Meta pauses its new ad terms, and the old AI rules still apply

Meta had previewed new versions of four documents: the Meta Commercial Terms, the Product Catalog Terms, the Self-Serve Ad Terms and the Ad Creative Generative AI Terms. Its Business Help Center notice now says those updates are paused and will not take effect on November 20, 2026. Meta says it is still consulting with stakeholders and will share more later. It gives no new date.

The sentence that matters for anyone running ads is short: "our existing terms remain in effect." The terms pages themselves now carry the same message. The Self-Serve Ad Terms page says its update is on hold, and the version in force dates to August 22, 2025. The Ad Creative Generative AI Terms page says the preview would have folded those AI rules into the Self-Serve Ad Terms, and the current AI terms date to May 6, 2024.

So if you spent this month bracing for a deadline, you can unbrace. If you never heard about it, congratulations: procrastination finally paid off. Either way, this is a good moment to read the rules that actually apply today.

The Ad Creative Generative AI Terms cover the AI features in Meta's self-serve ad tools, such as Ads Manager, and in the Marketing API: the features that create new versions of your images, text and video. Use them, and you have accepted these terms. Three parts matter most for marketers.

First, checking the output is your job. The terms say the AI can produce content that is inaccurate, incomplete or misleading, and that you alone are responsible for reviewing it before it runs. Meta makes no promise that the output is accurate, unique or free of anyone else's rights, and it says another advertiser can get the same or a similar result.

Second, the output stays on Meta. Meta lets you use what its ad AI makes anywhere on its own products, and the terms call using it outside Meta's platforms unauthorized. That AI-expanded background is not headed for your Google ads, your website or your email newsletter. You also may not present AI output as made by a person, or strip any watermark or content metadata that comes with it.

Third, your inputs come with a license. Anything you upload to these tools gives Meta a perpetual, worldwide, royalty-free license to use it for the purposes in the terms, and Meta says it may use ad content to improve its AI models. The terms also bar putting any personal information into these tools. Upload only what you have the rights to: your own product photos, yes. A customer's photo you never got permission to use, no.

The current Self-Serve Ad Terms say Meta's automated setup features, which can pick your audience, placements and objectives, are optional. You can opt in or out. The same terms then say you are solely responsible for the order, the ad content, the targeting and the placements.

Together, those two lines mean the responsibility stays with you, even when an automated feature chooses where your ad shows or who sees it. Breaking the rules can mean removed ads, a suspended account or a forfeited ad balance. "The algorithm did it" might fly in a team meeting. It is not a defense in these terms.

That doesn't make automation bad. Automated placements and audiences often find buyers a person would miss. It just means someone on your team should look at where your ads actually ran and what the AI changed, the same way you would check a new hire's first week of work.

Meta's pause notice ends with a reminder that its crawling practices haven't changed, and that site owners can still signal their preferences in robots.txt. That is the small text file at yoursite.com/robots.txt that tells automated visitors where they may go. Meta's crawler documentation, updated in May, lists five crawlers. Each one does a different job.

FacebookExternalHit builds the preview card (title, description and image) when someone shares your link on Facebook, Instagram or Messenger. Meta-WebIndexer crawls the web for Meta AI's search results, and Meta says allowing it helps Meta AI cite and link to your content. Meta-ExternalAds crawls for advertising and other business products. Meta-ExternalAgent crawls for uses such as training foundation AI models. Meta-ExternalFetcher fetches pages when a user asks Meta AI to.

These are separate switches, and that's the useful part. A business can welcome the crawler that gets it cited in Meta AI answers and still block the one that trains models. Meta notes two limits: FacebookExternalHit may skip robots.txt for security checks, and Meta-ExternalFetcher may skip it because a person asked for the page. Changes can take up to 24 hours, since the crawlers cache your file. Also worth knowing: the contract goes further than the crawler page. The Self-Serve Ad Terms say that by placing any ad order, you give Meta the right to crawl your domains to develop and improve its ad and business products, including for research and training. The AI terms add the right to crawl the URLs you attach to your ads. Neither document mentions robots.txt.

A pause is not a cancellation. Meta says it will share an update, and it hasn't said what the next version will look like. Meanwhile, the rules in force already make every AI-made image, every automated placement and every claim in your ads your responsibility.

Treat Meta's AI like a fast junior designer: useful, prolific and never left to publish on its own. Want help setting up a review process for AI ad creative? Talk to our founders below. See you in the next edition.