How to Check If Product Listings Appear in AI Shopping Results

Two AI surfaces put product listings in front of buyers today: ChatGPT shopping, and Google's AI Mode and Gemini. Different inputs control each one. Check them separately or you will misread both.

Six steps below. The first four you can do this afternoon with a spreadsheet. Steps five and six are where the work actually gets done.

The two surfaces

Surface	Where it applies	What controls inclusion	Crawler to check
ChatGPT shopping	Live in the U.S.; OpenAI says it will expand	Your crawlable site, or a product feed via ACP (Shopify merchants covered by Shopify Catalog)	OAI-SearchBot
Google AI Mode and Gemini (UCP checkout)	Products eligible in the US, Canada, Australia; select merchants	Merchant Center in good standing, products approved for free listings, native_commerce(checkout_eligibility) set to true	Storebot-Google

Sources: OpenAI's merchant page and March 2026 announcement; Google's UCP help article and native_commerce attribute. On ChatGPT, purchases complete on the merchant's own site. On Google's UCP checkout, the buyer pays on Google's surface with Google Pay and you remain merchant of record.

Step 1: write 15 prompts with constraints

Generic prompts return generic answers. Constraints force the assistant to pick products, which is what you want to test.

Take an espresso machine brand. Ten category prompts:

• "espresso machine under $500 with a built-in grinder"
• "quietest espresso machine for a small apartment"
• "best espresso machine for a beginner who drinks oat milk lattes"

and seven more like them, each with a price, a use, or a buyer attached. Then five purchase prompts that name your products: "where to buy [model] online". Fifteen total.

Step 2: fix your test conditions

Hold these steady and write them down.

• Login state. Run every prompt logged out, then logged in. Label each row. A signed-in account carries its own history, so don't mix the two sets.
• Location. ChatGPT shopping is a U.S. surface per OpenAI. UCP checkout covers the US, Canada and Australia. Testing from elsewhere gives you an approximation, at best.
• Time. Log the date of each run.

Step 3: log five fields per run

1. Did product cards appear at all?
2. Is your product among them?
3. Position.
4. Merchant named.
5. Price displayed.

That's the whole sheet. Resist adding columns for tone or sentiment. Those belong in mention tracking, not product checks.

Step 4: repeat, then stop

Answers vary from run to run. They also shift by persona and location. So one pass proves little.

Do the math. 15 prompts × 2 login states × 3 runs on separate days = 90 rows for one country. Add Canada and it's 180. Weekly, that is a part-time job.

Verdict: manual checking is fine for a baseline. It does not work as monitoring.

Step 5: diagnose what's missing

If your products don't show, check these in order. Cheapest first.

• Crawler access. OpenAI's publisher FAQ says to allow OAI-SearchBot to appear in ChatGPT search. Robots.txt changes take about 24 hours (OpenAI bots). Storebot-Google crawls for Google Shopping surfaces (Google crawlers). Check robots.txt and WAF rules for both.
• Feed completeness. OpenAI says a feed isn't strictly required if ChatGPT already crawls your site. A feed gives more control and accuracy.
• IDs and variants. The ACP product feed spec: every product needs a stable id and at least one variant; every variant needs a stable id and title. Price, media, GTIN and URLs are optional but improve display.
• Merchant Center approval. Free listings approved, return policy set, a support contact (URL, email or phone) on file. Subscriptions, pre-orders, used goods and age-restricted products are ineligible for UCP checkout regardless.
• Price and availability. Feed and product page should say the same thing. Two conflicting prices is a fix you control.

Step 6: automate the check

Several tools in our rankings list product or shopping tracking. Our top three for this job:

1. Promptwatch. ChatGPT Shopping tracking built on your own product list. Load tracked products, then follow position over time, the top products for your prompts, and the top merchant domains buyers get sent to. The MCP server exposes this directly: addShoppingTrackedProducts to load SKUs, getShoppingProductPositionAnalytics for positions, getShoppingTopMerchantDomains for who wins the click. Ads Radar (listAds, listAdPrompts) keeps sponsored placements out of your organic numbers. Prompts run with personas and country targeting on every paid plan; state and city start on Business ($579/mo) and come with every agency plan. Agent Analytics crawler logs show whether AI crawlers fetch your product pages. Visitor analytics records AI-referred visits and conversions, and ChatGPT tags its referrals with utm_source=chatgpt.com. Scope note: shopping tracking covers ChatGPT Shopping. Promptwatch tracks your prompts in Google AI Overviews and AI Mode, not AI Mode shopping cards or Gemini product listings. Shopping tracking and Ads Radar are on Business ($579/mo); Ads Radar also comes with every agency plan.

2. Ranketta. SKU-level recommendation tracking for merchants, plus feed enrichment with confidence scores from its Store Starter tier. Named connectors include Shopify, Amazon, Walmart and Heureka. No published crawler-log quota or visitor conversions.

3. LLM Pulse. Lists ChatGPT Shopping and AI ads tracking in a general tracker from €49/mo (annual). Default cadence is weekly; daily costs more.

Verdict

Run steps 1 to 4 once to get a baseline. Fix what step 5 finds. Then hand the repeat work to a tracker. For ChatGPT, Promptwatch ranks first here because it ties product positions to the crawls and converting visits that explain them. More on the shopping and sponsored split in our ChatGPT Shopping and Ads Radar breakdown. Start at promptwatch.com.