Meta's ad moderation is in turmoil once again. Media buyers are reporting that standard video creatives—without deepfakes, famous personalities, or explicit scams—are increasingly getting flagged and blocked. Most often, these issues are linked to videos heavily modified using filters, retouching, or AI processing.
What Is Happening
Meta is actively stepping up its crackdown on deepfakes and scam advertisements featuring celebrity likenesses. The platform's automated systems analyze copy, visuals, and video frames, while utilizing facial recognition technology on select ad units involving public figures. If the algorithm identifies a potential scam, the ad creative is blocked instantly.
The issue lies in automated moderation's inability to reliably distinguish actual deepfakes from standard, heavily edited user-generated content (UGC). Consequently, buyers attribute a significant portion of recent ad disapprovals to aggressive retouching, AI filters, and unnatural facial enhancements.
At present, this remains an unconfirmed trigger. Meta does not publicly disclose the specific technical signals evaluated by its detection algorithms.
Which Creatives Are Currently at Risk
Extra caution should be exercised with videos where face and body features undergo significant post-production altering:
-
heavily smoothed skin texture;
-
modified facial structure or contours;
-
AI voice cloning or facial expression swapping;
-
actors closely resembling well-known public figures;
-
video formatting styled after news broadcasts or authority figure appeals.
High-risk verticals—such as nutra, crypto, and investment offers—are particularly sensitive, as Meta aggressively targets fraudulent ads leveraging fake medical experts, authorities, and celebrities.
Action Plan for Media Buyers
Because Meta keeps its exact algorithmic criteria undisclosed, overhaul your production workflow only with clear testing data. If disapproval rates spiked following recent moderation updates, run split-tests across multiple variations of the same creative: raw/unfiltered, minimally edited, and heavily processed.
If an unedited video passes inspection while its filtered counterpart gets disapproved, post-processing can be isolated as a primary review trigger. However, avoid forming definitive conclusions from a single ad account or test launch—Meta's automated moderation frequently applies inconsistent enforcement across identical assets.
