Meta advertisers pull AI-altered ads as new terms shift liability to brands
Bookshop.org and Exverus say Meta's automated tools changed ads without their approval, as new terms shift AI output risks onto advertisers.
Meta advertisers are removing live campaigns after discovering unexpected AI changes to their creative assets, just as the platform prepares to apply new commercial terms assigning broad responsibility for AI-generated results to businesses using its tools.
The problem is no longer hypothetical. Adweek reported on October 22 that online bookseller Bookshop.org and digital agency Exverus encountered advertising changes they had not intended to publish. The incidents put the practical limits of automated creative optimization under scrutiny ahead of Meta's October 30 terms update.
Key Takeaways
- Bookshop.org and agency Exverus reported unexpected AI changes to Meta ads, with some live campaigns removed for correction.
- Meta's revised commercial terms take effect October 30 and place broad responsibility for AI feature output on business users.
- Advertisers should review AI enhancement settings and inspect the ads audiences actually see, including after account updates.
Table of contents
Jump to each section:
- How unwanted AI edits reached live ads
- What Meta's October 30 terms change
- Why existing creative reviews can miss changes
- What advertisers should check before October 30
How unwanted AI edits reached live ads
Bookshop.org said an automated account update enrolled it in a generative AI ad feature that inserted an AI-generated illustration into an ad originally made by people. The company said it had not known about the change and subsequently disabled Meta's AI ad features.
Agency Exverus described a wider range of alterations. Its team saw additional sitelinks, translated copy, generated pictures and carousel variations based on client imagery, including cases where buyers said they had not selected the relevant enhancement. Exverus told Adweek that it had pulled live advertisements to remove alterations that escaped its review process.
Ryan Adams, a paid search and social specialist at Exverus, said Meta "auto-applies the enhancements" and that options can be difficult to find inside collapsed menus. That creates an operational gap between the creative a brand approves and the version ultimately displayed to users.
What Meta's October 30 terms change
Meta's preview of its Commercial Terms says its revision takes effect on October 30, 2026. The updated section covering AI features acknowledges that automated outputs can be inaccurate or unsuitable and says businesses assume responsibility for assessing and using the resulting content.
Advertisers already had obligations under Meta's generative AI ad terms. The important distinction in Adweek's reporting is whether those obligations apply only when a business voluntarily chooses a tool or also when automated changes are applied without the buyer's deliberate selection. Meta told Adweek the update is part of consolidating its existing advertising AI terms and said individual creative options can be disabled.
That language represents Meta's contractual allocation of risk. It does not establish how every potential dispute would be decided under applicable law. For marketers, however, the immediate commercial concern is clear: a buyer may need to answer for creative that was changed after the original approval.
Why existing creative reviews can miss changes
AI creative production is now part of Meta's broader advertising roadmap. In its October 6 advertising announcement, Meta outlined expanded Ads Creative Studio access and tests that let its business assistant generate imagery and text, set up campaigns and recommend adjustments. Meta said some of those newer assistant tests include previews, approvals and controls.
The reported unwanted-edit cases raise a separate, more immediate question: whether an account's existing enhancement settings reliably reflect what the advertiser believes it has approved. An agency can check its original image and copy, then still miss a platform-generated version running in another placement or appearing after a settings change.
This matters beyond aesthetics. A changed product image, translated claim or unexpected illustration can affect brand standards, rights clearances and the meaning of an ad. Consumers generally see the brand's name, not the configuration choices inside Ads Manager. The cost of catching a mistake late can include rework and a loss of confidence in campaign controls.
What advertisers should check before October 30
Advertisers should treat AI-enhanced creative as a live production output requiring its own checks. Start by reviewing enhancement options across active campaigns and checking whether account or campaign updates have changed the configuration. Do not assume a previously approved asset guarantees that every delivered variation remains identical.
Teams should also preview available variations and placements, compare them against approved claims and visuals, and recheck live ads after major account changes. Where a feature is inappropriate for a brand, disable the applicable option when the account provides that control and document the setting for the next reviewer.
Agencies managing several advertisers need a clear owner for these inspections and a quick escalation process if an unintended version is already live. Legal and brand teams should review the updated terms before October 30, particularly for regulated products, licensed creative and campaigns involving sensitive claims.
Meta's automation can reduce repetitive creative work, but the Bookshop.org and Exverus experiences show why final-output inspection still matters. The crucial control is knowing what was actually delivered under the brand's name, not merely what the team uploaded.
