Meta commercial terms cover firms that buy no ads. Google and OpenAI tie ad bots to ads

From November 20, 2026, Meta commercial terms let the company crawl the site of any business with a Facebook Page or Instagram business account. That includes firms that never buy an ad. We read the new text and set it next to the ad crawler rules of Google and OpenAI. Both tie their ad bots to pages that carry ads. Meta ties its permission to the account itself.

Meta moved the start date to November 20 and kept the clause

A Meta spokesperson gave Reuters the new date on October 23. The old date was October 30. Meta said it wants time to explain “what’s changing and what’s not.” The delay covers the Commercial Terms and the Self-Serve Ad Terms. The text did not change.

One change came earlier. On October 17, The Verge reported that Meta had cut a sentence letting it use crawled content to train its products. The rest of the clause, which we covered when it first appeared, stayed. Meta may still send bots to a business site. It may also create “test credentials,” meaning a login that lets a bot pass a sign-up page.

The terms apply through use. A business that keeps using Meta’s products after the notice accepts them. There is no opt-out and no partial acceptance. The terms also cap what Meta can owe a business. The cap is $100 or the amount paid to Meta in the past 12 months, whichever is higher.

Google, OpenAI and Meta draw the ad crawler line in different places

All three companies run separate bots for separate jobs. The difference is how far each ad bot reaches.

Google runs AdsBot for ads. Its crawler documentation says the bot checks the quality of pages that Google Ads sends people to. AdsBot ignores the general “*” rule in robots.txt, the file where a site tells bots what to skip. Google says it does this with the advertiser’s permission. The AdSense crawler works the same way, as we reported when Google explained what blocking Mediapartners-Google does. For Gemini training, Google offers a separate control called Google-Extended. Blocking it does not affect a site’s place in Google Search.

OpenAI launched OAI-AdsBot with ads in ChatGPT. Its bot page says the crawler “only visits pages submitted as ads.” OpenAI also says data from this bot is not used to train its foundation models. Training has its own bot, GPTBot, which sites can block.

Meta lists five crawlers on its developer page. Meta-ExternalAds collects content for ads and business products. Meta-ExternalAgent collects content for training AI models. The page says robots.txt controls both. Two others, FacebookExternalHit and Meta-ExternalFetcher, may bypass robots.txt in some cases. The new terms are wider than a landing page check. They cover domains linked to a firm’s products and links that other users share.

Meta promises robots.txt control in interviews but not in the terms

Meta told Reuters that site owners can still use robots.txt to signal their wishes. A Meta spokesperson told Nieman Lab the same. The terms themselves do not say this. We checked the Polish preview of the new text on October 26. It does not mention robots.txt. It still showed October 30 as the start date, three days after Meta announced the delay. PPC Land found the same old date on the UK English preview on October 23.

The bot clause names web scrapers, bots, spiders, crawlers and user agents. The stated purpose is to build and improve Meta’s “advertising and business products.” It also covers keeping Meta’s services safe. James Grimmelmann, a Cornell law professor, told Nieman Lab the first version gave Meta permission to train AI on publishers’ content. That first version still had the training sentence.

European publishers asked Brussels and London to stop the terms

The European Publishers Council asked the European Commission to look at the terms on October 16. Its executive director, Angela Mills Wade, said other uses should need “separate, express authorisation.” On October 22, the Movement for an Open Web filed a complaint with the Commission and the UK Competition and Markets Authority. It asked for interim measures to suspend the terms. The Commission said it would assess the complaint under its usual procedure.

The UK’s News Media Association called the delay “a pause not a solution.” Its chief executive, Theo Bamber, asked the competition authority for public commitments from Meta. In Poland, the Polish Chamber of Publishers said on October 19 that publishers should consider whether their pages stay on social media. The complaints in Brussels and London came from publishers. But the clause covers any business with a Page, from a hotel to a dentist.

EU lawmakers gave sites a machine-readable opt-out in 2019

Article 4 of the EU copyright directive allows text and data mining, meaning the automatic reading of large amounts of text. Rights holders can reserve their content. Under Article 4(3), online content needs a reservation by “machine-readable means.” Poland, for example, copied it into Article 26³ of its copyright act.

Lawyer Matt Topic told Nieman Lab that publishers should assume a court would enforce Meta’s clause. Meta also stayed out of the EU’s code of practice for general-purpose AI models. Joel Kaplan, Meta’s chief global affairs officer, announced that on July 18, 2025. The code asks model makers for voluntary copyright commitments. Meta did sign a separate EU code on labelling AI content.

Every Page owner falls under the new clause on November 20

On November 20, the clause applies to every business that keeps using a Facebook Page or an Instagram business account for work. Until then, the current terms from March 6, 2024 still apply. Site owners who want to keep Meta’s training bot out can add a “Disallow” line for meta-externalagent in robots.txt. Meta’s own crawler page gives that method.