Will saying we used AI hurt sales?

We found no reliable published sales figure either way. The IAB research above shows a gap between how advertisers and younger consumers feel about AI ads. Where the law allows both, test a labeled and an unlabeled version, and watch returns and complaints as well as clicks.

Do I need to tell customers if AI only replaced a background?

Generally no. The EU guidelines treat background replacement for looks as a minor change, and the IAB framework does not ask for a label on routine editing or on AI backgrounds behind a real product photo. The product itself must still look the way it is.

Our agency used AI without telling us. What should we ask for?

Ask which deliverables used AI, which tools and plans, what references went in, and whether any real person’s likeness was used. Then check ownership, labels and releases against those answers. If the work was passed off as human-made, the agency may also have broken its tool’s terms.

Is "AI-assisted" a good label?

It tells people less than "AI-generated". The IAB found "AI-generated" and "Generated with an AI tool" the most consistently understood terms, and the EU Code builds its label around the letters "AI". Use "AI-generated" for images made from prompts.

Should you tell clients and customers you used AI?

Checked September 24, 20265 min read

Short answer

Tell clients every time, in the contract and on each delivery, because ownership, tool terms and labels all depend on it. Tell customers when AI could change what they believe they are seeing: realistic scenes, AI-generated people, or a product shown differently. The EU has required a visible label on deepfakes since August 2, 2026, New York requires one for AI people in ads, and the IAB’s August 2026 framework asks for one on realistic images generated from a prompt.

Two audiences, two duties

With a client, disclosure is about the deal. The client needs to know which parts it will own, which tool terms apply, and who carries the risk. OpenAI’s own Terms of Use forbid users to represent output as human-generated when it was not, so hiding AI use from a client can breach the tool’s terms as well as the client’s trust.

With customers, disclosure is about not misleading them. The law asks for a label in specific cases, and industry groups ask for more.

What to put in the client contract

Settle these points before work starts. The list is a starting point for your lawyer; it is not a finished clause.

AI terms to cover in a statement of work

1. Disclosure: which deliverables use AI, and which tools and models.
2. Ownership: the client receives all rights the agency holds in the work. Both sides accept that parts generated by AI may not be protected by copyright.
3. Tool terms: the agency confirms each tool allows commercial use on the plan used, and says whether an indemnity applies and to whom.
4. Inputs: who supplies product photos, references and brand assets, and a promise that they may be used as AI inputs.
5. People: releases and consents for any real person, digital twin or union performer, covering AI use.
6. Labels: who decides whether an image needs an AI label, and who applies it. In the EU the agency is usually the deployer.
7. Records: the agency keeps and hands over prompts, model names, reference lists and final files.
8. Approvals: the client approves each AI image before use and signs off on product accuracy.

For showing AI routes to a client earlier in the job, see how to present AI concepts to clients. For who owns what, see who owns AI-generated images.

When customers must be told

Where	Rule	What triggers it	Binding?
EU	AI Act Article 50(4)	A realistic AI image of people, products, places or events that would pass as real	Yes, since August 2, 2026
New York	General Business Law § 396-b	An ad showing an AI-generated person who is not a real, recognizable performer	Yes, since June 9, 2026
US federal	FTC Act Section 5 and the Endorsement Guides	An image that would mislead, or AI people presented as real customers	Yes, where the ad would otherwise mislead
UK	CAP Code, enforced by the ASA	No AI label rule. The code applies however an ad was made, and a realistic AI celebrity endorsement is likely to breach it	Yes, for misleading ads
Amazon (all stores)	Seller rule announced July 2026	Listing and A+ images or videos with a photorealistic AI-generated person need the keyword contains-synthetic-performer in their XMP metadata	Yes, on Amazon
US industry	IAB AI Transparency and Disclosure Framework (August 2026)	Images generated from prompts or reference images, unless clearly unrealistic; photoreal AI people in main roles	No, voluntary

Rules and guidance as of September 24, 2026.

The IAB framework also says what does not need a label: routine retouching, color correction, background cleanup, and a real product photo placed on an AI background, as long as the product is shown accurately. The EU guidelines draw a similar line for real products on AI backgrounds. For Meta, TikTok and Google’s own rules, see do you need to label AI images in ads, and for the EU detail see the EU AI Act guide.

What customers think

The IAB and Sonata Insights study "The AI Ad Gap Widens" (January 2026), as reported in the IAB’s framework, found that 82% of advertisers believe consumers feel positive about AI-generated ads, while 45% of Gen Z and Millennial consumers do. Among Gen Z, 34% would call brands that use AI in ads creative and 30% would call them inauthentic. The same study found 83% of advertising executives say their company has used AI in creative work, up from 60% in 2024.

When H&M described its plan for digital twins of 30 real models in March 2025, it told Business of Fashion it would likely watermark those images to show they were AI, partly to be transparent and partly to see how people responded.

Sample disclosure lines

Lines to adapt

On the image, EU Code style: AI
On the image, spelled out: AI-generated
Social caption: Made with AI from photos of our real product.
Product page, under lifestyle images: Lifestyle images on this page were made with AI. Product photos show the item you receive.
Ad with an AI person (New York): The person in this ad was generated with AI.
Ad with an AI "customer": Not an actual customer. Image generated with AI.
Client delivery note: Images 1 to 10 were generated with [model] in [tool] from product photos supplied by [client]. Human work: art direction, selection, retouching, type. Labels applied: [which images, where].

Keep the words short and put them where the viewer looks first. The IAB reports that "AI-generated" and "Generated with an AI tool" are the terms people understand most consistently.

Questions people also ask

Will saying we used AI hurt sales?: We found no reliable published sales figure either way. The IAB research above shows a gap between how advertisers and younger consumers feel about AI ads. Where the law allows both, test a labeled and an unlabeled version, and watch returns and complaints as well as clicks.
Do I need to tell customers if AI only replaced a background?: Generally no. The EU guidelines treat background replacement for looks as a minor change, and the IAB framework does not ask for a label on routine editing or on AI backgrounds behind a real product photo. The product itself must still look the way it is.
Our agency used AI without telling us. What should we ask for?: Ask which deliverables used AI, which tools and plans, what references went in, and whether any real person’s likeness was used. Then check ownership, labels and releases against those answers. If the work was passed off as human-made, the agency may also have broken its tool’s terms.
Is "AI-assisted" a good label?: It tells people less than "AI-generated". The IAB found "AI-generated" and "Generated with an AI tool" the most consistently understood terms, and the EU Code builds its label around the letters "AI". Use "AI-generated" for images made from prompts.

Where Overs fits

Overs exports the whole campaign with a record of the prompt, model and reference pictures behind each photo, and can produce a client review file (one HTML file) that sends the client’s decisions back into Overs. Together they show a client where AI was used in every image.

Free for 40 photos a month. The AI that makes the photos is billed separately, on your own key, with no markup from Overs.

Start free How it works

Sources

Checked on September 24, 2026. Prices, specs and rules change; follow the links for the current versions.