How does ChatGPT shopping pick products to show?

What is ChatGPT's new shopping question box?

ChatGPT is testing a shopping question box where shoppers choose a product type and a budget, then press a 'Find' button that turns those choices into a new query. The test surfaced in October 2026, and it closely resembles a recent Google AI Mode experiment. The shopper no longer has to write a long prompt explaining what they want and how much they'll spend. They tap a couple of options and ChatGPT builds the question for them. It's a small change on screen, but it matters a lot to anyone who runs an online store. Once the question is structured, the answer tends to be structured too. The products that come back are the ones whose data clearly matches a category and a price band, and fuzzy listings have far less to match against.

This didn't come out of nowhere. In late November 2025, OpenAI launched a shopping research mode that asks clarifying questions before it recommends anything. It asks about budget, who the item is for and which features matter most. Shoppers can also mark results as 'Not interested' or 'More like this' to steer the search as it runs. The new question box moves the two most decisive of those questions, type and budget, to the very start of the conversation. That shifts the balance. Before, a vague prompt let ChatGPT interpret loosely and cast a wide net. Now the shopper's constraints are stated up front, so a product either fits the type and price range or it doesn't.

For ecommerce owners, my take is simple. This box rewards dull data hygiene far more than clever copywriting. A beautifully written product description won't help if your price is buried in a script, your category name doesn't match what shoppers call the product, or your stock status is missing. Filters act like a sieve, and your product has to make it through before anyone weighs its relevance. That's good news for small and mid-sized stores, because tidy data costs effort rather than ad budget. A store that clearly says 'trail running shoe, $89, in stock' on its pages, markup and feed is easy for ChatGPT to slot into a filtered answer. The rest of this article explains how to become that store.

How do ChatGPT shopping filters decide which products show?

ChatGPT shopping filters pick products by matching the shopper's type and budget against structured product data, then ranking the matches by how relevant they are to the request. OpenAI says product results are chosen independently by ChatGPT, aren't ads and aren't influenced by partnerships. The data behind them includes price, description and availability, and it comes from third-party data providers, merchants and web pages crawled by OpenAI's search crawler. In practice, the type and budget choices work like hard constraints. A product priced at $140 won't appear for a shopper who set a $100 ceiling, however good its reviews are. Among the products that pass, the ones whose descriptions, specs and outside reputation best fit the shopper's intent rise to the top. When several stores sell the same item, price, availability and quality signals help decide which store gets shown.

Merchants can also send product data straight to OpenAI through a product feed. The feed is a regularly refreshed CSV or JSON file listing each item's ID, title, description, page URL, brand, seller, image, availability and price. Recommended extras such as richer media and reviews help with relevance. Shopify stores' catalogs are already connected, and Etsy listings are available too. Other retailers can request direct feed access, but they may land on a waitlist. Feed-based checkout has mostly targeted US shoppers so far. A feed isn't the only way in, because products can still appear through crawled pages and third-party data. It is, however, the one channel where you fully control what ChatGPT reads about each product, and that's exactly the control a type-and-budget filter needs.

Recent tracking data shows how much feeds now matter. One large third-party study of ChatGPT shopping prompts reported that the share of product recommendations tied to merchant feeds jumped from about 8% to about 62% in July 2026. Over the same period, the ten most-referenced merchants grew their share from 22.5% to 41.8%, and the number of distinct merchants cited fell. Treat these figures as observations from tracked prompts, not official numbers, but the direction is hard to ignore. Shopping answers are concentrating around stores with clean, machine-readable catalogs. Crawled pages still matter, though. Descriptions, reviews, comparisons and third-party mentions feed the relevance judgment that ranks products after the filter has done its work. The winning setup is both: a reliable feed plus pages worth citing.

What product page data helps you show up in filtered results?

The product page data that matters most is a visible, machine-readable price with its currency, a clear product type, current availability and the specs that define the item. Everything else comes after those four. Put the price in the page's HTML, not only in a script that loads after the page renders, so that crawlers read the same number a human sees. Add Product structured data with an Offer that states price, currency and stock status, and include brand and a global identifier such as a GTIN where you have one. Name the product the way buyers search for it. 'Women's waterproof hiking boot' gets through a type filter, while 'The Summit Collection No. 4' doesn't. Write the key specs, like size, material, capacity or compatibility, as plain text on the page and not only inside image graphics.

Category pages deserve the same care, because they often decide whether ChatGPT understands what kind of store you are. A category page called 'Footwear' tells a crawler very little. 'Trail running shoes' with a short intro listing the price range, the main use cases and how the models differ tells it a great deal. Make your filter options, such as price bands, sizes and materials, readable as text and not hidden behind interactions that only work after JavaScript runs. Check your robots.txt to confirm you aren't blocking OpenAI's search crawler, which happens more often than you'd think when a store copies an aggressive bot-blocking template. Make sure each variant has its own price and stock status. When a shopper filters to 'under $60', the $55 variant of your product should be able to show up on its own.

Consistency is the part most stores get wrong. If your page says $79, your schema says $85 and your feed still shows last month's $69 sale price, ChatGPT has three conflicting answers and no reason to trust any of them. A mismatch like that can quietly push a product out of a budget-filtered result, or worse, show a price the shopper never finds at checkout. Pick one source of truth, usually your store platform, and make sure the page, the structured data and the feed all pull from it automatically. Mark out-of-stock items accurately instead of leaving them listed as available, because recommending something nobody can buy damages trust for the whole store. None of this is glamorous work. But each fix compounds across every product and every filtered query from then on.

Which content helps you win budget-and-type shopping questions?

The content that wins budget-and-type questions is buyer guides and comparisons built on the same two axes the shopping box uses: what kind of product it is and how much it costs. Think 'best espresso machines under $300', 'carry-on vs checked luggage for business trips' or 'which standing desk to choose for a small apartment'. The Find button turns a shopper's taps into almost exactly that kind of query. A page that answers it directly, with named products, real prices and honest trade-offs, gives ChatGPT both a match and a reason to trust it. This is the core of ecommerce GEO, meaning generative engine optimization for stores. Product pages prove that an item exists and fits the filter, while editorial content explains why it's the right choice for this buyer at this budget.

How you structure these pages matters as much as the topics you pick. Open with a one-sentence answer, such as 'For most home baristas under $300, a single-boiler machine with a built-in grinder is the best value'. Then back it up. Add a comparison table with price, the key specs and who each option suits, because tables are easy for AI systems to parse and quote. Say plainly who each product is wrong for, since that honesty is what makes a guide worth citing. Link every pick to its product page so the recommendation and the data reinforce each other. Since ChatGPT is also experimenting with badges on its product picks, it's worth reading up on how stores can earn ChatGPT's product labels alongside this work.

Here's the contrarian part. Don't churn out forty thin 'best X under $Y' pages with the price swapped in each one. AI answer engines are good at spotting pages that say nothing new, and a cluster of near-duplicates can weaken trust in your whole site. You're better off writing six substantial guides that match the real price bands your customers shop in, each with genuine comparison and first-hand detail. Outside signals count too. Reviews on your own product pages, mentions in independent roundups and community discussion all shape the relevance judgment after filtering. If you want to know how citation-ready your pages are today, a quick GEO score check will show where structure and clarity fall short before you write anything new.

Should you pay for shopping placement or earn it organically?

For ChatGPT's organic product results, you can't pay for placement, so earning it is the only option. OpenAI says those results are selected independently and aren't ads. ChatGPT has been testing a separate, clearly labelled ad system since early 2026, but that is a different slot from the products that appear inside the answer itself. The distinction matters for your budget. An ad shows up only while you're paying for it. A product that earns its place through clean data and helpful content keeps showing up for every matching filtered query, month after month, with no cost per click. If you're weighing both routes, this breakdown of paid ChatGPT ads versus organic citations covers when a paid test is worth running and when it just rents attention you could own.

The economics favor organic for most stores, and the reason is compounding. With paid ads, every visitor costs money, and traffic stops the day the budget stops. With organic growth, the work is front-loaded: fixing your schema, syncing your feed, renaming categories, writing six strong buyer guides. After that, the asset keeps paying off across Google, Bing and AI answers at once, because the same structured pages rank in search results and get cited by assistants. That doesn't make it instant. Crawling, feed processing and building trust take time, and a brand-new store with no reviews will still lag behind established competitors. To put rough numbers on your own situation, an SEO traffic calculator can estimate what steady organic visits would be worth compared with buying the same clicks.

When is organic the wrong fit? If you're launching a seasonal product next week, paid ads are the faster lever, and that's fine. Organic visibility is for stores that plan to be around next year and want a channel that doesn't charge rent. That's the work Artiql automates. It runs your organic marketing on autopilot, covering the product-focused buyer guides, comparison pages and structured content that help your catalog earn placement in Google results and in AI shopping answers like ChatGPT's new Find box. There's no content team to hire and no agency retainer to manage. Pricing is tailored to each business after a short discovery call, because a 50-product boutique and a 5,000-SKU store need very different plans. If you'd like to see what organic growth could look like for your catalog, book a short demo and we'll walk through it together.

Frequently asked questions