AI video ad disclosure checklist
Before you publish an AI video ad
Not every AI-assisted video needs the same label. The EU AI Act separates machine-readable marking by AI providers from human-perceivable disclosure by the people or companies using the content. For marketers, the first question is whether the final ad may qualify as a deepfake, not simply whether an AI tool touched the file.
1
Write down what AI changed
Separate standard editing from generated or substantially manipulated image, audio, video, or text. Provider marking rules cover synthetic outputs; standard editing and changes that do not substantially alter the input can fall outside that marking duty.
2
Check what the ad resembles
Does it closely resemble an existing or plausibly existing person, object, place, entity, or event?
3
Check whether it could look authentic
Under Article 50, resemblance and authenticity tests help determine whether generated or manipulated image, audio, or video is a deepfake.
4
Keep provenance intact
Preserve machine-readable marks and content credentials supplied by the AI system. A visible label does not replace provider-side marking, and metadata alone does not satisfy a deployer's deepfake disclosure duty.
5
Make required disclosure perceivable at first exposure
If the content is a deepfake, disclose that it was artificially generated or manipulated in a clear and distinguishable way that people can see or hear. A visible or audible notice must be available by first exposure; do not hide it only in metadata, terms, or a separate page.
6
Record the review
Keep the source assets, consent or usage rights, tool and model, human reviewer, final label, platform checks, and launch date with the campaign file.
This checklist is operational guidance, not legal advice. Article 50 has applied since August 2, 2026, but the correct treatment depends on the content, audience, deployment context, platform rules, and other applicable law.