How to Label AI-Generated Images: Step-by-Step Guide for Retailers
How to Label AI-Generated Images: Step-by-Step Guide for Retailers
From August 2, 2026, the EU AI Act requires photorealistic AI images to be labeled. What does that mean for your product images in practice, and how do you implement the requirement with as little effort as possible?
This guide walks you through the implementation step by step: from taking inventory of your images to choosing a labeling format to checking every channel. For the legal background (copyright, liability, misleading advertising), see our in-depth legal guide for AI product images.
Note: This article provides general legal information and does not constitute individual legal advice.
Quick Answers
Do I have to label AI images?
From August 2, 2026, yes, if they are photorealistic and could be perceived as real (EU AI Act Art. 50). That applies to most AI-generated product and lifestyle images in eCommerce.
How do I label correctly?
Two layers: machine-readable inside the image file (the AI provider’s job) and visible to the viewer (your job as the retailer). For the visible layer, a notice directly on the image is the safest option, for example a badge reading “AI-generated”.
What happens in case of violations?
Violations of the transparency obligations can result in fines, and competitor warning letters cannot be ruled out. Setting up the process early avoids both.
Who has to label: provider or retailer?
The AI Act splits the obligation between two roles:
| Role | Obligation | Implementation |
|---|---|---|
| Provider (the AI system) | Machine-readable labeling in the image file | Watermarks and metadata, e.g. SynthID or C2PA |
| Deployer (you as the retailer) | Visible notice for the viewer | Badge or watermark on the image, notice near the image |
Reputable providers ship the machine-readable layer with their images. Your actual task is the visible labeling, and that is what the following steps cover.
Step by step: implementing labeling in your shop
Step 1: Take inventory
Get an overview of which of your images are AI-generated. Document it internally, for example via a field in your PIM or a simple marker in the file name. Important: images from suppliers and manufacturers count too. Actively ask whether AI was involved and add a disclosure obligation to your procurement terms.
Step 2: Choose a labeling format
For visible labeling, you have two basic options:
- On the image itself: A badge or watermark with a notice like “AI-generated” is applied directly to the image. Advantage: the label travels with the image to every channel, whether shop, marketplace, or social media.
- Near the image: A caption, a notice in the product description, or a note on an info page. Advantage: the image stays untouched. Disadvantage: on channels you do not control, the notice gets lost.
The most robust solution is labeling on the image, optionally complemented by a notice near it.
Step 3: Apply the visible notice to the image
Done manually, this step means image editing for every single photo. In showcase it is automated: activate the built-in AI badge once via checkbox, and an “AI-generated” notice is stamped onto every download and marketplace export. The customizable badge designer lets you match the badge to your brand, and your originals stay unchanged.
Whatever tool you use, pay attention to three things:
- Legibility: The notice must be recognizable at the size the image is actually displayed, including on mobile.
- Clear wording: “AI-generated” or “Created with AI” is understood by everyone. Cryptic abbreviations are riskier.
- Consistency: One consistent badge across all images looks professional and signals a clean process.
Step 4: Check the machine-readable layer
Ask your AI provider how technical labeling is implemented. With showcase, the image models in use embed machine-readable watermarks (e.g. SynthID), and complementary metadata standards like C2PA are being implemented as part of product development. If you post-process images, make sure metadata is not lost on export.
Step 5: Go through every channel
Review your publishing channels one by one:
- Your own shop: Check whether your shop system offers fields for image notices, and add a notice to the product description if needed.
- Marketplaces: Dedicated labeling fields for AI images are still the exception. If the image carries the notice itself, you are independent of what each marketplace offers technically.
- Social media: Many platforms now offer their own AI labeling options at upload. Use them in addition to the on-image notice.
Step 6: Document the process
Record who on the team is responsible for labeling, which images are affected, and how new images enter the process. That way you are also prepared for questions from marketplaces, customers, or competitors.
Frequently Asked Questions (FAQ)
When do I have to start labeling AI images?
From August 2, 2026, the labeling requirement under EU AI Act Art. 50 applies to photorealistic AI images that could be perceived as real. Until then, labeling is voluntary, but many platforms already ask for it and it builds trust with your customers.
What should the labeling notice look like?
The law does not prescribe an exact wording. Short, clear notices like “AI-generated” or “Created with AI” work well, clearly legible directly on the image or right next to it. What matters is that an average viewer notices and understands the label.
Is a notice in the product description enough?
This has not been conclusively settled legally. The safest option is a notice directly on the image, for example as a badge or watermark, because the image then stays labeled on every channel: your shop, marketplaces, and social media. An additional notice near the image never hurts.
Do I have to label cutouts and color variants too?
Pure AI background removal without new image content is generally considered not subject to labeling. AI color variants are a gray area: if the result looks photorealistic and viewers cannot recognize the change, the precautionary principle favors labeling.
Who is liable if supplier images are not labeled?
As the deployer, from August 2026 you must label images when you know or should know that product images are AI-generated. It does not matter who created them. Proactively request information about AI use from suppliers and add a disclosure obligation to your procurement terms.
Conclusion: label early, relax in August 2026
The labeling requirement is no reason to give up AI product images. It is a process topic: identify images, apply the notice to the image, check your channels, document everything. Set it up now and there is nothing left to do by the deadline, while you already score with transparency today.
The biggest effort saver is labeling directly on the image when it happens automatically. See how showcase solves this on the AI labeling page.
Further reading:
- AI Product Images eCommerce: Legal Guide 2026
- AI Product Photography for Home & Living: Photorealistic Lifestyle Images Without a Photoshoot
showcase is an AI platform for Home & Living product photography. Simple cutouts turn into photorealistic lifestyle images, color variants, and room scenes in seconds, fully labeled with the built-in AI badge on request. Create your first images for free today.
About the author
Author
Tim Hoffmann
Chief Product Officer, getshowcase.ai
Tim Hoffmann leads the product strategy for the AI image studio at showcase (getshowcase.ai). He brings years of e-commerce experience in product data, marketplace integrations, and visual content creation. His focus: helping Home & Living retailers turn product cutouts into photorealistic lifestyle images and room scenes in minutes - without expensive shoots, with measurably better conversion. Tim shares practical strategies for product images that perform on marketplaces and in your own shop.
