Product photos, video and AI generation
A product photo is the main way to show a shopper something they can't hold in their hands. AI generation helps create a background, an angle or an illustration, but it comes with clear rules: in Google Merchant Center such images must be labeled, and the EU has new requirements for labeling synthetic content.
What works on a product page
Baymard Institute (a 2025 benchmark, 155+ US and European sites, material updated in March 2026) found these gaps:
- 37% of sites don't show the product "in scale" (next to a familiar object or a hand); 42% of users try to work out the product's size from the photo.
- 23% of sites have no photo on a model, and for wearable items (clothing, accessories, cosmetics) it's needed most.
Practical conclusions follow: one photo where the size is visible, and for wearable items a photo on a person. Source: Baymard, the state of product page UX. There are no fresh independent figures on the effect of video on conversion.
What is allowed with AI images
Google Merchant Center. Every AI-generated image must keep the embedded IPTC metadata `DigitalSourceType` with the value `trainedAlgorithmicMedia`. The rule covers the main, additional and "lifestyle" images, and the metadata must not be removed (Google Merchant Center Help; the page is dated February 2024 and is a current requirement). If your system compresses images and strips metadata, check this. For AI-generated titles and descriptions, Google requires you to state this separately and flag it in Merchant Center (Google Search Central).
The EU. Article 50 of the EU AI Act came into force on August 2, 2026. Providers of generative systems must label the output in a machine-readable form, and anyone who publishes a deepfake with real people, places or events must disclose it. For systems that were already on the market before August 2, 2026, the labeling deadline is extended to December 2, 2026 (Goodwin, 2026). This is a legal overview, and the main obligation lies with the tool developers. If you sell in the EU, ask a lawyer what this means for you.
What's safe in AI photos and what isn't
Safer | Risky |
|---|---|
Background, shadow, removing clutter, a neutral studio scene | A changed color, shape, texture or composition of the product |
An advertising illustration labeled as such | A model "wearing" an item you never showed |
Straightening the frame, retouching glare | Invented details the product doesn't have |
A shopper compares what they see with what they receive. The difference leads to returns (returns). This is practical advice, not a quote from a study.
Checklist
- A real photo of the product is always there; AI only for the background and additional shots.
- Every AI image carries a label in its metadata, and your system doesn't strip it.
- There is a photo in scale and, for wearables, on a model.
- A person compares the result with the physical product.
- `alt` text describing the product is added.
Common mistakes
- Replacing all photos with generated ones and losing trust (trust in a store).
- Not checking whether the metadata survived compression.
- Showing colors that aren't in stock.
FAQ
Can AI product photos be used in Google Merchant Center? Yes, provided the `trainedAlgorithmicMedia` metadata is preserved. The image must not mislead about the product.
Do AI photos need to be labeled for shoppers? Merchant Center requires a technical label in the metadata. The EU's requirements for visible disclosure concern deepfakes first of all; check them for your case.
How many photos does a product need? There is no single number. Go by what needs to be shown: scale, details, on a model.
Related terms
- Product page
- AI for catalog content
- Google Merchant Center and free product listings
- Product description: how to write it
Sources
- Baymard Institute. The Current State of E-Commerce Product Page UX (updated 18.03.2026) — baymard.com/blog/current-state-ecommerce-product-page-ux
- Google Merchant Center Help. Preserving metadata tags for AI-generated images (February 2024) — support.google.com/merchants/answer/14572008
- Google Search Central. Guidance on using generative AI content — developers.google.com/search/docs/fundamentals/using-gen-ai-content
- Goodwin. EU AI Act transparency obligations now in force (August 2026) — goodwinlaw.com/en/insights/publications/2026/08/alerts-technology-dpc-eu-ai-act-transparency-obligations-now-in-force
Related
Trust signals
Trust signals are elements of a site that help a shopper believe the store is real, honest and safe to pay: reviews, contacts, delivery and return terms, visual signs of protection on the card form, familiar payment methods.
Free shipping threshold
A free shipping threshold is a rule: shipping is free if the order total is at least a set amount (for example, from 1,500 UAH). For the store it is a way to raise the average order value, for the shopper an incentive to add one more item.
Online store catalog structure
Catalog structure is how products are arranged into categories, subcategories, filters and page URLs. It answers two questions: can a customer find what they need in a few clicks, and does a search engine understand which pages matter.
The mobile version of a store
The mobile version of an online store (mobile UX, m-commerce) is how the site looks and works on a smartphone: from the catalog and search to checkout. Mobile traffic in Ukraine is large: by recent data, about 60-78% of online purchases are made from a…