Bookshop.org noticed something strange in one of its Meta ads: an illustration nobody at the company had designed. An account update had quietly switched on a generative AI feature, and Meta’s system had decorated the ad without asking. The bookseller called the result “enhanced (or ruined, really)” and turned off every Meta AI feature it could find.
What looked like an isolated creative complaint became a contract story. Adweek reported on October 22 that agencies are pulling live Meta campaigns to remove unwanted AI additions. Three days earlier, Meta’s own engineers had explained why one of the company’s AI agents had to be split into 14 specialist subagents after it hallucinated in production. Roughly a third of the system’s changes still need correcting. Yet the new terms say that when the AI inside Ads Manager gets things wrong, the advertiser answers for them.
What changes on October 30
Meta’s revised Commercial Terms add an “AI Features” section. Under the previous wording, an advertiser was responsible for AI output only when they chose to use Meta’s AI creative tools. The new section drops that condition. Adweek’s reading: advertisers become solely responsible for whatever Meta’s AI creates or changes in their ads, even when Meta switches the feature on itself. The terms also acknowledge that the tools may make mistakes and produce unsuitable results. No one signs anything; continuing to use Ads Manager after notice is treated as acceptance.
Meta calls the update housekeeping and says advertisers can switch off individual Advantage+ features in Ads Manager. What Meta did not say, according to Adweek, is whether an advertiser is liable for a feature it never switched on.
Why this matters for performance teams
Exverus, a digital agency, told Adweek that Meta’s AI had added sitelinks, translated copy and generated images and carousels from brands’ own photos—sometimes on campaigns where nobody had enabled the features. Ryan Adams, its paid search and social specialist, said many options sit in collapsed menus: “If every one of those isn’t manually unchecked, the enhancements still apply.”
The risk is operational and reputational. Kayla Loecker of AdCellerant warns that pushing legal liability onto the advertiser creates huge operational risk. Asa Hiken of Emarketer points out that consumers don’t read Meta’s ad policies, so they’re likely to assume the brand intentionally used AI.
Not every agency is pulling spend. Go Fish Digital is reviewing the terms with legal, but it won’t tell clients to cut Q4 budget. Its president, David Dweck, called the shift “kind of the tax of advertising on Meta.”
What to audit before the deadline
Meta has a history of switching creative features on by default. PPC Land points to Dynamic Media being enabled for catalog ads between September 1 and October 20, 2025, and the REI bike ad from June 26, 2026, where an AI image tool distorted a vendor photo without REI knowingly enabling it. October 30 doesn’t change the behaviour; it changes who carries the consequences in writing.
This pre-flight list will help:
- Open every account-level Advantage+ menu, including collapsed ones, and switch off enhancements you did not choose.
- Screenshot your settings so your team has a record of what was off before October 30.
- Review live ads weekly for new sitelinks, translated copy, images or carousels you did not approve.
- Share the revised terms with legal before the holiday push, not after a campaign ships.
The practical frame: treat Meta’s AI creative features as opt-in, not opt-out. Audit them as if the default is always on, because in many accounts it effectively is.
Source: PPC Land
