AI-generated product images and the law: what marketplace sellers need to know in 2026

Three regulatory clocks tick simultaneously this year — the EU AI Act, the FTC's enforcement guidance, and state-level synthetic-media laws in the United States. For D2C brands and marketplace sellers using AI product photography, August 2, 2026 is the date that matters: the EU AI Act's transparency rules apply from then.

This guide is general information, not legal advice. Check with a lawyer about your own situation.

Most coverage of AI transparency law has focused on deepfakes and political content. Marketplace sellers and D2C operators have been left to read 200-page legal texts to find the two paragraphs that affect them. This is the version of those paragraphs that we wish someone had written for our team last year.

The three rules that affect product images

1. 1EU AI Act Article 50 — providers and deployers of generative AI systems must ensure outputs are 'marked in a machine-readable format and detectable as artificially generated.' These rules apply from August 2, 2026 (Regulation (EU) 2024/1689).
2. 2FTC Section 5 (US) — the long-standing prohibition on 'unfair or deceptive practices' has been clarified to cover synthetic product imagery that materially misrepresents the product being sold.
3. 3New York Synthetic Performer Law and similar US state laws — narrower than the EU rule but enforceable since mid-2026; they apply where a real person's likeness is depicted in a synthetic image without consent.

What 'substantially AI-generated' actually means

Neither the EU AI Act nor the FTC defines a numerical threshold. A practical working test (general information, not legal advice): an image is 'substantially AI-generated' if the AI step contributes content a buyer would consider material to the purchase decision. A removed background or a colour-corrected white is not substantial. A generated lifestyle scene around a real product is. A fully synthesised second angle is. The deciding factor is whether the modification affects 'material characteristics' — fit, colour, finish, surroundings — not technical retouching.

AI metadata standards (C2PA and IPTC), in two paragraphs

The EU AI Act asks for AI output to be 'marked in a machine-readable format' but doesn't name a single format. The most widely adopted open standard for this is C2PA (Coalition for Content Provenance and Authenticity): it writes a signed manifest into the image's metadata describing what was captured, what was AI-generated, and which tool produced the output. A lighter-weight option is the IPTC 'digital source type' field, which tags an image as AI-generated in its embedded metadata.

Crucially, neither is a watermark. Neither appears in the visible image, so the marking doesn't change how the product image looks to the buyer. Check what your tools actually write. AngleForge does not write C2PA manifests today; its Google Shopping export preset embeds IPTC/XMP metadata marking the image as AI-generated (DigitalSourceType = trainedAlgorithmicMedia, which matches Google Merchant Center's requirement to label AI-generated images), and its other export presets don't add AI metadata.

Platform-level rules vs legal rules — how they interact

• Amazon: no explicit AI disclosure rule. Listings must accurately represent the product; an AI-generated lifestyle slot of a real SKU is fine. Hero must be the actual product on white.
• Etsy: structured disclosure ('Made with AI assistance') required where the listing item itself is AI-generated. Lifestyle/illustrative AI imagery of real handmade items does not require per-image disclosure.
• Shopify: no marketplace-wide rule (Shopify is the storefront, not the seller). Sellers remain liable for misrepresentation under their local consumer law.
• Flipkart: requires accurate product representation; AI generation is not separately regulated. Listing-level enforcement is consistent with the EU rule.
• Walmart: AI imagery permitted; the platform reserves the right to demote listings that materially misrepresent the product.

What good disclosure looks like in 2026

For most sellers, full compliance is invisible to the buyer. The recipe: make sure every AI-touched image carries machine-readable AI metadata (an IPTC digital-source-type tag or a C2PA manifest) — check what your tool writes, and add the IPTC tag with a metadata editor such as ExifTool if it writes nothing; use accurate alt text that describes the actual product, not the AI scene; if any single image is fully synthetic — a generated angle of a real product, for example — note that the listing 'uses AI-assisted imagery' in your shop policies or description.

When you must use a visible label

A visible badge or note is required in three narrow situations: (a) the product itself is AI-generated and shipped as such — an AI illustration, an AI-designed printable, a generative 3D model file; (b) the image depicts a real person's likeness; (c) advertising into a jurisdiction with explicit visible-label rules (New York for performer likeness, parts of the EU for political/deepfake content — neither of which typically applies to product photos).

Penalties on the books

EU AI Act penalties top out at €15 million or 3% of global annual turnover, whichever is higher, for the marking obligation breach. FTC Section 5 cases are decided case by case. US state law penalties vary widely. Worth noting: nobody is going to fine a small D2C brand for a missing metadata tag. For most small sellers the practical risk is a rejected or suppressed listing and lost trust, more than a fine.

A 20-minute compliance checklist

1. 1List every AI image tool in your workflow and check what AI metadata each one writes, if any. (AngleForge adds an IPTC AI tag to Google Shopping exports only; it doesn't write C2PA.)
2. 2For every listing currently using fully synthetic imagery, add a one-line note: 'This listing uses AI-assisted product imagery.'
3. 3Update your shop / brand policies page to include a brief AI imagery statement.
4. 4Audit any image that depicts a person — confirm consent or use AI-generated stand-ins explicitly.
5. 5If you advertise into the EU, keep a record of the AI metadata on every active product image — proof of marking compliance.
6. 6Set a recurring quarterly review for new regulation (especially state-level US laws).

What to ignore

There is a cottage industry of 'AI compliance consultants' charging four-figure retainers to scare small sellers into 'AI risk audits.' Most of what they sell is the checklist above plus a PDF cover sheet. If your AI images carry machine-readable AI metadata, your listings accurately represent your products, and your shop policies acknowledge AI-assisted imagery, you have done most of the work. Spend the saved budget on better photographs of the real product — that is what marketplaces and buyers actually care about.