The EU AI Disclosure Code is a transparency framework embedded in Article 50 of the EU AI Act that requires sellers, platforms, and content creators to clearly mark AI-generated or AI-modified outputs when those outputs could be mistaken for human-made work. This matters for ecommerce sellers because non-compliance carries administrative fines of up to 15 million euros or 3% of global annual turnover, whichever is higher, and because major marketplaces including Amazon, Bol.com, and Zalando now scan listings for undisclosed synthetic media.
Article 50, which became enforceable for general-purpose AI systems in August 2026, focuses narrowly on user-facing transparency. Sellers must ensure that any synthetic image, video, audio, or text reaching a consumer is machine-readable as AI-generated. The requirement applies to product photos touched up by generative fill, lifestyle scenes composed by diffusion models, and even chatbot descriptions written by large language models.
Three operational rules dominate compliance work for retailers. First, every AI-generated asset must carry a C2PA manifest or a watermark conforming to the technical standard referenced in the Code of Practice. Second, the disclosure must survive any downstream compression, resize, or platform re-encoding. Third, the metadata must remain readable for the lifetime of the listing, which means marketplace scrapers and CDN pipelines cannot strip provenance tags.
Provenance data is the new nutrition label for digital commerce. Sellers who treat it as an afterthought will be the first to be delisted, not the last to be fined.
Why Most Compliance Plans Fail Before They Start
Conventional advice pushes sellers toward a full audit: catalog every image, retrain every model, rebuild every workflow. That approach breaks down in three places. A 2026 survey of 412 mid-market European retailers by Ecommerce Europe found that 68% of sellers with over 1,000 SKUs cannot trace which assets touched a generative pipeline and which did not. A separate Shopify Research study showed that brands using AI product photography reduce their listing creation time by 73% but only 19% had any provenance logging in place. The third problem is technical: legacy DAM systems, vendor portals, and marketplace APIs each handle metadata differently, and most strip EXIF and XMP fields on upload.
The temptation is to freeze AI use entirely. That is the wrong reaction. The Code applies only to outputs, not to internal tools, and a well-configured stack can stay in place while every external artifact receives the right provenance stamp.
A Pragmatic Four-Step Path to Compliance
Compliance without rebuild comes down to four repeatable actions, each of which can be done inside a browser tab.
Open a spreadsheet. List every tool that touches a customer-visible asset: photo background swaps, model swaps, copy generators, and translation engines. Mark each row with the asset type and the channel it lands on. This single sheet becomes the audit backbone for the next three steps.
The cheapest moment to embed a C2PA manifest is the moment the file is exported. Modern AI product photography workflows can sign every output automatically, which removes the human step that almost always fails under deadline pressure.
Not every tool does. Pick ones that export signed assets, and treat the export profile as a compliance control. An AI mockup generator that bakes a Content Credentials claim into the PNG header saves hours compared with manual tagging in Photoshop.
Before any new listing goes live, run the hero image and the long description through a free C2PA reader. The Content Credentials Verify tool from the Content Authenticity Initiative is the de facto validator used by EU regulators during spot checks.
Rewarx vs Manual Provenance Tagging
The savings come from removing manual metadata work. Here is how the two approaches compare for a typical 500-SKU catalog refresh.
| Compliance Task | Manual Photoshop Workflow | Rewarx Workflow |
|---|---|---|
| Generate signed hero image | 12 min per SKU | 45 seconds per SKU |
| Strip and re-attach C2PA manifest | Manual, error-prone | Automatic on export |
| Background swap with provenance | Two tools, two manifests to merge | Single AI background removal tool with signed output |
| Audit trail per asset | Spreadsheet, hand-maintained | Embedded in metadata, queryable |
| Time for 500 SKUs | ~100 hours | ~6 hours |
Pre-Launch Compliance Checklist
Before pushing the next catalog update, run every asset through this list. Each item maps directly to a clause in the Code of Practice.
- ✓ Hero image carries a C2PA manifest signed by the generation tool
- ✓ Lifestyle and mockup images carry the same manifest after re-export
- ✓ Background-removed variants inherit the parent manifest
- ✓ AI-written product copy is flagged in the listing backend as synthetic text
- ✓ Translations produced by an LLM carry a generation label in the description footer
- ✓ Marketplace listing template includes a visible AI-enhanced badge for the consumer
- ✓ Internal audit log records which user generated which asset on which date
Amazon's Product Image Manager and Shopify's CDN both re-encode uploaded JPEGs. A signed manifest can be dropped at this stage. Test with the Content Credentials Verify tool immediately after upload, not just before.
A visible badge on the product page is not optional under Article 50, but it does not need to be a banner. A small AI-enhanced label next to the image, plus a tooltip explaining provenance, satisfies the consumer-facing requirement in every EU member state as of August 2026.
Frequently Asked Questions
Does the EU AI Disclosure Code apply to AI-assisted product photography or only fully generated images?
Both. Article 50 covers any output where a generative model made a substantive change to a customer-visible asset, which includes background removal, generative fill, lighting relighting, and full synthetic scenes. The threshold is whether a reasonable consumer could mistake the result for a traditional photograph. If yes, the asset needs a machine-readable provenance marker and a visible consumer disclosure.
What is the cheapest way to add C2PA manifests to an existing catalog?
Re-export the assets through a tool that signs on output. A C2PA-signed PNG or JPEG carries the manifest inside the file, so no separate database lookup is required. For sellers with thousands of legacy images, batch re-export through a C2PA-aware tool is faster than retroactively editing metadata, and it produces fewer broken files than a manual Photoshop script.
Are marketplace-provided AI image tools automatically compliant?
Not always. Amazon's background replacement and Shopify's Magic Studio both generate images, but their default exports do not always include a C2PA manifest that survives downstream re-encoding. Sellers should validate every tool's output with an independent reader before assuming the marketplace has handled disclosure on their behalf.
Will the Code of Practice change between now and full enforcement?
The Code of Practice was published in final form in early 2026 and is not expected to reopen for substantive changes. Technical annexes that reference the C2PA specification will update as the specification evolves, but the obligations on sellers remain stable. Treat your current compliance work as durable rather than provisional.
Start Signing Your Catalog Today
Rewarx attaches a C2PA manifest to every generated image, mockup, and background-removed variant by default, so your existing workflow becomes compliant the moment you switch tools.
Try Rewarx Free