How to Check If Product Listings Appear in AI Shopping Results
A buyer's checklist for testing whether your products show up in ChatGPT shopping and Google AI Mode, what to fix when they don't, and what to ask a tracking vendor before you sign.
Most ecommerce teams discover AI shopping results the same way. Someone asks ChatGPT for a product in their category, sees four competitor products in the cards, and forwards a screenshot to marketing with a question mark. That screenshot is a signal. It isn't a measurement.
This guide gives you three things: a checklist for running the check properly by hand, a fix list for when your products are missing, and the questions to put to a tracking vendor before you pay someone to do the checking for you.
Decide which surfaces you're checking
Two surfaces matter for most merchants right now, and they are controlled by different inputs.
ChatGPT shopping. In March 2026 OpenAI added visual product browsing and side-by-side comparison to ChatGPT, and extended the Agentic Commerce Protocol (ACP) so merchants can share product feeds and promotions for discovery (OpenAI). Buyers complete the purchase on the merchant's own site or app. OpenAI's merchant page says shopping in ChatGPT is live in the U.S. and will expand.
Google AI Mode and Gemini. Google's Universal Commerce Protocol (UCP) lets eligible listings in AI Mode and in Gemini show a Buy button, with checkout on Google's surface while you remain the merchant of record. It starts in Merchant Center: your account needs to be in good standing with products approved for free listings (Merchant Center Help). That help article covers products eligible in the United States, Canada, and Australia, and the checkout feature is open to select merchants for now.
Write down which surface you care about before you start. A US-only Shopify brand and a Canadian retailer running Merchant Center feeds will end up with different checklists.
The manual spot-check checklist
- Write 10 to 20 buyer prompts that carry real constraints. "Best running shoes" is too loose. "Waterproof trail running shoes under $150 for wide feet" is what a buyer actually types, and it forces the assistant to choose specific products.
- Add a few purchase prompts that name your products, such as "where can I buy [product] online". These test whether a known product surfaces at all.
- Run each prompt logged out and again logged in, and label every result with the state you used. A signed-in account carries its own history and settings, so keep the two sets apart.
- Note where you're testing from. ChatGPT shopping is a U.S. surface according to OpenAI, and UCP checkout applies to the US, Canada, and Australia. A team in Amsterdam on a VPN is not seeing what a buyer in Ohio sees, so treat those results as rough.
- Record what appeared: whether product cards showed at all, which products, your position, the merchant named, and the price displayed.
- Run the same prompts again on a different day before drawing conclusions.
Why the spreadsheet stops working
Run one prompt twice and you'll often get different cards. Answers vary from run to run, and they shift with who is asking and from where. A single check tells you what one person saw at one moment.
The arithmetic gets ugly fast. Twenty prompts, two login states, and three countries is 120 checks, and you need to repeat that weekly to see a trend. Most teams quietly stop around week two, which is exactly when the data would start to mean something.
If your products are missing
Work through these in order. The early items are cheap and catch most problems.
- Feed or crawl. OpenAI says you don't strictly need a feed if ChatGPT already crawls your site, but a feed gives you more control and accuracy. Shopify merchants are covered through Shopify Catalog with no extra work. Others can share feeds through ACP, including via providers such as Salesforce and Stripe.
- Stable IDs and variants. The ACP product feed spec requires a stable
idfor each product and at least one variant, and every variant needs its own stableidandtitle. Price, images, GTINs, and URLs are optional but improve how the product displays. If your PIM regenerates IDs on every export, fix that first. - Merchant Center status. For Google's UCP checkout you need approved free listings, a defined return policy, and at least one customer support contact (website, email, or phone). Some product types are excluded from checkout outright, including subscriptions, pre-orders, used or refurbished items, and age-restricted goods.
- Crawler access. OpenAI's publisher FAQ says to allow OAI-SearchBot if you want to appear in ChatGPT search, and its bot documentation notes robots.txt changes take about 24 hours to apply. Google's Storebot-Google crawls for Google Shopping surfaces (Google). Check robots.txt and your WAF rules for both.
- Price and availability that agree. If the feed says in stock at $120 and the product page says sold out at $135, you've handed the assistant two different answers. Pick one source of truth and make everything else read from it.
Questions to ask a tracking vendor
Put these in the demo call, and ask for the answers in writing.
- Which shopping surfaces do you extract product results from? Plenty of tools track text mentions and label the report "shopping".
- Can I load my own product list and see each product's position over time?
- Do you show which merchant domains the assistant sends buyers to?
- Are sponsored placements reported separately from organic product results?
- Which locations can prompts run from: country only, or state and city too?
- Can I see whether AI crawlers fetch my product pages, and whether AI-referred visits convert?
- Which plan includes shopping tracking, and is it an add-on?
How the shortlist answers
Promptwatch covers the most of that list for ChatGPT. Its ChatGPT Shopping tracking lets you load tracked products and follow their positions over time, and it reports the top products and top merchant domains appearing for your prompts. Ads Radar tracks sponsored placements separately. Around that sit prompt tracking with personas, country tracking on every paid plan (state and city from Business), Agent Analytics crawler logs, and visitor analytics with conversion tracking. Be precise about scope: shopping tracking is ChatGPT Shopping. Promptwatch also tracks your prompts in Google AI Overviews and AI Mode, but it does not track AI Mode shopping cards or Gemini product listings. Plans start with Essential at $95/mo, but ChatGPT Shopping tracking and Ads Radar are on Business at $579/mo, so that's the plan to buy for shopping. Our ChatGPT Shopping setup guide walks through loading a catalog with the right IDs.
Wildcard is built for retailers. It syncs your catalog from Shopify, BigCommerce, Salesforce, or a PIM, and its tracking covers Walmart Sparky and Alexa+ Shopping alongside ChatGPT, Gemini, Google AI Overviews, Perplexity, and Claude, with Amazon Rufus in beta. It attributes revenue per AI surface. Basic is $99/mo with weekly refresh and 20 products, and it doesn't publish crawler logs.
Trackbase is an EU-hosted tracker whose ChatGPT Shopping tracking records product positions, displayed prices, and tags, with change alerts. Shopping is an add-on on its Professional plan, and it includes crawler analytics through Cloudflare or a WordPress plugin.
Who should pick which
- A US merchant who wants ChatGPT product positions and the reason behind them (crawls, citations, converting visits): Promptwatch.
- A retailer whose buyers also shop through Walmart or Amazon assistants: trial Wildcard alongside.
- A team that needs EU hosting and only wants shopping positions as one module: Trackbase.
Run the manual checklist once either way. It tells you which questions matter for your catalog, and it gives you a baseline to hold any vendor's first report against. When you're ready to automate the ChatGPT side, start on Business at promptwatch.com.