AI-created ad visuals: keep an asset register before choosing a label.
A practical creative handoff for provenance, permitted use, factual claims and platform declarations, without treating a label as approval.
In this article
A practical creative handoff for provenance, permitted use, factual claims and platform declarations, without treating a label as approval.
Know what changed and what did not.
Microsoft’s 19 November 2026 announcement describes AI content transparency controls as generally available, with declarations and supported visible labels for image and video assets. It also says controls vary by format and workflow. A declaration is not a guarantee of policy approval or legal compliance.
The useful operational change is to make provenance part of the creative handoff. If the person uploading an asset cannot explain how it was made, they cannot reliably decide which disclosure setting fits. Do that review before the asset is distributed across campaigns.
Build one row per final asset.
Use a register containing the final filename, version, author or supplier, source materials, editing method, permitted uses and approval owner. Record whether a visual depicts a real product or an illustrative concept. Keep supporting permissions in a controlled location rather than inside a public campaign folder.
An asset can combine photography, illustration and generated material. Describe the meaningful edits clearly enough for a reviewer to understand them. A single checkbox called creative approved loses the distinction between permission to use a photograph and approval of a changed product appearance.
Review the visual promise.
Illustrative example: a cosmetics brand uses a generated scene around a real product photograph. The packaging may be accurate, while the scene implies an ingredient, texture or outcome the product cannot support. Review the whole impression rather than only checking whether the logo is correct.
For healthcare, counselling and professional services, avoid visuals that imply a real patient, practitioner endorsement or proven outcome when none exists. A concept should be presented as a concept. Route regulated or sensitive claims to an appropriately qualified reviewer instead of treating design approval as specialist clearance.
Check the actual platform workflow.
Have the upload owner inspect available declaration controls for the specific asset and format. Capture the chosen setting, date and approved file version. A label on one exported image should not be assumed to carry into a newly cropped or re-uploaded version.
Test legibility after platform previews crop or resize the asset. Important qualifications should remain understandable on a phone. Where the format cannot communicate the offer accurately, choose a different creative rather than trying to solve the problem with tiny text.
Give changes a review trigger.
Name the changes that require another approval: altered product features, a new person’s likeness, revised claims, a new territory or a different usage channel. Keep archived versions separate from approved launch assets so an older export is not reused accidentally.
A disclosure setting can help communicate how material was created. It cannot establish ownership of source material, substantiate a performance claim or grant a model release. A stronger workflow makes each of those questions visible before media spend begins.
- Track provenance and permissions for the exact final file.
- Review implied claims as well as written copy.
- Record declarations by platform, format and version.
What would this
look like for you?
Bring the message, creative and customer journey together around a clearer next step. Explore our creative & conversion service, or tell us what you are working on so we can discuss the scope.
Put the thinking into practice.
Sources & further reading
Platform references support the factual background. The planning recommendations are AdzUp’s editorial interpretation.
A useful read.
A clearer next move?
Keep exploring, or bring your questions to AdzUp.

