Surfaces our work more for you across Google Search and AI answers.
Quick read: Research from AI-visibility firm Profound, tracking 1,757,723 ChatGPT Shopping prompt runs during July 2026, found that the share of product picks sourced from structured product feeds jumped from 8.26% to 61.54% in a single day — July 10, 2026 — coinciding with OpenAI's release of the ChatGPT 5.6 model on July 9 (Search Engine Journal). Of 687 retail customers Profound tracked through that shift, 450 lost at least a third of their ChatGPT Shopping visibility comparing the July 7–9 window to July 10–12, while 67 gained by the same margin. A product feed is a structured data file — not a web page — submitted directly to OpenAI through its Merchant portal at chatgpt.com/merchants after a business verification step.
You sell something. Somebody asks ChatGPT to help them find it. Two months ago, there's a real chance it read your website to answer — your product page, your prices, your photos. Today, for a fast-growing share of those answers, it isn't looking at your website at all. It's reading a file you never knew you were supposed to submit, and if you haven't, you're not in the running, regardless of how good the page is.
That's not a slow drift. Per the numbers above, it happened to a large share of tracked shopping answers in the space of one day.
What flipped on July 10, 2026
ChatGPT Shopping used to work roughly the way you'd expect an AI search tool to work: ask about a product, and it pulls together an answer partly from pages it can read on the open web — much the same retrieval most AI answer engines use for everything else they cite. Profound's tracked data shows that changed abruptly on July 10, 2026. The share of ChatGPT Shopping recommendations sourced from a structured product feed — rather than from crawling a merchant's own site — jumped from roughly 1 in 12 to nearly 2 in 3, practically overnight. The timing lines up with OpenAI shipping its ChatGPT 5.6 models the day before, which is the only dated event with enough weight to explain a shift of that size on that specific day.
A product feed, in this context, is a separate structured file — think a spreadsheet of every product's name, price, availability, and image, formatted to a specification OpenAI publishes — that a business registers to submit through the ChatGPT Merchant portal, independent of the website itself. OpenAI verifies the business, then the feed gets delivered and kept current, separately from anything a visitor would ever see rendered as a page. When ChatGPT answers a shopping question now, for a large and growing share of the time, it's checking that file, not fetching your product page the way it might fetch a blog post to answer a different kind of question. That's a different track entirely from the page-level work covered in our walkthrough of optimizing a site for ChatGPT — that guide is still correct for how ChatGPT reads a page; it just isn't what shopping answers are running on anymore.
Why this is a real problem, not a rounding error
The part that should get a product-based small business's attention isn't the mechanism — it's the visibility swing tied directly to it. Profound's before-and-after comparison, watching the same 687 retail accounts across the exact week the shift happened, found 450 of them lost at least a third of their ChatGPT Shopping visibility in that window. That's roughly two out of every three tracked merchants taking a real hit, in days, for a reason that had nothing to do with their prices, their reviews, or the quality of their site. It had to do with whether a feed existed.
Meanwhile 67 merchants gained visibility by that same margin — the flip side of the same mechanic. Nothing about a losing merchant's product necessarily got worse and nothing about a gaining merchant's product necessarily got better. What moved was whether ChatGPT had a feed to read for that business at that moment, which is a completely different kind of variable than anything Answer Engine Optimization has asked a business to manage before. This is worth being precise about: Profound's numbers describe prompt runs its own monitoring system issued and tracked, not confirmed real-shopper sessions or sales — but the size and abruptness of the swing, and the fact that it lines up exactly with a dated model release, is hard to explain any other way.
This sits apart from everything DBO has written about local service businesses showing up in AI answers, because a service business — a plumber, a groomer, a dentist — doesn't have products in the retail sense, and none of this applies to them directly. It matters specifically for a small business selling physical goods: a boutique, a specialty food shop, a maker selling online, a retailer with an actual product catalog rather than a list of services.
What to check first
Find out if you already run a Google Shopping or Meta product feed. If you advertise products through Google Shopping or Meta, you likely already maintain a structured feed of your catalog for that purpose. That existing file is the head start — adapting a feed you already build is a fraction of the work of starting from nothing.
Register through OpenAI's ChatGPT Merchant portal (chatgpt.com/merchants) and go through business verification. This is a separate registration from anything you've set up for Google or your website — there's no automatic carryover just because a feed exists somewhere else.
Check the current feed specification before building anything. OpenAI has adjusted which file formats it accepts since first publishing the spec, so confirm the up-to-date requirements at the time you build the file rather than working from an old guide.
Ask whoever manages your website or ads whether this is already on their radar. If an agency or a platform like Shopify handles your product listings, this is the kind of infrastructure change worth a direct question rather than an assumption that it's automatically covered.
Re-test your own products in ChatGPT periodically. Ask it to help find something you sell and see whether you show up. It's a blunt check, but right now it's one of the only ways to see your own status from the outside.
Moves that don't get you back into ChatGPT Shopping
Rewriting your product pages. This shift is about a separate data file ChatGPT reads instead of your page, not about how your page itself is written. A better product description doesn't put you in a feed you never submitted.
Assuming your Google Shopping feed automatically counts. It's the fastest starting point for building one, not a substitute for registering and submitting through OpenAI's own process.
Waiting to see if this settles back down. The July 10 shift held through the rest of Profound's tracked window rather than reversing — treat it as the new baseline, not a spike to wait out.
Buying an AI-visibility monitoring tool instead of building the feed. A dashboard that confirms you're missing from ChatGPT Shopping results doesn't put you back into them. The feed is the fix; monitoring only tells you whether the fix worked.
FAQ
What is a ChatGPT product feed?
A structured data file — not a web page — that lists a merchant's products, prices, availability, and images in a format OpenAI specifies. It's submitted through the ChatGPT Merchant portal after business verification, and it's how ChatGPT increasingly sources what it recommends for shopping questions, separately from anything it might read on the merchant's own website.
Why did my products stop showing up in ChatGPT?
Per Profound's tracked data, ChatGPT's shopping results shifted sharply toward feed-sourced products on July 10, 2026, tied to a ChatGPT model update. A business without a submitted product feed is competing for a shrinking share of results still sourced by other means, which is a plausible explanation if your visibility dropped around that window with nothing else on your end changing.
How do I submit a product feed to ChatGPT?
Register through OpenAI's ChatGPT Merchant portal at chatgpt.com/merchants, complete business verification, then build and submit a feed matching OpenAI's current product feed specification. Check the specification at the time you build it, since the accepted formats have changed since the initial announcement.
Does this replace regular SEO or AEO work?
No. This is an addition specific to shopping-style queries for businesses that sell physical products, not a replacement for Search Engine Optimization (SEO) or the broader Answer Engine Optimization (AEO) work a site still needs. A service business with no product catalog doesn't need a product feed at all.
I already run Google Shopping ads. Do I still need to do anything?
Yes. An existing Google Shopping or Meta feed is a genuine head start on the data you'll need, but it doesn't automatically transfer — registering with OpenAI and submitting through its own process is a separate step.
Is this just a ChatGPT thing, or will Google and other AI tools do the same?
This specific shift is documented for ChatGPT. Whether other AI shopping experiences move the same direction isn't something this post claims to know — treat this as what's confirmed for ChatGPT today, and watch for the same pattern elsewhere rather than assuming it.
What if I only sell a handful of products, not a full catalog?
The mechanism doesn't require a large catalog — a feed can list as few products as a business has. The registration and verification steps are the same regardless of catalog size.
Catching a shift like this before it costs a season of visibility is exactly the kind of thing ongoing Care exists for — someone watching the mechanics change, not a business owner finding out from a quiet sales dip months later.
Want a back office for your site?
Builds and Care are by application — scoped and quoted after a short audit, against the actual work, not tiers.
Apply for a build slot →RELATED READS