Telling the Client: How Agencies Should Disclose AI Use in Campaign Work
Updated: 1 day ago
An agency should tell its client, in writing and before production starts, where AI will be used in the campaign work, which parts of each deliverable it touched, which external systems and which production partners are involved, what material of the client's goes into those systems, and who on the human side checked and approved the result. After delivery, the agency should confirm what was actually done, because plans change during production. Put the duty into the contract as well: a disclosure clause, a warranty about rights in the outputs, a clear statement of who bears the risk if a rights claim arrives, and a deletion commitment for the client's material. Disclosure to the client is a separate matter from labelling for the public. It protects the brand's ability to make its own decisions about rights, reputation and compliance, and it is much easier to agree at the start than to reconstruct after a complaint. This is general information, not legal advice; have counsel review the final terms.
Why the client needs to know
A brand that publishes an image carries the risk for it, whoever made it. If the image draws a rights claim, a regulator's question or a public complaint, the first call goes to the brand, and the brand then asks the agency what happened. An agency that cannot answer quickly has a problem, and so does the client.
Three things make AI use different from a conventional retouching or post-production step:
The rights chain gets longer. The output depends on a generation system, on the material fed into it and on the terms attached to both. Who Owns an AI-Generated Campaign Image? shows why that chain has to be written down.
Client material travels further. Product files, unreleased collections and images of models may pass through systems the client has never heard of.
The label may be a legal question. Some imagery has to be disclosed to the public, and the brand cannot judge that without knowing how the image was made.
What to disclose before production starts
A short disclosure note, attached to the brief or the statement of work, covers the following points.
Where AI is used
State the stages: concept visuals, background generation, product variants, retouching, a full synthetic image, a digital twin of a real model. A brand reads "AI-assisted" very differently depending on whether it means cleaned-up skin tones or a generated person. Name the deliverables concerned instead of describing the studio's approach in general terms.
Which systems and which partners
List the generation systems and the production partners that handle the work, including freelancers and a studio working for the agency. The client does not need a technical specification. It needs to know who sees its material and under which terms, so that confidentiality, data protection and ownership can be checked against the right documents. Agree that the agency tells the client before it adds or replaces a system. The agency's own promise only reaches as far as its suppliers' terms, which is why How to Vet an AI Production Studio for Brand Safety asks about them.
What client material goes in
Say what the agency uploads or feeds into a system: product images, reference sets, brand guidelines, model images. Say whether anything is used to train or fine-tune a model, who controls that model, and what happens to it at the end of the project. The confidentiality side is covered in Sharing Assets With an AI Studio; here the point is only that the client gets the answer without having to ask.
Whose likeness appears
If an image shows a real person or a digital twin of one, the client needs the consent documents or a clear statement that none are needed because the figure is synthetic. If the figure resembles a real person by accident, say that too. The brand will otherwise learn about it from the person.
Who checks and approves
Name the human reviewers and the checks: product accuracy, anatomy, brand rules, rights review. Clients accept AI work more readily when they can see that a named person looked at every final image.
What to confirm after delivery
The note written at the start describes a plan. At handover, send a confirmation that records what happened:
1. For each final asset, whether AI was used and at which stages.
2. Which systems and partners were involved in the end.
3. What changed from the plan, and why.
4. Where the production record for each asset is kept, and for how long.
This record is what the brand opens when a question comes in months later. Studios that keep one file per asset, as described in What You Receive From an AI Production Studio, can pass most of it on directly.
Contract clauses that carry the duty
Disclosure that lives only in an email is hard to enforce. These clauses give it weight:
Disclosure clause. The agency discloses AI use in the way described above and updates the disclosure when it changes.
Rights warranty. The agency states that it has the rights it needs in inputs and outputs, or says exactly what it cannot warrant. A silent contract leaves the question to argument later.
Allocation of risk. Who answers if a third party claims a violation, and whether the agency's liability for rights claims is capped. Indemnities are negotiated, not standard, and each side's insurer will want to see them.
Client data. No use of the client's material to train systems that serve others, and deletion at the end with written confirmation.
Approval of changes. A new system or partner on the client's work needs the client's consent, or at least prior notice.
Cooperation. If a regulator or platform asks how an image was made, the agency supplies the production record on request.
Flow the same obligations down to the studio. An agency that promises its client full disclosure but receives nothing from its production partner cannot keep the promise.
Disclosure to the public is a different job
Telling the client and labelling the campaign are two separate duties. The first sits in the contract between agency and brand. The second depends on the law that applies where the campaign runs. In the EU, Article 50 of the AI Act addresses deep fakes and is covered in its own article. The agency's contribution is information: the brand can only decide on a label if it knows which images are generated, and how realistic they are.
One US example shows how disclosure can reach a different audience. The US Copyright Office's registration guidance says that AI-generated content that is more than de minimis should be explicitly excluded from the application when a work is registered. A brand that wants to register a campaign work needs exactly the stage-by-stage account described above to complete that form correctly.
A practical order of work
1. Add a disclosure section to the brief or statement of work.
2. Collect the matching information from the studio and freelancers.
3. Agree the contract clauses before the first asset is delivered.
4. Send the confirmation with the final files.
5. Keep the record in the same place as the approvals.
Frequently Asked Questions
Do agencies have to tell clients when they use AI?
There is no single rule that covers every campaign. The duty usually comes from the contract, from the agency's existing duties of care and honesty, and from the fact that the brand carries the legal risk for what it publishes. In practice, an agency that wants to use AI on client work should say so in writing before it starts, and many client contracts and procurement rules now ask for it.
What should an AI disclosure to a client contain?
It should name the stages where AI is used, the deliverables affected, the systems and partners involved, the client material that goes into them, whether any real person's likeness is used, and the people who review and approve the results. A confirmation at delivery should record what actually happened.
Can an agency use client material to train a model?
Only if the client agrees in writing. The safer default is a contract clause that forbids using the client's material, and anything derived from it, to train or improve systems that serve other clients, with deletion at the end of the project.
Who is liable if an AI-generated image infringes someone's rights?
That depends on the contract and on the facts. Without a clause, the brand that published the image is the first target. Agencies and brands therefore negotiate a rights warranty and an indemnity, and they should also check what the studio and its suppliers are willing to stand behind.
Is disclosing AI use to the client the same as labelling it for the public?
No. Disclosure to the client is a private duty between agency and brand. Public labelling depends on the law of the countries where the campaign runs. The client disclosure supplies the facts the brand needs to decide whether a public label is required.
What should the agency keep after the project?
A production record for each final asset: stages where AI was used, systems and partners, source material, rights documents and approvals. Keep it as long as a rights question could still arise, and agree the period with the client.
Planning a campaign with AI imagery or video? Request an AI campaign.
About The Bridges AI
The Bridges AI is a hybrid AI creative agency: a creative studio and production partner for fashion, beauty, luxury and hospitality brands and for the agencies that work with them. It combines booked real talent, real products and human creative direction with AI production, and offers licensed digital twins of real models, only of models under contract. The Bridges AI works in strategic partnership with Most Wanted Models®, the German model agency founded in 1997, which brings nearly 30 years of experience in bookings, buyouts and usage rights.
Written by the editorial team of The Bridges AI. Last updated November 2026.




Comments