Turn off Meta AI enhancements
How to turn off Meta AI enhancements: what each Advantage+ toggle does, which to disable for offer-sensitive campaigns, and how to verify delivery.
How to turn off Meta AI enhancements: what each Advantage+ toggle does, which to disable for offer-sensitive campaigns, and how to verify delivery.
The worst ad edit I have seen ship was made by nobody. A promo ad went out with a hard deadline in the primary text, and somewhere in delivery Meta's text enhancements smoothed the copy into a friendlier sentence that no longer carried the deadline. People kept claiming the offer the following week. Support honored it, because the screenshot they were sent was real. The discount came straight off the margin, and not one person on the buying side had touched the ad.
That is what Meta's AI enhancements do when they run unaudited. Advantage+ creative is a bundle of automatic edits that Meta applies to your ads in delivery: rewritten text, image overlays, generative image expansion, music, animation, extra links. Some of these are harmless. A few are dangerous for any ad where the exact wording is the offer. This is the audit I now run on every account I work in.
Short answer: To turn off Meta AI enhancements, open each ad in Ads Manager, find the Advantage+ creative section, and switch off enhancements individually. There is no campaign-wide kill switch, so audit per ad. Disable text enhancements, overlays, and image expansion on any offer-sensitive ad, then verify the delivered ad in the Meta Ad Library.
The takeaways
Meta AI enhancements, grouped in Ads Manager under Advantage+ creative, are automatic modifications Meta applies to your ad at delivery time: rewriting or re-pairing text, adding image overlays, expanding images with generative fill, animating statics, adding music, appending catalog items or extra site links. Meta's help center frames them as per-person, per-placement variations that should lift performance.
The framing is fine for a brand awareness ad. It breaks down for direct response, where the copy is a commitment: a price, a deadline, a regulated claim, a guarantee. An enhancement that "improves" that sentence can quietly change what you are legally and commercially promising, and you will not see it in your own feed because variations are assembled per impression.
This is also a different failure mode from your own automation misfiring. I covered the mistakes buyers make with their own automated rules separately: rules act on metrics you chose, with thresholds you wrote. Enhancements act on the creative itself, on Meta's judgment, with no readable log of what was changed for whom. Google runs the identical opt-out play on the search side, which is why auditing Google's recommendations and auto-apply is the same job in a different account.
At the ad level: open the ad in Ads Manager, scroll to the Advantage+ creative section, click the edit control next to the enhancements summary, and switch off each enhancement individually, then publish. There is no single master kill switch that covers a whole campaign; the settings live per ad, so the audit lives per ad too.
Two complications to know before you start. First, the list of toggles you see depends on your objective, the ad format, and which rollout wave your account is in, so two accounts rarely show the identical panel. Second, Meta has been moving these controls gradually toward account-level defaults; some ad accounts can now set default enhancement behavior in the account settings. If yours has that control, set the defaults to off so new ads start clean. Then verify at the ad level anyway, because defaults govern new ads and tell you nothing about the ones already running.
The names below shift as Meta renames features, so audit by what the toggle says it does. These are the 10 families I check, ordered by how much damage they can do.
Switch off for any offer-sensitive ad:
Decide deliberately, default to off:
Usually fine to leave on:
Because the off state does not travel. Four patterns account for nearly every case I have investigated. New ads are born with the current defaults, regardless of what you set on the last ad. Duplicating an ad or campaign can re-apply default enhancements in some creation flows. Meta introduces new enhancement types over time, and a new toggle arrives in the on position, which means an audit from March says nothing about a feature added in May. And certain streamlined campaign types bundle enhancements with fewer individual controls than the standard flow.
There is a fifth pattern that deserves its own line: delivery lag. I have seen ads still serving an enhanced variant after the toggle was switched off, presumably while the change propagated. Whether that is caching or a bug makes no difference to the customer holding the screenshot. It just means the settings panel can never be your only check.
The verification habit takes 10 minutes a week and has paid for itself many times over:
If an enhanced variant is live and the numbers are sliding, treat it like any other underperformer and apply the same discipline you would use when deciding to kill an ad: check the context before you blame the creative.
I am not against automation in ad accounts. I built an automation layer myself, in Adscalr, and the rule I held it to was simple: every evaluation gets logged and every kill is reversible for 24 hours, because automation without an audit trail is exactly how the deleted-deadline story happens. That standard is what Meta's creative enhancements currently lack, so this checklist is you building the audit trail by hand. If you want to see what accountable ad automation looks like when the trail is built in, the automation page walks through it.
This is the thinking behind Adscalr.
See the product →