Short answer: over the past month two major players have shown the same thing in different ways. Meta introduced an AI agent that browses shops and can buy, and OpenAI announced a visual ad format inside ChatGPT. If AI shopping agents increasingly take part in deciding what gets bought, your product is no longer judged only by a person on a page. It is also judged by what the assistant can read and show. That means, first of all, the product photography and the product data, whether you sell furniture, interiors or homeware.
What has actually been announced
Meta, 23 September 2026. At its Connect conference Meta presented new capabilities for Muse, its personal AI agent, which launched at the start of September. According to TechCrunch’s overview, Muse can browse Shopify’s catalogue and buy through Link by Stripe and Shop Pay, and PayPal support has been added. Retailers named include Best Buy, Gap, Sephora, Walmart and Wayfair, and Expedia is connected too (TechCrunch, 23 Sept 2026). Not everything is available yet: smart glasses, an avatar and the agent’s own email address are listed as still to come. TechCrunch does not say which countries the service works in, and the retailers named are mostly American. For sellers outside the US, then, we do not know when or in what form it will arrive.
OpenAI, 5 October 2026. OpenAI announced a new visual ad format in ChatGPT and expanded measurement. The format uses images that show inspiration for a product, how it is used, or the experience it enables. Testing starts this month in the US, initially during image generation for Free and Go users; the ads are labelled and kept separate from the image being created (OpenAI, 5 Oct 2026). For measuring results OpenAI names outside partners, among them AppsFlyer, Triple Whale and Adjust (Search Engine Land).
What this is not: evidence that most European shoppers will be ordering through an assistant tomorrow. These are early releases and tests, mostly in the US. But the direction of agentic commerce is consistent: the buying decision is moving into the conversation.
What changes for sellers of furniture and homeware
When a shopper opens your site they see everything at once: photographs, typography, the atmosphere of the brand. An assistant sees differently. It takes names, descriptions, prices, availability and images, and shows the shopper a summary. That is our own reading of these announcements, not a claim by Meta or OpenAI, but it has three practical consequences:
- Product data becomes part of the shop window. If the same product is named and described differently on your webshop, in your catalogue, in your product feed and on social media, it is harder for an assistant to tell what is what.
- The photograph has to work at small sizes. A reduced image in a card or inside an answer still needs to show the product: a clear shape, clean space around it, a consistent style across the whole range.
- Brand remains the reason someone picks you. If an assistant offers five similar products, a recognisable seller with well-kept materials has the edge over an anonymous one.
Five steps a retailer can take this week
- Compare three places. Pick your five best sellers and compare the name, description, dimensions and price on your webshop, in your catalogue and on social media. Fix the differences.
- Look at the hero image at thumbnail size. Shrink it to the size of a small icon. Can you still see what is being sold? If not, that is the first frame to reshoot.
- Add text descriptions to images. Every product photograph needs clear alt text. It helps accessibility and search, and it helps assistants too.
- Track your traffic sources. In your analytics, check whether any traffic arrives from ChatGPT and similar assistants. If none does, that is also information.
- Set a rule for yourself. Write one sentence: who on your team confirms an order that came through an agent, and what gets checked before dispatch.
What this means for a studio like ours
Our job is to make a product and a brand look valuable and sell: photography, catalogue, web and campaign as one whole. For example, we photographed the webshop range of a homeware retailer, and a range like that demands a consistent style and data that match. It is a sound foundation whether the material is read by a person or by a machine.
A photograph that sells to a person is probably also one that an assistant will present better. That is why we do not recommend chasing every new tool, but building a foundation that holds in each of them.
Next step
If you run a webshop or a catalogue and wonder whether your range is ready for a world where a growing share of choices passes through an assistant, we are glad to review it with you: photographs, product names, structure, and how your brand looks at small size. Tell us what you sell and we will suggest a first step: get in touch.