AI disclosure
AI agents produce the drafts of everything we make for you. People review it and you approve it. Every deliverable carries a record of how it was made, and the label that a law or an ad channel asks for.
Last reviewed 2026-10-14
What AI makes, and what people do
- Ad copy
- Headlines, descriptions and body text are written by language models from your brief, your approved brand profile and your approved claims.
- Images
- Ad images are generated by image models from the concept's visual direction and your brand's visual rules.
- Pages and email
- Landing pages and lifecycle emails are drafted by language models in the same way.
- Checks
- Separate model checks read every concept against your brand profile, your claims library and the ad platform's policies before a person sees it.
- People
- Every piece passes automated checks, then a person at Plunge reviews it and decides what reaches you. You approve it, ask for changes or decline it. Every piece is approved before anything is published.
The AI manifest on every deliverable
Each piece of copy, image, page or email we make carries an AI manifest from the moment it is produced. It records:
- The run that produced it, by id, so every step of the agents' work can be traced.
- The models used, by role: writing the concepts, checking them, generating the image.
- Synthetic people: whether the piece shows a person who was generated rather than photographed.
- Human contributions, in order: who steered a concept and what they asked for, who edited it, who approved it, and when.
- The disclosures the piece needs on the channels it was made for, with the rule behind each one.
Your workspace shows the disclosure each approved asset needs next to the asset. We can send you the full manifest of any piece.
Labels, rule by rule
These are the rules our manifest checks today. Each one is applied per deliverable and per channel, not once for the whole account.
EU AI Act, Article 50
What it asks: the providers of AI systems that generate images, video, audio or text must mark the output in a machine-readable way so that it can be detected as artificially generated, and content that imitates real people or events must be disclosed as such.
What we do: the machine-readable marking is the model providers' duty. On our side, every AI-generated image or video is flagged in its manifest as needing an AI disclosure, and that flag travels with the file to every channel.
Google Ads
What it asks: Google Ads has its own disclosure rules for ad creative made or altered with generative AI.
What we do: every deliverable made for a Google placement (Search, Display or YouTube) carries the Google Ads AI disclosure in its manifest, and we apply it when we publish.
New York General Business Law §396-b
What it asks: an advertisement that uses a synthetic performer, a digitally created person who looks like a real human, must say so conspicuously.
What we do: each concept records whether it shows synthetic people. When it does, the ad carries the synthetic-performer disclosure wherever it runs.
Before anything goes live
We publish only approved work, into ad accounts you own (Google Ads today, once Google's API access is live). Before publishing, we take the disclosure list from the deliverable's manifest for that channel and apply every label on it. If a required label cannot be applied on that channel, we do not publish the piece.
Copy-only deliverables that show no people need no AI label under the rules above, except where the channel itself asks for one. The manifest still records how they were made.
Questions
Ask us at legal@plungeai.com. See also Rights for who owns the work and Privacy for what we keep and how to have it erased.