Meta's Updated Terms Would Let It Scrape Business Websites for AI Training

Earlier this month, researchers noticed that Meta's updated Terms of Service — set to take effect October 30th — included new language that went significantly beyond the company's existing terms regarding automated data collection. 

Businesses continuing to use Meta's platforms after that date would authorize Meta to "use automated software (such as web scrapers, bots, robots, spiders, crawlers, and user agents) and create test credentials to access, index, cache, analyze, collect, retrieve, or crawl the domains, web pages, or URLs (and any content available through them) that are associated with your products or services."

The new Terms of Service also authorized Meta to "periodically pull updates from such domains, web pages, or URLs" for use in its own training and research processes.

In plain terms: any business operating a Facebook Page would, by continuing to use the platform after October 30th, be consenting to Meta systematically scraping their website — and using whatever it found to train its AI systems. The only way to opt out would be to close the Meta account.

Legal experts quoted by Nieman Lab noted the language was broad enough to justify widespread scraping of business data, with no narrow limiting principle that would prevent Meta from crawling sites far beyond any direct advertising relationship.

What Meta Changed — and What It Didn't

After the backlash, Meta revised the terms over the weekend. The company's spokesperson said that the scraping provisions "only applied to our ads and business products" and that the updated language clarifies this — it "does not change how Meta uses web crawlers" and doesn't "give us new rights that supersede" robots.txt preferences.

Businesses with Facebook Pages will still be authorizing Meta to crawl their websites for purposes related to its ads and business products — meaning the data collected feeds Meta's advertising targeting infrastructure. What's been walked back is the claim that this authorization extends to general AI training and research. The ad targeting piece, which is what Meta's business model actually runs on, remains.

The practical distinction matters less than it might appear. Ad targeting and AI training aren’t cleanly separate in Meta's infrastructure. Data collected to power advertising systems feeds the models that power ad recommendations, which are themselves AI systems. The line between "for ads" and "for AI training" isn’t one that most businesses operating Facebook Pages are equipped to audit.

The Underlying Pattern

Meta has faced multiple lawsuits over scraping and using copyright-protected works to train its AI tools. The revised ToS language — both the original and the walked-back version — appears designed to give the company legal cover in those proceedings by obtaining forward-looking consent from businesses using its platform.

That consent mechanism has a structural problem: it's bundled into a terms update that takes effect on a specific date, and the only meaningful opt-out is closing your business's Facebook presence. For many small businesses, that's not realistic. The consent is real in a legal sense and coerced in a practical one.

We've written about how platforms' privacy decisions tend to be technically voluntary and structurally compelled. The Meta ToS revision is a clean example: a company that half a billion businesses depend on for customer reach can obtain sweeping data permissions by updating a terms document, and the businesses that object can leave. Meta's partial reversal of the most alarming language doesn't change the mechanism's design.