta's platforms, and the right it gave Meta went as far as permission to "display, distribute and deliver your content within the Meta Products."

Automation appeared once in that version, and the actor was the business, not Meta. Section 5(c) excluded from commercial claims "efforts to interfere with our Products or engage with our Products in unauthorized ways (for example, automated ways)." The words crawler, scraper and spider do not appear in the 2024 text at all. What Section 1 granted was "a license to content that is covered by intellectual property rights (like photos or videos) you share, post, or upload on or in connection with our Meta Products."

In the version that takes effect on October 30, 2026, a new paragraph sits in that place. This is how it reads in the UK English edition of the preview.

You authorise us (or any entity acting on our behalf) to use automated software (such as web scrapers, bots, robots, spiders, crawlers and user agents) and 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, or that are made available through your or users' use of Meta Company Products and services, and to periodically pull updates from such domains, web pages or URLs.

What changed is the direction the permission faces. The thing being crawled is not what the company posted on Facebook but the company's own website. And the party saying that the website may be crawled is not Meta. It is the owner of the website, because whoever agreed to the terms is treated as having said it.

The paragraph is not the only thing that widened. The new version states that "except where we state otherwise, the Meta Terms of Service and Instagram Terms of Use ('Terms') also apply to access or use of any Meta Company Products for any business or commercial purpose." The 2024 version has no sentence pulling in the Instagram terms. The number of documents a business agrees to went from one to two.

There is no box to decline. No checkbox, no settings screen. The remedy the document offers is stated in a single sentence: "If you do not agree to the updated terms, you may stop all access or use of our Products for business or commercial purposes." As the International News Media Association summarized it, that is the only exit on the page. Close the Page, stop the ads, leave Business Suite.

The outlet that first pulled this paragraph into view was Nieman Lab, at Harvard's Nieman Foundation. In an October 15, 2026 story, reporters Andrew Deck and Hanaa' Tameez showed the terms to three lawyers and asked them to read. Each focused on something different, and the three readings pointed the same way.

Matt Topic, a media and intellectual property attorney at Loevy & Loevy, started with where the clause sits. He called it "a remarkable provision to bury in terms of service," and said the material it opens would cost many millions of dollars in a licensing market. Whether a court would enforce it is unsettled, and his advice to publishers was to assume that it would be.

James Grimmelmann, a professor at Cornell Law School, looked at how far the permission reaches. He read the wording as a license to train AI on any content from a publisher's site, with the possibility of that content then being deployed in competition with the publisher. The phrase he singled out was "test credentials." It means Meta can go through a signup flow and hold an account, and an account reaches material an anonymous visitor never sees. Nieman Lab asked whether that extends past a paywall. Meta did not answer.

Benjamin Sobel, a professor at the University of Wisconsin Law School, looked at the bundle. To use Meta's business products a company has to accept the scraping and crawling permission carried inside the same terms, and he compared that structure to tying under antitrust law. The comparison he drew was with Google, which for a long time bundled having your content appear in AI Overviews and AI Mode with appearing in ordinary search, until it opened an opt-out in August 2026. No tying suit has been filed over Meta's terms.

What let three lawyers read this paragraph as an AI story was not a word inside it. The authorization paragraph says neither AI nor artificial intelligence. Where the terms do say it is a separate clause about tools offered to the business: "Meta may use or make available products and features powered by artificial intelligence, machine learning, generative AI or other automated products or features that help you understand and use our Meta Company Products and/or optimise, enhance or create Content." The lawyers were not reading that clause. They were reading the next sentence after the authorization, the one that lists what the collected content may be used for, and on the day the story ran that list included train.

Trade bodies spoke up one after another on October 16, the day after the story. Danielle Coffey, president and chief executive of the News/Media Alliance, called the terms "outrageous and probably illegal" and said that withholding access to Meta's audience unless a business allows scraping is "a clear abuse of the company's market power" that "cannot be allowed to be implemented on October 30." She took the point past news organizations: if Meta gets away with this, she said, "every publicly-hosted word, song, and image is up for grabs." In Brussels the European Publishers Council said it would ask the European Commission and telecom regulators to look at the terms urgently, on copyright, contract fairness and abuse of market power. Angela Mills Wade, its executive director, wrote that "burying such far-reaching claims in platform terms is no substitute for consent," and that "a user sharing a news article cannot confer rights they do not hold." In the UK, Theo Bamber, chief executive of the News Media Association, called it a "shameless landgrab attempt from Meta to scrape anything and everything from news publishers and business websites to train their AI products" and asked ministers and regulators to step in.

Meta's first move was denial. Andy Stone, its vice president of communications, replied on X on October 16 to a post summarizing the Nieman Lab story and called it false. The update, he said, was "designed to make sure we're able to pull accurate information about products for sale on third-party websites when making on-platform ad and content recommendations, not about scraping news sites." Nothing about how the crawler works had changed, he added, and content owners could still manage access through robots.txt.

A day later, on October 17, The Verge reported that Meta had edited the wording. Spokesperson Daniel Roberts said: "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." Two things actually moved. A narrowing to advertising and business products went into the text, and the word train came out of the list of purposes.

More stayed next to the gap than left it. Provide, develop, improve, research and maintain integrity are all still there. Sam Quigley, a spokesperson for the News/Media Alliance, said the revision falls short: taking a word off a list is not the same as prohibiting the use, so training is "still potentially on the table." Topic, reading the revised text, said publishers "should remain concerned," since the clause still lets Meta use what it collects to develop a business product. The terms also do not say Meta will respect robots.txt. The place Stone said that was a post on X, not the document anyone is agreeing to.

This is not the first time the Alliance has spoken about Meta and publisher content. On December 5, 2025, Meta signed multi-year deals with seven news publishers, among them CNN, Fox News, People Inc. and USA Today Co., covering new and archival articles for use in Llama, on undisclosed terms. Coffey said at the time that the deals "reinforce the strength of the licensing market and also demonstrate that there is value in our content and that licensing is, in fact, possible." In the ten months since, Meta's route to publisher content has become two routes. One is a contract that is paid for. The other is a permission that arrives with a business account.

Earl J. Wilkinson of the International News Media Association took a different angle the same week. Meta's defense that its crawler has not changed may well be true, he wrote, but the crawler is not what changed. What changed is the range of material Meta claims it may crawl. Unless Meta formally re-issues the terms, he argued, publishers may leave. The precedent runs the other way. When Canada's Online News Act came into force in 2023 it was Meta that left, not the publishers, and almost three years on Canadian users still cannot see news on Facebook or Instagram.

After the edit was reported, the advertising trade publication PPC Land opened the preview page itself. It captured the UK English edition at 20:44 UTC on October 18, nearly thirty hours after The Verge's report went up at 5:04 p.m. Central European Summer Time on October 17, and the sentence was still there, with no narrowing to ads and business products either. The question the publication left on the table is the useful one. Which edition did The Verge see, and is that edition live in every language and region?

So on October 19 we opened the same preview address three times, changing only the locale. The list of purposes was different each time.

The sharpest comparison here is between two editions of the same language. In US English the purposes are folded into the authorization sentence itself and narrowed, exactly as Roberts described: the business authorizes Meta to crawl and pull updates "in order to provide, develop, improve, research, and maintain advertising and business products … and maintain the integrity and security of Meta Company Products." In UK English the purposes stand as their own sentence, and nothing is narrowed: "Meta may use any content from such domains, web pages or URLs to provide, develop, improve, research, train and maintain the integrity of, the Meta Company Products, services and features provided under these Commercial Terms and any other applicable supplemental terms." Same company, same document, same day, two English editions.

The Korean edition tracks the UK one on both counts, and adds a wrinkle of its own. Its purposes run to Meta Company products, services and features rather than to advertising and business products, and the slot where train sits is filled by 교육, a word that reads back as education rather than training. A Korean reader who went looking in the terms for the AI training word would not find it, though the clause is where it has always been.

A preview page can update at different times by region and language, and this is one day's reading. Even allowing for that, two things hold. Eleven days before the effective date, a business cannot tell from the document which edition binds it. And what was deleted is one item in a list of purposes, while the permission to crawl survives in every edition.

This story usually reads as a fight between Meta and the press. From where we work with data, something else comes into focus first. It is a story about where the permission to collect training data gets decided. Nobody built a better crawler and reached further. The range was set by adding a paragraph to a document businesses had already agreed to.

That is close to our own work. Data quality usually begins after the data has arrived. Fill the gaps, fix the labels, drop the duplicates. But what may be collected in the first place was settled a step earlier, and that step lives in a contract rather than in a technical document. Build the cleanest dataset in the world out of material that had no right to be there and no quality metric will show the flaw.

So four things are worth having an answer to before anyone asks. Nothing in the terms or the reporting sets them down. This is a list we recommend.

• Is there a list of which platform business accounts your company's domains are attached to? Facebook Pages, ad accounts, the careers page, the event signup form, all of it.
• Whose inbox receives that platform's terms-change notices? If it ends in one marketer's mail, the data lead and legal never learn that anything moved.
• Do you know which language and region edition of the terms you agreed to? As this case shows, the same clause at the same address can read differently from one edition to the next.
• Does your robots.txt name the user agents of AI training crawlers? Meta's is known as meta-externalagent. As long as the terms do not say Meta will honor robots.txt, though, that file is a request and not a contractual block.

The fourth item is the one to sit with. Blocking by technical means and blocking by contract are different layers, and the terms as they stand promise neither. The places Meta has said it respects robots.txt so far are a post on X and a statement to reporters. Whether that sentence makes it into the document is the thing to watch between now and October 30.

Thanks for reading this far. Open your company's robots.txt once today and look at who last edited it. If no name sits there, that is the whole of what this article has to say. We would like to hear how your team handles a platform terms change, and who on it does the reading.

Primary Sources

• 1.Meta. (2026). "Commercial Terms (preview, effective 30 October 2026)." US English, UK English and Korean editions, read October 19, 2026.
• 2.Meta. (2024). "Commercial Terms (effective March 6, 2024, previous version)." Read October 19, 2026.

News Coverage

• 3.Deck, A., & Tameez, H. (2026, October 15). "Meta's new commercial terms will allow the scraping of news sites, legal experts say." Nieman Lab.
• 4.Nieman Lab. (2026, October 17). "Legal experts warn Meta's new terms still grant broad scraping permissions, despite language changes." Nieman Lab.
• 5.PPC Land. (2026, October). "Meta's new terms let it crawl business websites to train AI from Oct 30." PPC Land.
• 6.PPC Land. (2026, October 18). "Meta's AI training clause still in Oct 30 terms after report of removal." PPC Land.
• 7.O'Brien, T. (2026, October 17). "Meta removes language that may have allowed it to scrape news sites." The Verge.
• 8.Stan, A. M. (2026, October 17). "Meta drops 'train' from its website-crawling terms after backlash." The Next Web.
• 9.Wilkinson, E. J. (2026, October 17). "Publishers may leave Meta unless it formally re-issues new terms." INMA.
• 10.Digiday. (2025, December 5). "Meta enters AI licensing fray, striking deals with People Inc., USA Today Co. and more." Digiday.
• 11.Search Engine Journal. (2026, October). "Meta's New Terms Expand Crawling Rights Beyond Advertisers." Search Engine Journal.

Industry Body Statements

• 12.News/Media Alliance. (2026, October 16). "News/Media Alliance Statement on Changes to Meta Commercial Terms of Use."