Related: AI crawlers are already a load and consent problem for small sites, as in our look at the AI bot traffic flood and robots.txt rate limits.
Meta has updated the commercial terms that apply to businesses on its platforms, and the wording has publishers worried. According to a report by Nieman Lab, relayed by The Verge, the terms, which take effect October 30, 2026, appear to let Meta crawl websites "associated" with a business account and use what it finds to train its AI products. Meta disputes that anything is changing.
This is a claimed-versus-verified story: the text is public, the interpretation is contested, and nobody has tested it. Below is what each side says, what is confirmed, and what a site owner should do this month.
Update (October 30, 2026): UK CMA opens a Competition Act investigation
The UK Competition and Markets Authority has opened an investigation into whether Meta is abusing a dominant position by imposing "exploitative terms" on publishers and advertisers, Press Gazette reports (Reuters carried the same story). Under the Competition Act 1998 the CMA will assess whether Meta is dominant in a relevant UK market and, if so, whether the terms amount to abuse. It says forcing businesses to accept them "could amount to an unlawful imposition of unfair trading conditions for thousands of UK businesses".
CMA chief executive Sarah Cardell said the regulator is "acting swiftly" and wants "a swift and satisfactory resolution". Press Gazette says Meta has delayed the change and pledged not to impose it until it has consulted publishers fully; Reuters quotes a Meta spokesperson saying there are "no grounds" for an investigation. Press Gazette adds that Norwegian media companies won a court order barring the terms in Norway pending review, that some small publishers have left Meta in protest, and that publishers argue the revised wording still covers generative AI even though the explicit AI-training reference was removed. Meta has said it will respect robots.txt blocks on AI scraping, but that is not written into the terms.
Separately, the CMA says it is looking at a trial that would make Facebook pages pay a Meta One subscription to post external links; that is not yet a full investigation. Nothing here changes the confirmed-versus-unconfirmed table below: no finding has been made, and the CMA has not said Meta did anything unlawful.
Update (October 23, 2026): Meta delays enforcement from October 30 to November 20
Meta is pushing back the date its new business terms take effect. The Verge reported on October 23, citing an earlier Fast Company report, that Meta spokesperson Daniel Roberts said the terms will now go into effect on November 20 "to allow more time for us to explain to our partners and stakeholders what's changing and what's not." The original date in this post, October 30, is therefore no longer the deadline.
The pressure came from several directions. The News/Media Alliance sent a letter to Meta CEO Mark Zuckerberg on Thursday demanding revisions over concerns of "severe harm" to publishers, and digital advertisers and publishers in the European Union and the UK urged regulators to force Meta to halt the terms. The Verge notes that legal experts argued the earlier removal of the "training" language was not enough to rule out AI training.
Roberts said the extra time will be used to lay out how Meta uses web crawlers and that content owners can still use robots.txt to signal preferences. What is confirmed: the new date, and that Meta has not withdrawn the terms. What is not: the text itself has not been reported as changed in this delay, so the disagreement over interpretation stands, and the checklist below still applies with three more weeks.
Update (October 22, 2026): legal experts say the revised wording still may not rule out training
The Verge reported on October 22, citing a follow-up by Nieman Lab, that the edits do not settle the experts' concerns. James Grimmelmann, a Cornell Law School professor, said the change "does nothing to restrict" the language about content "associated with your products or services" that would let Meta scrape entire sites. Matt Topic, a media and intellectual property attorney at Loevy & Loevy, argued the revised text still lets Meta use scraped content to develop any "business product," a term that is not defined. In his words, "develop" is probably broad enough to include "train," and Meta is presumably developing AI business products.
Meta's position is unchanged. Spokesperson Daniel Roberts told The Verge that content owners can use robots.txt to signal their preferences and that the update "is not about scraping news sites, and these updates don't give us new rights that supersede those preferences." The terms still take effect October 30. Nothing here has been tested in court, so this is a disagreement about interpretation, not a ruling; the checklist below still applies.
Update (October 17, 2026): Meta removed the "train" sentence
After the backlash, Meta edited the terms. The Verge reported on October 17 that this sentence no longer appears on the commercial terms page: Meta "may use any content from such domains, web pages, or URLs to provide, develop, improve, research, train, and maintain the integrity of" its products. The Verge notes that Nieman Lab and Cornell law professor James Grimmelmann had read the clause as giving Meta a right to use scraped site content for AI training that could override robots.txt.
Meta spokesperson Daniel Roberts gave The Verge a new statement: "the crawling terms we consolidated only applied to our ads and business products, and we've clarified that in the terms themselves." He added that it "does not change how Meta uses web crawlers," that content owners can still use robots.txt, and that the updates "don't give us new rights that supersede those preferences." Earlier, communications chief Andy Stone had called the reporting false.
What is still unsettled: one secondary analysis, FourWeekMBA, reads Meta's terms preview page as keeping the permission to crawl domains associated with business accounts, with remaining uses tied to "advertising and business products". We have not independently diffed the page, so check the current text yourself. The advice below on robots.txt and licensing still stands.
TL;DR
| Question | Answer |
|---|---|
| What changes? | New commercial terms for business accounts, effective October 30, 2026 |
| What does the text allow? | Automated access to domains and pages associated with your products or services, and use of that content to "develop, improve, research, train" Meta products |
| Who says it is a big deal? | Legal experts quoted by Nieman Lab, including Cornell's James Grimmelmann |
| What does Meta say? | Spokesperson Daniel Roberts: terms are being consolidated, "no change to how Meta uses web crawlers" |
| Is robots.txt still honored? | Meta says yes; critics say the terms could be argued to override it |
| Verified in practice? | No. No dispute has tested the language |
What does the clause actually say?
The Verge quotes the key sentence. Businesses "authorize 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, 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, or that are made available through your or users' use of Meta Company Products and services."
A second sentence says Meta "may use any content from such domains, web pages, or URLs to provide, develop, improve, research, train, and maintain the integrity of" its products, services and features. Businesses that do not agree "may stop all access or use of our Products for business or commercial purposes," per the same quoted terms.
Three phrases carry the weight: "create test credentials" (more than public crawling), "associated with your products or services" (broad), and "train" (explicitly AI training, not just indexing).
What does Meta say?
Roberts told Nieman Lab that Meta is "consolidating" its commercial and self-serve ad terms, and that "any suggestion to the contrary is false." He added that content owners can still use robots.txt flags to signal preferences, as explained in Meta's developer documentation. That position is consistent with Meta's long-standing crawler guidance: its AI-training crawler, Meta-ExternalAgent, was reported by Fortune in 2024 to honor exclusion rules, and Meta said then that it trains generative models on publicly available content.
Why legal experts are alarmed
The concern is contractual priority. Robots.txt is a convention, not a contract. If a publisher accepts commercial terms that explicitly authorize crawling, Meta could argue the contract overrides the robots.txt file for domains tied to that account. Grimmelmann told Nieman Lab the terms "clearly" represent permission to train on any content from a news publisher's sites, and "possibly to deploy some of that content, even in competition with the publishers themselves."
There is a second irritant. Last month Meta launched a Meta One subscription that requires businesses and creators to pay to include links in posts and reels. As Nieman Lab puts it, Meta has begun charging users to post links to content it may crawl for free. That is commentary, but the two facts, a paid link feature and a crawl clause, are both from public sources.
Why this matters beyond Meta
Meta has struck AI licensing deals with CNN, Fox News, USA Today and People Inc., following OpenAI's lead, as The Verge notes. If the terms let Meta reach a much broader set of content without payment, they weaken the bargaining position of publishers that sell licenses. It also lands as the industry fights over training data: see USA Today's $250 million copyright suit against OpenAI and Google's multiyear AI partnership with CNN, which show the two models, litigation and licensing, side by side.
The crawler traffic itself is big. Media Copilot, citing a DataDome report, said Meta's crawlers hit websites 9.1 billion times in Q2 2026, separating a training crawler from an indexing crawler. We have not verified that count.
How consent to AI crawling has worked so far
To judge whether this is a real shift, it helps to recall the layers of control a site owner has had. The first layer is robots.txt, a voluntary file that tells named crawlers which paths to avoid. It has no legal force by itself, and its value rests on crawlers choosing to obey. The second layer is a site's own terms of service, which can forbid scraping but only bind people who accepted them. The third layer is a direct contract, such as a licensing deal, where a company pays for access and both sides sign.
Meta's clause sits in the third layer, but it is attached to a relationship that is normally about advertising, not content licensing. A publisher that buys ads or promotes articles accepts commercial terms to do so. If those terms also grant crawl and training rights, consent for AI training travels inside a document that business teams click through to run campaigns. That is the structural worry: the person accepting the terms may not be the person who manages the site's AI policy, and the two may not talk before October 30.
Meta's defense is that this is a housekeeping merge of two existing documents, a commercial agreement and a self-serve ad agreement, and that its crawlers still respect the same exclusion flags as before. That could be entirely true in practice. The difficulty is that a contract describes what a party is permitted to do, not only what it currently does. Even a company with no plans to cross a line can leave the permission in the text, and later management or litigation positions may rely on it. Publishers negotiating licenses will therefore read the words, not the reassurance.
For the wider ecosystem, watch three signals: whether Meta publishes a clarifying FAQ that names robots.txt as controlling, whether trade groups for news publishers send a joint letter, and whether regulators in the EU, where training on copyrighted works has an opt-out regime, take an interest in contractual overrides of an opt-out.
What is confirmed and what is not
Confirmed (from the quoted text and Meta's statement): the terms exist, take effect October 30, authorize automated access to associated domains, and include "train" among permitted uses. Meta says crawler behavior is unchanged.
Disputed: whether the clause overrides robots.txt or other blocking instructions, and whether it can bypass a licensing agreement a publisher already has.
Unknown: which domains count as "associated," whether Meta will use test credentials on content behind logins or paywalls, and whether any regulator or publisher will challenge the language.
What site owners should do before October 30
- List your domains tied to Meta accounts. Ad accounts, Pages, Business Manager assets and any linked websites.
- Keep robots.txt rules explicit. Add rules for Meta-ExternalAgent and other Meta user agents, and log whether they are respected.
- Check existing licensing contracts. If you already license content to an AI company, ask counsel whether the new terms conflict.
- Separate commercial and editorial domains where possible. Content on a domain not associated with a business account is less clearly in scope.
- Watch the paywall. The "test credentials" wording is the part most worth asking Meta to clarify in writing.
- Get the clarification in writing. Meta's spokesperson statement is not the contract; ask for an amendment or side letter if you rely on robots.txt.
How this fits Meta's wider AI push
Meta is simultaneously shipping AI features to consumers and businesses, from its Muse assistant on WhatsApp in a Brazilian election dispute to deepfake likeness controls, and it is under scrutiny over spending, as in the Senate Democrats' letter on Meta's R&D tax credits. Training data is the raw material for all of it, which is why a terms-of-service paragraph can matter as much as a model release.
Bottom line
Read the clause, not the headline, and not Meta's reassurance either. If Meta is right that nothing changes, the new text is harmless but unusually broad. If the experts are right, the terms give Meta a contractual argument that outranks robots.txt for any site tied to a business account. Either way, the safest assumption for publishers is that consent given in a terms update is consent, and that the time to push back is before October 30.
Related reading
-
OpenAI will honor a shared robots.txt AI-training signal: what is confirmed
-
Meta commercial terms: customers "solely responsible" for AI mistakes
-
Stanford study: news sites blocking AI crawlers and AI voting advice in Japan
Sources: The Verge, citing Nieman Lab · Fortune on Meta-ExternalAgent
Accurate as of October 16, 2026. Quotes come from The Verge's reproduction of the terms and statements; we could not open Meta's terms page directly. We will update if Meta amends the language.