On October 30, Meta’s revised Commercial Terms take effect, and Section 1 contains a clause absent from the March 6, 2024 version it replaces: Meta grants itself the right to crawl any website “associated with your products or services.” That phrase, flagged by Cornell’s James Grimmelmann, pulls millions of small-business sites inside a perimeter that previously only touched advertisers who had placed an ad order under the August 22, 2025 Self-Serve Ad Terms. A five-person shop running a Facebook Page and one boosted post is now in scope.

The crawlers are already named. Meta-ExternalAgent handles AI training. Meta-WebIndexer feeds Meta AI search. Meta-ExternalFetcher and FacebookExternalHit can bypass robots.txt under user-request and integrity-check carve-outs. Meta’s share of AI crawling went from 0.9% in July 2024 to 7.5% a year later, and its Muse agent took 33% of agentic browser traffic within nine days of launch. The infrastructure isn’t theoretical.

After the UK News Media Association condemned the terms on October 16, Meta edited them on October 17, stripping the word “train” and narrowing stated purposes. Spokesperson Daniel Roberts said the clause “only applied to our ads and business products.” VP of Communications Andy Stone wrote on X that the intent is 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.”

Legal readers aren’t convinced. Loevy & Loevy’s Matt Topic, who represents The Intercept against OpenAI, notes that “‘develop’ is broad enough to include ‘train,’” and Grimmelmann calls the revised purposes “still broad enough to drive a truck through.” The UK English preview on October 19 still carried the original “train” language. A permission to use “test credentials,” he adds, could let Meta “sign up as a fake user.”

The liability cap under the new agreement is the greater of $100 or twelve months of fees paid. The product-browsing default flips on October 28. The crawling rights land two days later. For an owner-operator whose website is the one piece of customer-acquisition infrastructure they actually own, the window to edit a robots.txt file, gate pricing pages behind a login, and audit what’s publicly linked to the business’s Meta accounts is measured in days, and the only Meta-sanctioned alternative is to stop advertising entirely.

Sources