Don’t get flagged by Amazon
for your AI listing images

The compliance calendar

Amazon's rule: one exact keyword

Amazon translated this legal landscape into a hard technical gate. Effective around 22 July 2026, listing images, A+ content, and product videos containing photorealistic people generated entirely by AI must carry the keyword contains-synthetic-performer in the dc:subject (XMP) field of the media file, before upload.

The scope is narrow, and that matters. Only photorealistic, fully AI‑generated people need the tag. Product‑only images, images of real people, and real people edited with AI tools are exempt. Enforcement is operational: non‑compliant media can be flagged or removed, a listing without a compliant main image can be suppressed from search, and repeated issues risk ASIN removal or seller‑account action.

One nuance sellers get wrong: Amazon's requirement is a plain metadata keyword, not a signed C2PA credential. It needs no certificate and no trust list, which is why this free tool can write it for you honestly. You declare which images show AI people, and the tag travels inside the file.

The EU AI Act: a visible label, not just metadata

The AI Act's transparency obligations (Article 50) apply EU‑wide since 2 August 2026. Two duties matter for product imagery, and they land on different parties:

• Provider marking (Art. 50(2)).Whoever generates the image must mark outputs machine‑readably as artificially generated. The European Commission's draft Code of Practice names C2PA Content Credentials as an example mechanism.
• Deployer disclosure (Art. 50(4)–(5)).If you publish an AI image that could appear authentic, you must disclose that it is AI‑generated via a visible label a person understands without a detection tool: clear, distinguishable, and at first exposure. Metadata alone does not satisfy this.

Penalties reach €15 million or 3% of global annual turnover. In Germany the sharper near‑term risk is civil: Article 50 labelling duties likely qualify as market‑conduct rules under §3a UWG, which means competitors and consumer bodies can issue cease‑and‑desist warnings (Abmahnung) without any regulator involved.

Content Credentials, briefly

C2PA Content Credentials bundle signed statements about an asset (which tool produced it, what was done to it) and cryptographically bind them to a hash of the pixels. Trust depends on the signing certificate chaining to the official C2PA Trust List; that is why this tool verifies existing credentials but never re‑signs your uploads. A re‑signed manifest from a tool that didn't generate the image would vouch for a claim it can't verify, exactly what a provenance signature must not do. Images generated in Scalable are signed at the source, which is where a credential belongs.

The 60‑second seller checklist

1. Does the image contain a photorealistic, fully AI‑generated person? Yes → add contains-synthetic-performer to dc:subject before uploading to Amazon, and add a visible disclosure.
2. Selling into the EU or Germany? The image needs a visible AI disclosure at first exposure, and it must not mislead about the real product (§5 UWG).
3. Advertising into New York? Ensure a clear, conspicuous in‑ad disclosure whenever a synthetic performer appears.
4. Generated by a large AI provider? The machine‑readable mark is the provider's duty. Keep it intact and don't strip metadata on re‑export.
5. Best practice: generate with a tool that embeds provenance at the source, so disclosure travels with the asset.

Sources

European Commission — Transparency obligations under Article 50 AI Act
EU AI Act — Article 50 full text & guide
Bundesnetzagentur — KI‑MIG market‑surveillance announcement
New York — Synthetic Performer Disclosure Law announcement
California — SB 942 / AB 2013 · C2PA — Specification 2.4 · Trust List conformance
Amazon policy coverage — GeekSeller · Forbes · Ecomclips tagging how‑to