Platform disclosure guide
Facebook AI labels: posts, ads and AI info
Direct answer
Facebook may show AI info when it detects technical AI signals or when a publisher discloses AI use. An organic post and a paid ad have different viewer-facing surfaces: an ad may show AI info in About this ad or beside Sponsored. Neither surface settles whether EU law, advertising rules or a client policy requires a separate disclosure. Check the delivered version and keep publication evidence.
Practical decisions
Facebook AI labels: posts, ads and AI info: worked examples
Each case describes a specific final asset. The platform decision is separate from legal duties, rights and client policy.
Real product photo with a generated background — organic post
Asset: The product is photographed; only the background is generated. The export may carry an AI assertion for that component.
Platform decision: AI info may be shown from a detected signal or publisher disclosure. Do not describe the whole product photo as synthetic without checking the affected component.
Action: Review the final export and the available post disclosure control; use precise copy such as ‘Product photographed; background generated with AI’ if a visible explanation is needed.
Evidence to keep: Save the original and final file hashes, edit record, metadata inspection and screenshot of the published post and any AI info surface.
Inspect embedded AI signals before posting. This does not predict whether a platform will definitely add an AI label.
The same creative is used in a Facebook ad
Asset: A paid placement uses the generated-background asset, possibly with further changes from Meta's ad creative tools.
Platform decision: Ad transparency is a separate surface. Meta describes AI info in About this ad and, for some uses, beside Sponsored; rollout and region may affect what appears.
Action: Inspect the served ad, not only the organic post or approved upload. Check the three-dot About this ad view and any label beside Sponsored.
Evidence to keep: Keep the approved asset version, ad ID, served-variant capture, label location and review timestamp.
AI info appears without a manual declaration
Asset: The team did not select an AI control, but the published image receives AI info.
Platform decision: Detection can be based on embedded industry signals. A label is not proof that the entire asset was generated, and an absence of a label is not proof of no AI use.
Action: Compare component-level edit history with the final file's signals. If the label overstates the edit, investigate the export and platform's available review route before republishing.
Evidence to keep: Retain the final file, metadata result, screenshot and any appeal or correction correspondence.
Inspect embedded AI signals before posting. This does not predict whether a platform will definitely add an AI label.
A scheduled post or ad is published
Asset: A third-party scheduler or ad workflow uploads an approved asset.
Platform decision: Do not assume the scheduler exposed every native control or that the served ad matches the approved upload.
Action: Verify the live post or served ad and the visible disclosure after publication; use a direct platform workflow where the needed control is unavailable.
Evidence to keep: Record scheduler job, platform post or ad URL, publication time and a contemporaneous label screenshot.
Do you need disclosure for your asset?
Check the final media, AI use, audience and destination separately. This is not an inspection of the file's metadata.
First-party policy text was checked on 23 September 2026. These are documented workflows, not a live test of each account, region, post format, ad placement or scheduler interface. Verify the actual publishing UI and served content. See the source register.
Applies when
- AI-generated or AI-modified image, video and audio content.
- Trigger scope: Detected industry signals or publisher self-disclosure.
- Image: Use the AI disclosure control; Meta can also detect industry indicators.
- Video: Self-disclose generated or materially altered video before publishing.
- Audio: Meta's AI info approach covers detected or declared generated audio.
Doesn't apply when
- Text: Use clear post copy where a separate disclosure is needed.
Worked decision
Example
Real product photo with a generated background — organic post
Assess the final asset and delivered label
AI info may be shown from a detected signal or publisher disclosure. Do not describe the whole product photo as synthetic without checking the affected component.
Next steps
Required action
- Review the final Facebook asset and identify generated or materially altered media.
- Use the available AI disclosure control before publishing.
- Publish and verify whether AI info appears on-post or in the post menu.
- For ads, also verify the About this ad disclosure surface.
Publishing layer
Platform differences
EU legal layer
Meta's AI info system can be broader than Article 50. Assess the legal deepfake or public-interest-text trigger independently.
Scheduler and API
Third-party interfaces may not expose every Meta disclosure control. Verify the published post or ad.
Known limitation
Label prominence differs between generated and only modified content, and Meta can apply labels based on detected technical signals.
Evidence
Source
Freshness and accountability
Last verified
Last reviewed 2026-09-23 · Current
Run this exact scenario
A representative scenario and Facebook (Meta) are preselected. Confirm the final media, AI operation and audience.
Native AI-label capability
| Native label | full AI info label |
|---|---|
| Applies to | AI-generated or AI-modified image, video and audio content. |
| Surface | Facebook organic posts and Meta ad-transparency surfaces |
| Trigger scope | Detected industry signals or publisher self-disclosure |
| Permanence | Platform-controlled |
| Reversible after publication | not verified |
| Rollback path | No verified path recorded |
| Rollback confidence | unknown |
| Post-publish edit | unknown |
| Label state verifiable | not verified |
| Legal threshold | broader Signals detected or declared AI involvement under Meta policy; it is not a legal conclusion. |
| API / scheduler | uncertain Third-party interfaces may not expose every Meta disclosure control. Verify the published post or ad. |
Content and media types
| Type | Native handling | Manual action |
|---|---|---|
| Image | native | Use the AI disclosure control; Meta can also detect industry indicators. |
| Video | native | Self-disclose generated or materially altered video before publishing. |
| Audio | native | Meta's AI info approach covers detected or declared generated audio. |
| Text | manual | Use clear post copy where a separate disclosure is needed. |
Platform policy can change independently of EU law. Disclara records both layers and never treats a native label as a legal-compliance verdict.