Labelling and rejection are different questions. Assess an AI ad through its content, presentation and the current requirements for its intended placement.
A label provides information about content
Meta's June 2026 update describes extending AI transparency to ads created or edited with third-party tools. It is therefore incorrect to assume only content from Meta's own tools can receive AI information.
What Meta, Google and TikTok each label, and when, is set out in our guide to AI labels on Meta, Google, TikTok and EU ads.
AI information on an ad does not mean the ad is rejected. Separate how origin is disclosed from whether the particular ad meets relevant requirements. A label does not fix inaccurate products or unsupported claims.
A label can also be a legal duty outside the platform. Since 2 August 2026, deepfake ads shown in the EU need a clear label under Article 50 of the AI Act, and since 9 June 2026 New York has required a conspicuous disclosure when the business that produces or creates an ad knows it contains a synthetic performer.
Who applies the label is covered in who must label an AI ad under the AI Act and the New York and California rules. This is general guidance, not legal advice.

Check the actual ad and placement
Record where the ad will run and what was generated or changed: background, setting, person, voice or larger parts of the video. This helps the uploader consult relevant guidance and choose appropriate settings.
A review for one market or platform does not automatically cover every other one. Assess specific uncertainties against the actual ad and relevant platform and market conditions.
A hypothetical product presentation
Imagine a synthetic person holding a drinking bottle and explaining its documented size. First check the bottle and whether the person appears as a presenter rather than a real customer with invented experience.

Then check audio, captions, disclosure requirements and destination. If the ad runs in New York, that includes the conspicuous disclosure for a synthetic performer. If a later rejection concerns a claim in the copy, removing AI information does not solve it. Investigate the stated reason and correct the problem. This example illustrates a workflow, not advance approval for a particular ad.
If the correction changes a product claim, the product owner needs to approve the new version. A technical fix and a changed message are separate matters even within the same file.
