Short answer: Tell viewers what matters to their decision: whether the scene is a concept, a simulation, an edited photograph, or a real product demonstration. A generic “Made with AI” label can be too vague, and a provenance credential does not establish that the scene is factually true.
The first review is about the claim the image makes. In U.S. advertising guidance, the FTC evaluates words and pictures together, including implied claims, and expects support before an ad runs. Its discussion of a staged vehicle demonstration shows why an image that appears to demonstrate performance needs to represent what the product can do. That is U.S. guidance and an example, not a statement that every concept render needs the same legal label in every market.
Ask what a reasonable viewer would infer
Before approving the image, write the most likely viewer inference in one sentence. “This building is open now,” “this vehicle can perform this maneuver,” and “this is a possible design direction” are different messages. If the intended meaning is a concept, make the concept status visible near the image. A small note elsewhere on the page may not correct the main visual impression. If the image shows a feature that does not exist, consider changing the image rather than relying on a qualifier.
Next record how the material was created: photographed, rendered from a model, composited, generated, or lightly edited. The C2PA user-experience guidance says a generic “Made with AI” label does not distinguish full generation from a small AI-assisted edit. A more specific caption can answer the actual question, such as “Concept visualization of a proposed interior” or “Illustrative AI-generated scene; product shown separately.” Use precise wording that the production team can substantiate.
Finally, consider provenance. C2PA's explainer describes Content Credentials as a way to communicate origin and editing history, while explicitly saying provenance alone cannot establish whether an image is true, accurate, or factual. Keep the underlying model, photograph, approvals, and version history as evidence of what was made. Do not use a credential as a substitute for checking the product claim.
Expect mixed reactions to an AI label
In three online experiments with participants recruited in China, Shi and Jiang found opposing routes: AI disclosure increased perceived novelty but reduced perceived authenticity, with effects varying by audience orientation and product type. The studies used scenario ads and self-reported responses, so they do not predict how Archv's or a client's campaign will perform in market. They do support testing whether the proposed caption is understood and whether the image still communicates the right offer.
ARCHV has used 3D and concept visuals in automotive brand work. For any future campaign, the useful review question is not simply “Was AI used?” It is “What does this picture lead someone to believe about the real thing, and is that belief accurate?”