Meta's 2026 AI Compliance Crackdown: What Every Advertiser Needs to Know

Practitioner sources describe Meta's 2026 review shift as increasingly multimodal and AI-driven, evaluating multiple data layers at once (AGrowth.io, 2026). The term MARS — Multimodal Ad Review System — is used by practitioners; Meta does not publicly document an internal system by that name. Practitioner reports frame this as a move from reactive, keyword-based review toward proactive enforcement, but the specific implementation details are not published by Meta.

What Changed in March 2026

Meta's published guidance describes automated review and indicates that creative, text, targeting, and destination content may be considered. Practitioner reports describe increasingly multimodal review, but the previously stated count of 47 new rules was not supported by a primary Meta source and has been removed.

How Automated Multimodal Review Works

Practitioner analysis describes Meta's automated review as covering five layers (AuditSocials, 2026): text (semantic NLP that goes beyond keyword matching — synonyms, euphemisms, and indirect phrasing like “clinically shown” are evaluated similarly to “guaranteed”), visual (images and video frames for before/after imagery, medical visuals, financial charts, and HEC-related content), audio (transcribed speech and sentiment analysis), destination (crawling landing pages for consistency between ad claims and on-page content), and account history (cross-referencing the current ad against the advertiser's prior violations). Meta does not publish the internal implementation details; these descriptions come from practitioner sources.

AI Disclosure Is Now Mandatory

Any ad creative where AI was used to generate, substantially modify, or composite visual or audio content must carry an “AI-generated” label (Meta Business Help Center). Practitioner analysis (AdAmigo, 2026) estimates undisclosed AI use accounts for 14% of all ad rejections; Meta does not publish an official figure. Advertisers should follow the current labeling guidance for their format and market, and outcomes vary.

Account Health Scores

Practitioner sources (Tridigiam, 2026) report Account Health Score thresholds — for example, scores below 70 slowing review times and scores below 25 resulting in Restricted Delivery. Meta does not publish these numeric thresholds, so they should not be treated as platform fact. The system is also reported to enforce retroactively — an ad that passed review in January can be flagged and paused in March if policy thresholds are recalibrated. Every pixel, claim, and landing page may be subject to ongoing automated review, not just a one-time check at launch.

The FTC Is Watching Too

It's not just Meta. The FTC's maximum civil penalty for deceptive advertising is now $53,088 per violation (FTC, 2025). Per violation — not per campaign. In June 2025, supplement brand Evoke Wellness paid a $1.9M civil penalty (DTCskills, 2026). In June 2026, the FTC filed a contempt motion against Amare Global for violating a 2005 order (FTC, 2026). Running non-compliant ads now carries both platform-level risk (rejections, account bans) and regulatory risk (civil penalties, consent orders).

What This Means for Advertisers

Pre-submission review can reduce avoidable issues, especially in regulated categories. It is one useful control alongside current platform documentation, substantiation, destination-page consistency, and account monitoring; it does not guarantee approval or prevent enforcement.