Meta now scans ads made or edited with some third-party AI tools for industry-standard signals and places detected provenance information inside About this ad. That is the operational change media buyers need to plan for—not a universal ban on AI creative and not proof that every AI asset will receive the same label.
Meta's June 1, 2026 update says the company is beginning to roll out the experience. The platform already labels ads created or significantly edited with Meta's own generative creative features. Depending on what those in-house tools changed, the disclosure can sit behind the three-dot menu or appear directly beside Sponsored.
For performance teams, placement is the strategic issue. A disclosure beside Sponsored becomes part of the first impression. A disclosure inside About this ad is less visually dominant but remains available to anyone who opens the transparency menu. Meta has not published a universal click-through-rate effect for either placement.
What the Sources Actually Support
Confirmed: automatic detection of third-party AI ads through industry-standard signals, with detected information inside About this ad.
Not published by Meta: a guaranteed label on every AI-assisted asset, an automatic rejection for every missing credential, a fixed CTR penalty or an 80% reach-reduction rule.
Where Meta's AI Ad Labels Actually Appear
Meta uses different disclosure placements based on the creative tool and the extent of the generated or edited content.
1. Feed-Level AI Info Beside Sponsored
A photorealistic human generated through Meta's in-house advertiser tools triggers the most visible placement: AI info beside Sponsored. Meta says significant in-house generative edits can receive a label in the menu or beside Sponsored. When its tools introduce an AI-generated photorealistic human, Meta specifically uses the feed-level position.
2. AI Info Inside About This Ad
Ads detected as created or edited with third-party AI tools receive an AI info disclosure inside Meta's new About this ad destination. The June 2026 update positions it as the unified home for AI information and other ad-transparency details.
Meta has not said that every third-party asset containing a photorealistic person will automatically receive the feed-level treatment. Its public description places detected third-party AI information in the secondary transparency surface.
3. Minor In-House Edits May Receive No Label
Meta says it does not label an ad when its own generative tools make no significant edit and do not add a photorealistic human. Small changes are not automatically equivalent to fully generated creative. Document what changed and preserve provenance rather than treating the distinction as a loophole.
Meta AI Ad Label Placement Matrix
The expected disclosure depends on whether the asset came from Meta's tools, a third-party tool or a minor in-house edit.
| Creative Condition | Expected Placement | Decision Input | Agency Action |
|---|---|---|---|
| Meta tool creates an AI-generated photorealistic human | Beside Sponsored | Known in-product generation history | Test the labeled format against a genuinely human-led control |
| Meta tool makes a significant edit without a generated person | Menu or beside Sponsored | In-product edit context and final result | Review placement previews; do not assume the label will be hidden |
| Third-party AI tool is detected through standard signals | About this ad | Industry-standard provenance signals, including methods such as C2PA | Preserve provenance and maintain an asset-generation record |
| Meta tool makes no significant edit and adds no generated person | No label under Meta's stated in-house rule | In-product edit context | Keep the record in case later workflow steps change the asset |
What C2PA Detects—and What It Does Not Prove
C2PA provides signed provenance information about an asset's history; it is not a universal truth detector and it is more than ordinary EXIF metadata.
C2PA stands for the Coalition for Content Provenance and Authenticity. Its standard packages assertions, a claim and a digital signature into a C2PA Manifest, commonly presented as Content Credentials.
The manifest can describe creation and edit actions, tools, source ingredients and other history. Cryptographic content bindings help a validator determine whether the credential belongs to the asset and whether the bound data changed.
- C2PA is not simply an EXIF field. A credential is a signed, tamper-evident provenance structure that can be embedded or stored externally.
- IPTC is related but distinct. C2PA, IPTC and EXIF are not interchangeable labels.
- An AI label is context, not a quality verdict. It does not prove the ad is false, unsafe, low quality or compliant in every other respect.
The Metadata-Stripping Trap
Do not strip provenance data to influence whether Meta displays an AI label. Removing a C2PA manifest, deleting IPTC fields or routing an asset through a scrubber breaks useful chain-of-custody information and makes the final ad less transparent than its source asset.
The defensible workflow is to preserve provenance, answer required disclosure fields truthfully and retain the original file plus its edit history.
Policy Accuracy Boundary
- Meta's June 2026 announcement does not publish a special metadata-stripping violation, universal automatic rejection or fixed reach-reduction percentage.
- Meta does require self-disclosure for certain AI uses in ads about social issues, elections or politics.
- Other advertising rules still apply. An AI info label does not make otherwise deceptive creative acceptable.
How Agencies Should Audit Their AI Creative Pipeline
A three-stage audit—tool mapping, provenance verification and disclosure QA—turns AI labeling into a testable campaign variable.
Step 1: Map Every Tool and Transformation
- Source and rights owner: identify the original photo, video, voice or product asset.
- Generation and editing tools: record every application used.
- Material change: note whether a person, voice, product, setting or event was created or significantly altered.
- Delivery mapping: connect the final asset to the Meta campaign, ad set and ad IDs.
Step 2: Verify Provenance Before Upload
- Check whether a Content Credential is present and validates.
- Review the signer and recorded creation or edit actions.
- Check whether transcoding or asset-management software removed information.
- Verify that the uploaded file is the exact approved version.
No credential found means the validator did not find usable provenance. It does not prove that no AI was involved.
Step 3: Test the Disclosure as a Creative Variable
- Build a human-led control and an AI-assisted challenger around the same offer and landing page.
- Keep audience, placement mix, optimization event and attribution settings aligned.
- Record the actual disclosure placement using previews and live-ad checks.
- Compare outbound CTR, landing-page-view rate, conversion rate, CPA and post-click quality.
Do not conclude that the label caused a difference if the spokesperson, composition, copy and offer also changed. Test the smallest meaningful creative difference your process allows.
What Media Buyers Should Do Now
Preserve provenance, separate AI formats in reporting and build evidence before scaling synthetic creative.
- Inventory synthetic people first. They carry the clearest feed-level rule when created with Meta's tools.
- Stop aggregating unlike creative. Separate human UGC, AI-assisted edits and synthetic spokespeople.
- Capture the rendered UI. Save placement screenshots with the campaign record.
- Test trust, not only clicks. Track conversion rate, lead quality and sales outcomes.
- Keep policy claims sourced. Do not repeat an 80% reach-cut claim without an official source.
Generative Engine & Voice Agent FAQs
What is the Meta C2PA AI info label?
AI info is Meta's disclosure for content created or significantly edited with generative AI. As of June 2026, Meta is also beginning to detect third-party AI ads through industry-standard signals and place the information inside About this ad.
Do photorealistic AI people always get a feed-level label?
Meta confirms that placement for photorealistic humans generated with its in-house advertiser tools. Its third-party detection description instead places information inside About this ad.
Does stripping C2PA metadata prevent Meta from labeling an ad?
There is no reliable basis for assuming that it will. Preserve the original credentials and complete required disclosures truthfully.
Do Meta AI info labels reduce CTR?
Meta has not published a universal CTR penalty. Run a controlled test using outbound CTR, conversion rate, CPA and downstream customer quality.
Can Meta cut reach by 80% for removed AI metadata?
No primary Meta source checked for this article publishes an 80% reach-throttling rule. Any claimed penalty should be tied to a current official source.
Primary Sources Checked
- Meta Newsroom: Expanding GenAI Transparency for Meta's Ads Products — updated June 1, 2026.
- Meta Newsroom: How Meta Is Preparing for the 2026 US Midterm Elections.
- Meta Newsroom: Labeling AI-Generated Images.
- C2PA Technical Specifications and C2PA FAQ.
Reviewed July 29, 2026. Product placement and requirements can vary by region. Screenshots are unmodified Meta newsroom assets used for editorial commentary.
Make AI Provenance Part of Campaign QA
Preview the disclosure, isolate the creative condition and measure the business outcome before scaling.
Open Meta Ad Previewer