Advantage plus enhancements: keep, kill or test
Advantage plus creative enhancements are not the enemy. Silent, unverified changes to tested creative are. Each enhancement below gets a plain description of what it changes, and a call: keep it if it adapts format without changing meaning, kill it if it alters the message, test it if the answer depends on your brand. The API notes at the end are for anyone controlling this at scale rather than one ad at a time.
Usually safe to keep
These adapt your creative to a placement without changing what it says
Automatic cropping and reframing
Fits your asset to placements you did not export for. Keep only if you have not exported every ratio yourself. If you have, this can override a deliberate composition.
Standard enhancements, format level
Minor visual adjustments to fit the surface. Low risk on lifestyle imagery, higher risk on anything with typography or precise brand colour.
Relevant comments surfacing
Promotes existing comments into the ad view. Harmless when your comment section is healthy, which is a reason to moderate it.
Usually kill
These change the message, the art, or where the click goes
Text improvements
Rewrites your primary text or headline. Every claim you legal-checked can be reworded into one you did not. The single most important one to turn off for regulated categories.
Image touch-ups, brightness and contrast
Adjusts your art. On a brand that spent weeks on colour, this is an unapproved edit. On a screenshot or a product shot it can be actively misleading.
Image expansion and uncrop
Generates image area that your designer never made. What appears in the invented region is not something you approved.
Music
Adds a track to a silent or narrated video. Changes the tone, and occasionally fights the voiceover.
Enhance CTA and CTA text swaps
Changes the button language, which changes expectation before the click and skews measurement of what you actually tested.
Site links and destination adjustments
Anything that changes where the click goes should be your decision, since it silently changes what the campaign is optimising toward.
Worth testing deliberately
Not harmful in principle, but test them as a variable rather than leaving them on by default
Catalogue and product tags
Can genuinely help ecommerce conversion. Test as its own variant instead of enabling it mid-flight on a running test.
Visual touch-ups on UGC
Creator footage is often improved by light adjustment. If it wins, adopt it in the edit rather than delegating it to delivery.
3D animation and motion effects
Sometimes lifts static assets, sometimes looks cheap. Test per brand and treat the result as brand-specific.
Doing this at scale
Opt out per feature, not per bundle
Since v22 the single bundle toggle is gone. Control moved to per-feature entries in degrees_of_freedom_spec.creative_features_spec, each with its own enroll status.
The eligible key set changes across API versions
Keys that existed last year may be absorbed into a catalogue entry this year. Reconcile your list against the live reference each release rather than hardcoding it forever.
Read the creative back after publish
Requesting an opt-out and receiving one are different events. Fetch the creative after it is live and compare the applied set against what you asked for. Anything else is an assumption.
Check again after review
Enhancement state can differ after an ad passes review on some accounts. One check at publish plus a weekly sweep catches it.
Account defaults do not stay put
Meta resets or introduces enhancements at the account level periodically. If your process depends on a default staying off, it will eventually fail quietly.
If you are managing this on one ad, the Ads Manager toggle is enough. Across a batch, it needs to be configuration with verification attached, because the failure mode is silent by design.