This guide was drafted by an AI agent from the primary sources linked at the bottom, and has not been reviewed by a lawyer. It is general information, not legal advice. The original is at https://mdodeja.github.io/eu-ai-act-article-50-guide/guides/ai-product-descriptions-article-50.html.
Short answer. Usually not. The Article 50(4) text duty covers text published to inform the public on matters of public interest, and an ordinary product description is not that. The machine-readable marking duty in 50(2) sits with the provider of the generation tool. The exception is images: a realistic AI product photo of a real item can meet the Act's definition of a deep fake, and that needs a visible disclosure.
The short answer for shop copy
For most online shops, writing product descriptions with an AI tool does not create an Article 50 labelling duty for the shop or for the agency writing the copy. Three paragraphs of Article 50 could be relevant, and each one usually misses product text for a different reason.
| Paragraph | Who it binds | Does it reach a product description? |
|---|---|---|
| 50(1) | Providers of AI systems that interact directly with people | No. A description on a page is not an interaction. A shopping assistant that chats with customers is, and is covered separately. |
| 50(2) | Providers of AI systems that generate synthetic text, images, audio or video | It binds the company that makes the generation tool, not the shop that uses it. |
| 50(4), text | Deployers publishing AI text "with the purpose of informing the public on matters of public interest" | Generally no. Selling a kettle is not informing the public on a matter of public interest. |
| 50(4), deep fakes | Deployers of AI that generates or manipulates image, audio or video content constituting a deep fake | Possibly, for realistic AI product images. See below. |
Why the text duty usually does not apply
The second half of Article 50(4) is narrow. It covers AI-generated or manipulated text "published with the purpose of informing the public on matters of public interest". The Act does not define "matters of public interest", but the wording points at news, commentary and civic information, not at sales copy. Product titles, bullet points, category intros and size guides are written to sell, so they generally fall outside it.
Some shop content sits closer to the line. A buying guide that reads like consumer journalism, a blog post on product safety recalls, or a page about a health or environmental claim could be read as public-interest text. If a piece could be taken either way, treat it as covered. Even then, the same paragraph exempts text that has gone through human review or editorial control where a person or company holds editorial responsibility for publishing it. The guide on labelling AI-written articles explains how to set that up.
The marking duty belongs to the tool's provider
Article 50(2) requires providers of AI systems that generate synthetic text, images, audio or video to mark the output in a machine-readable format so it can be detected as AI-generated. That is a duty on whoever develops the system and places it on the market under their own name, such as the company behind the writing tool or the image model.
Two practical points for agencies and shops:
- Do not strip the marks. If a tool embeds metadata or a watermark in its images, keep your export and upload pipeline from removing it where you can.
- Check whether you have become a provider. An agency that builds its own description generator and supplies it to clients under the agency's name may count as a provider of that system. Who the provider is when a model sits underneath someone else's product is not settled; the marking guide covers what is known.
The Digital Omnibus on AI (Regulation (EU) 2026/1744, Article 111(4)) gives generative systems already on the market before 2 August 2026 until 2 December 2026 to meet the 50(2) marking duty.
The catch: AI product photos
The first half of Article 50(4) applies to image, audio and video content, not text. Deployers of an AI system that generates or manipulates a deep fake have to disclose that the content is artificially generated or manipulated.
Article 3(60) defines a deep fake as AI-generated or manipulated image, audio or video content "that resembles existing persons, objects, places, entities or events and would falsely appear to a person to be authentic or truthful". The definition names objects, so it is not limited to faces. A photorealistic AI image that shows a real product, in a setting it was never photographed in or with features it does not have, can meet that definition if a shopper would take it for a real photo.
How that works out in practice:
- Likely covered: a photorealistic AI-generated photo of the actual product presented as a real photo; an AI model or person wearing the item; AI edits that change how the product looks.
- Likely not covered: obvious illustrations, clearly stylised renders, and plain background removal or colour correction that does not substantially change the image.
- What to do: put a short visible note next to the image, such as "AI-generated image", or in the gallery caption. The disclosure has to be clear and given at the latest when the shopper first sees the image (Article 50(5)).
Where the line between an edit and a manipulation falls is not spelled out in the Act. The Commission's guidelines on Article 50 address scope, so read them for borderline images.
Other rules still apply
Article 50(6) says the article is without prejudice to other transparency duties in EU and national law. For product pages, the bigger risk is usually consumer protection law: a description or image that misleads shoppers about the product is a problem whoever or whatever wrote it. AI tools invent specifications, materials and certifications, so a human should check every factual claim before it goes live. Marketplaces such as Amazon or Etsy may also have their own policies on AI images; check the platform's current rules.
A checklist for shop copy
- Product descriptions and titles: no Article 50 label needed in the usual case. Have a person check facts.
- Buying guides, blog posts and anything news-like: either label as AI-generated or put a named editor in charge and keep a record.
- AI product photos and AI models: add a visible "AI-generated image" note where a shopper could take them for real.
- Keep the tool's metadata or watermarks on exported images.
- On-site AI shopping assistants: those talk to customers, so the chatbot rule applies. See the disclosure wording guide.
Sources
- Regulation (EU) 2024/1689 on EUR-Lex, the official text: Article 3(3) (provider), Article 3(60) (deep fake) and Article 50.
- Article 50 text and Article 3 definitions at artificialintelligenceact.eu.
- European Commission: guidelines on transparency obligations for providers and deployers of AI systems, published 20 July 2026.
- European Commission: Code of Practice on Transparency of AI-generated Content.
