Online stores jumped on AI early, and it’s easy to see why. Descriptions for hundreds of SKUs, lifestyle shots without a photo shoot, cleaner backgrounds, size charts, FAQ answers. AI handles all of it now. So if you sell to customers in the EU, what does the AI Act’s transparency rule mean for your catalog?
Short answer: a lot less than you might fear, with one big exception. Let’s go through the usual ways stores use AI and see which ones actually need a label.
A tool, not legal advice. This guide explains public documents in plain language. It doesn’t cover consumer protection or advertising law, which have their own rules on misleading claims. If it matters for your business, ask a lawyer.
The two rules that matter for a store
Article 50 has four transparency duties. For a typical store catalog, two of them count:
- Deepfakes (first part of Article 50(4)). AI-generated or edited images, audio or video that look like real things and could pass as genuine have to be disclosed.
- Public-interest text (second part of Article 50(4)). AI text published to inform people on topics like health or consumer safety has to be disclosed, unless a person reviewed it and takes editorial responsibility.
Got a chat widget in your store? Then the chatbot rule applies too. See Chatbots on your website.
Product descriptions: usually fine
The European Commission is pretty direct here. One of its examples of text that’s not in scope:
AI-manipulated text that is part of a company’s advertisement or product descriptions (not including any claims related to e.g. health, consumer safety or sustainability).
So an AI-written description for a lamp, a T-shirt or a patio chair normally falls outside the text rule. No Article 50 label needed.
The catch: health, safety and green claims
Notice the part in brackets. Claims about health, consumer safety or sustainability are carved out. These are the descriptions to look at twice:
- “Supports healthy joints” on a supplement.
- “Safe for babies from 6 months” on a toy.
- “Made from 100% recycled materials” or “carbon-neutral shipping”.
- “Hypoallergenic” or “dermatologically tested” on cosmetics.
If AI wrote or rewrote copy like that, you’ve got two options. Label it. Or have someone who knows the product check the substance, including the facts behind the claim, before it goes live, and make sure your store’s legal info says who’s responsible. A spell-check or a glance doesn’t count. And if anyone runs the text through AI again after that review, the review no longer counts either.
Claims like these are also covered by other EU rules, such as the ones on misleading green claims and health claims for food. Those apply whether AI was involved or not, so a human check is smart anyway.
Your store blog and buying guides
Lots of stores also run a blog or buying guides. Those are treated like any other published text. “How to choose a sofa” is unlikely to be in scope. “What the new EU toy safety rules mean for your kids” is more likely. See AI-written blog posts and articles.
Product images: the one to watch
Images are where a store’s real responsibility sits. The guidelines name a product example in their list of deepfakes:
An AI-generated image of a product in advertisement or packaging that can affect the audience’s perception and mislead as to the actual product appearance, characteristics or use (e.g. making the product appear not identical to the real product, more appealing or with improved quality than in real life).
They also list edits that are likely to have only a minor effect and usually don’t make a product image a deepfake:
- AI color correction
- extending a background
- adjusting or swapping a background just for looks
- arranging existing products together
- resizing
And one of their “not a deepfake” examples: a real product in an ad against an AI-generated background, as long as the ad doesn’t mislead about the product itself.
Sorting your product images
| What you did | Deepfake? |
|---|---|
| Real photo, background removed and swapped for plain white | No |
| Real photo, dropped into an AI-generated living room | Usually no, if the product looks like it really does |
| Color corrected, resized, sharpened | No |
| A few real products arranged together in one shot | Usually no |
| AI “improved” the fabric or stitching so it looks finer than it is | Yes, it can mislead about quality |
| Fully AI-generated product image, not a photo of the real item | Often yes, if it can differ from the real thing |
| AI model wearing your real garment, shown realistically | Often yes. The person looks real and the fit may not match reality |
| Stylized illustration or an obvious 3D render | Usually no |
The test never changes: could a shopper be misled about what they’ll actually get? If yes, label it.
Where the labels go
The Act wants the disclosure to be clear and easy to spot, no later than when someone first sees the content.
- Images: on the image or right next to it, visible without clicking. The Code of Practice uses the top-right corner of an image as its example. Don’t forget the gallery. If the AI image is the third thumbnail, the label has to be there too, not just on the main photo.
- Text: at the start, like at the top of the description.
A note in your footer or your terms comes far too late.
What about your existing catalog?
The guidelines say deepfakes generated before 2 August 2026 don’t need labels after the fact, though adding them is encouraged. In-scope text that was generated and published before that date doesn’t need one either. Text generated before 2 August 2026 but published on or after it does.
For a store, that means you can focus on what you add or change from here on, and catch older images and claims next time you update those products.
How AuthorLabel works with WooCommerce
AuthorLabel treats WooCommerce products the same way it treats posts.
Label products and images. The “AI transparency” panel shows up on the product edit screen and in the media library. Mark the description and each image AI-generated, AI-assisted, reviewed by a human editor or not AI, and note the tool you used.
Labels where shoppers actually look. A labeled image gets a short caption with a small “AI” mark, like “Image edited with AI”, right inside the WooCommerce product gallery too. A labeled description gets a one-line notice above it. Every labeled item also gets machine-readable markers: Schema.org JSON-LD, a meta tag and a data-ai-generated attribute. Your product data and image files stay untouched.
Detection on upload. When you upload an image, AuthorLabel reads its C2PA, IPTC Digital Source Type and generator EXIF/XMP fields. If the file says an AI tool made it, the plugin suggests a label. It never guesses from how an image looks or how text reads.
Bulk labeling for big catalogs. Filter by post type, date, author or tool, and label up to 200 products at once. Generated a batch of descriptions with one tool last week? Label them all in one step. Every bulk action can be undone.
A record of what you did. The action log shows who labeled what and when. The dashboard widget gives you a readiness score from 0 to 100 and what’s left to do.
Plays nice with caching. When you change a label, AuthorLabel tells common caching plugins like LiteSpeed Cache, WP Rocket and W3 Total Cache to refresh the affected pages, so shoppers see the update right away.
If your store generates content with AI plugins, AuthorLabel Pro can label products automatically the moment they’re generated, and export PDF or CSV reports of your labels.
The short version
- Regular product descriptions are outside the Article 50 text rule.
- Health, safety and green claims written by AI need a label or a real human review.
- Product images that could mislead about the real product are deepfakes and need a visible label.
- Cosmetic edits like backgrounds, color correction and resizing usually don’t.
- Label on the image and at the top of the text, gallery included.
Sources
- Regulation (EU) 2024/1689 (AI Act), Article 3(60) and Article 50(4) and (5).
- European Commission, Guidelines on the transparency obligations under Article 50 of the AI Act, C(2026) 7702 final, Sections 6.1, 6.2 and 8.4. digital-strategy.ec.europa.eu
- Code of Practice on Transparency of AI-Generated Content, deployer section, Sub-measure 1.2.2. digital-strategy.ec.europa.eu
The Commission’s guidelines aren’t legally binding, and the Code of Practice is voluntary.
Last reviewed: November 2026. AuthorLabel is a tool, not legal advice.