E-commerce imagery spans a spectrum, and the Article 50(4) labelling duty doesn't reach all of it equally. The duty covers published AI imagery that resembles real persons, places, objects, or events and would falsely appear authentic, with no requirement of deceptive intent.
Clearest case first: AI-generated humans. A photorealistic AI model wearing your clothing, or an AI "customer" using your product, reads as a photograph of a real person and sits squarely in the duty. Label it with the harmonised "AI" icon in a fixed position, at every exposure, per the June 2026 Code of Practice.
AI-staged lifestyle scenes, such as your real product composited into a generated living room or a generated beach, follow the same logic: a real-seeming scene a shopper would take for a photo. Treated honestly, these need the label too.
Plain product renders are the genuinely closer call. A clean AI render of your actual product on a white background depicts a real object accurately, the way traditional 3D product renders long have. Whether that "falsely appears authentic" in any way that matters is not clearly settled by the draft guidance, and we won't pretend it is. The pragmatic position: if the render is faithful to the physical product, the Article 50 risk is low, but the consumer-law risk is the one to watch, because an AI render that flatters the product beyond reality can be a misleading commercial practice entirely independent of the AI Act. If the render deviates from the real item, fix the render, not the label.
One more layer: if you offer sellers or customers a tool that generates such imagery under your brand, you're also a provider of a generative system with marking duties under Article 50(2), a separate analysis from the labelling of what you publish yourself.
Not sure how this applies to your specific setup? The free 2-minute check tells you which of the four Article 50 duty areas likely apply to your company. No email required.
Take the free check Generate disclosure text