AI Content Disclosure
The requirement to label ads where AI generated or substantially modified visual, audio, or video content, enforced via automatic metadata detection on Meta and self-disclosure on Google from 2026.
You upload a product image retouched in Photoshop’s Generative Fill and submit your Meta ad. Within 24 hours, it comes back rejected — not for the copy or the offer, but because the file carries C2PA metadata identifying Adobe as the creation tool, and Meta read it at upload. The disclosure requirement doesn’t wait for the advertiser to declare it; it reads the file.
How it shows up in the wild
Meta (AI Info label enforcement, 2026) Undisclosed AI content is now the third-largest ad rejection category on Meta, at approximately 14% of all rejections, per AuditSocials’ 2026 policy tracking. The detection runs at upload: an image produced by Adobe Photoshop’s Generative Fill, Shutterstock’s AI Image Generator, or DALL-E carries C2PA metadata that Meta reads and flags before the ad enters the review queue. A first violation kills the ad and adds a policy strike; a second within 90 days triggers a 24-hour account hold.
Google (AI ad disclosure rollout, July 2026) On July 9, 2026 — 24 days before the EU AI Act’s Article 50 obligations became enforceable — Google launched a “How this ad was made” disclosure panel across Search, YouTube, and Discover, per TechCrunch. Ads built with Google’s own tools (Performance Max AI asset generation, Demand Gen image creation) label automatically via SynthID watermarks and C2PA metadata. Ads built with third-party tools require the advertiser to self-disclose — which Google has publicly stated it cannot independently verify.
Why it matters
C2PA metadata is embedded by default in exports from most major AI creative platforms — Adobe, Shutterstock, Canva — and most production teams don’t audit for it before submission. My hunch is that Meta’s rejection numbers understate the actual exposure: every brand that integrated AI creative tools in 2025 is running compliance risk without necessarily knowing it. The performance concern appears overblown: a MediaScience and Adelaide University study found AI disclosure labels had minimal impact on brand recall and ad sentiment.
Related terms
- AI Voiceover Ads — AI-generated audio is separately covered by Meta’s Ads Manager audio disclosure toggle
- Dynamic Creative Optimization — Advantage+ Creative enhancements (background generation, add animation) trigger the AI Info label automatically
- Meta Advantage+ Shopping Campaigns — the primary campaign type where Advantage+ Creative enhancements run by default
Frequently asked questions
Does disclosing AI content hurt ad performance? A MediaScience and Adelaide University study found AI labels had minimal impact on brand recall and ad sentiment. 73% of Gen Z and millennial consumers say clear disclosures wouldn’t reduce their likelihood to purchase. The concern is reasonable given the newness of the policy, but the data doesn’t support it.
If I don’t know whether my asset has C2PA metadata, what happens? Meta reads the file at upload. If C2PA metadata is present, the ad is flagged regardless of advertiser intent. Auditing which tools in your production workflow embed C2PA credentials by default is the only reliable way to know before submission.
Do Meta and Google require disclosure on the same assets? The scope differs. Meta covers visual, audio, and video content substantially generated or modified by AI in commercial ads. Google’s July 2026 rollout covers content created or edited with generative AI, with automatic labeling for Google-built assets and self-disclosure required for third-party tools. The same asset can require disclosure on both platforms through different mechanisms.