A sofa is sold with images. A white background for the product description, a detail of the fabric, a side view, and, above all, scenes in a setting that help the customer imagine it in their home. The problem is that each styling photoshoot is expensive and time-consuming, and a furniture catalog has many variations in color, upholstery, and finish.
Generative AI already allows for the creation of lifestyle scenes from a product photo. For a furniture e-commerce business, this is a real opportunity to scale, especially before the peak season. However, it also brings two risks that should be managed from day one: product loyalty y compliance with platform rules.
What can AI do today in furniture photography
In practical terms, generative AI is used in furniture e-commerce for:
- Setting the mood for a packshot: Place the photographed product on a white background inside a living room, bedroom, or terrace.
- Create scene variations: the same product in different styles (Scandinavian, Japanese, Mediterranean) for different campaigns or markets.
- Adapt formats: Clippings and backgrounds for marketplaces, social media, or banners.
- Generate seasonal scenes, For example, for Christmas or summer, without setting up a new set.
There is some adoption data available, though it comes from vendors. OmniRoom, a furniture rendering tool, published a usage analysis in August 2016 covering 167 companies in 55 countries. According to that report, its clients kept approximately one out of every six generated images, with an average of 6.2 images per product. This data comes from a vendor about their own platform, not an industry benchmark. Even so, it points to something useful: AI produces volume, but human selection is still necessary.
Risk number one: that AI will change the product
A beautiful lifestyle image is useless if the sofa in the photo isn't the sofa you're going to deliver. Image editing techniques can subtly alter what matters most in furniture.
- he color or the tone of the upholstery;
- he finish of wood or metal;
- the proportions: backrest height, arm thickness, legs;
- construction details such as seams, buttons, handles, or grain.
In October 2026, The AI Journal published a test of 96 runs with six image editing models on real products. The number of usable results varied greatly between models, from half to all cases depending on the model. The study was funded by a company that sells access to these models, and the sample size is small, so it should be taken as an indicator rather than a definitive ranking. The lesson is clear nonetheless: It's not enough to just choose a tool; you have to verify each image..
In furniture, a difference in tone or proportion is not just an aesthetic detail. It's an unmet expectation that can lead to a complaint or return.
Google Merchant Center rules for AI-generated images
If you sell on Google Shopping, there are specific rules:
- Required metadata. According to the Official help from Google Merchant Center, All images created with generative AI must include the IPTC DigitalSourceType metadata with the value TrainedAlgorithmicMedia. Google explicitly requests this. do not delete These metadata, and many compression or editing tools, delete them without warning.
- Where they are allowed. AI-powered images can be used in image_link, additional_image_link, and lifestyle_image_link.
- Lifestyle images. The specification of lifestyle_image_link It mentions, as an example, furniture displayed in a room. It requires a minimum of 600x600 px and allows up to 5 images per product.
- Main image. According to the specification of image_link, From January 31, 2027, the minimum size for product images will be 500×500 px; Google recommends 1500×1500 px. Each variant must show the exact color, material, and finish.
- Texts with AI. If titles or descriptions are also generated using AI, they must be submitted using structured_title and structured_description.
A recommended workflow
- Part of a good real photo. AI works best on a clean, well-lit packshot with accurate color. That image is your true reference.
- Define what cannot be changed. Color, material, proportions and key details of each SKU, in writing.
- Generate in batches and select. It assumes that only a portion of the results will be publishable.
- Perform a comparative quality control. Review each image along with the reference: tone, shape, legs, seams and scale with respect to the environment.
- Preserve the metadata. Check that your compression pipeline and your CMS do not remove IPTC.
- Label inside. Record which images are generated, with which tool, and who approved them.
- Measure. Compare conversion, time on page, and returns of products with and without generated scenes.
Catalog images and in-store visualization
AI-powered lifestyle scenes solve part of the problem: they showcase the product in a A nice atmosphere. But the customer doesn't live in that kind of environment. The question that really stops a furniture purchase is another: how does it look? at my place?
That's why the most comprehensive strategy combines two layers of generative visual commerce: AI-generated catalog images and fidelity control, and contextual visualization which places the actual product in the buyer's space photo. Both share the same rule: the product cannot be changed.
Conclusion
AI-powered product photography allows furniture e-commerce businesses to produce more scenes, faster, and at a lower cost. It works when treated as a structured process with rules: real-world reference, quality control, intact metadata, and accurate measurement. Speed without fidelity is costly.
If you work in furniture or home decor e-commerce, follow the DekorAI blog For more content on visual commerce. And if you want your customers to see your actual product in their own space, learn more. Home Try-On? y Request a demo.
References
- Google Merchant Center Help — AI-generated content — https://support.google.com/merchants/answer/14743464?hl=en
- Google Merchant Center Help — Lifestyle image link [lifestyle_image_link] — https://support.google.com/merchants/answer/9103186?hl=en
- Google Merchant Center Help — Image link [image_link] — https://support.google.com/merchants/answer/6324350
- The AI Journal — The pretty image problem: how to test whether AI changed the product you sell (Oct 6, 2026; vendor-funded study) — https://aijourn.com/the-pretty-image-problem-how-to-test-whether-ai-changed-the-product-you-sell/
- OmniRoom — Furniture product photography data 2026 (August 26, 2026; supplier data) — https://www.omniroom.ai/playbook/furniture-product-photography-data-2026
Frequently Asked Questions
Can AI-generated images be used on Google Shopping?
Yes. Google Merchant Center allows them in image_link, additional_image_link, and lifestyle_image_link, as long as they retain the IPTC metadata that identifies them as AI-generated.
What is IPTC Trained Algorithmic Media metadata?
This is a value from the IPTC DigitalSourceType field that indicates the image was created using a trained AI model. Google requires it for AI-generated images and asks that it not be removed.
Can AI change the color or shape of the product?
Yes. Editing models can alter tone, finish, proportions, or details. That's why each image should be checked against a real reference photo.
How many lifestyle images does Google allow per product?
Up to 5 images in lifestyle_image_link, with a minimum of 600×600 px.
Is AI replacing the product photographer?
Not entirely. The actual product photo remains the basis for fidelity; AI multiplies scenes and variations from it.
How to perform quality control of product images using AI?
Compare each image to the reference (color, material, proportions, and details), verify the metadata, and record who approved each image before publishing it.
Español de Chile 
Português do Brasil
English
