How to Check If Product Listings Appear in AI Shopping Results
Which parts of an AI shopping visibility check an agent can run on a schedule, which it can only recommend, and which still need a human with Merchant Center or PIM access.
Checking whether your products appear in AI shopping results sounds like one task. It's actually about six, and they automate very differently. Some an agent can run every day without anyone looking. Some it can only flag. A couple still need a person with admin rights to Merchant Center or your product information system.
This post splits the check into those jobs and sorts each one by the question we ask of every tool on this site: does it monitor, recommend, or execute?
What you're checking
Two surfaces show product listings today.
ChatGPT shopping lets buyers browse products visually and compare them side by side. OpenAI extended the Agentic Commerce Protocol (ACP) in March 2026 so merchants can send product feeds for discovery (OpenAI), and its merchant page says shopping is live in the U.S. Purchases complete on your own site.
Google AI Mode and Gemini can show a Buy button on eligible listings through the Universal Commerce Protocol (UCP). That runs off Merchant Center, needs products approved for free listings, and applies to products eligible in the United States, Canada, and Australia, for select merchants (Merchant Center Help).
Job 1: run buyer prompts and record what shows
Can an agent do it on a schedule? Yes, and it should.
By hand, you write prompts with real constraints ("cordless stick vacuum under $300 for pet hair"), run each one logged out and logged in, note your location, and log whether cards appeared, which products, your position, the merchant, and the price shown. Then you do it again tomorrow, because answers vary from run to run and shift by persona and location. A single pass shows you one buyer's view at one moment.
This is the job agents are best at. It's repetitive, the inputs are stable, and the value comes from volume. Promptwatch runs it as ChatGPT Shopping tracking, included from Business ($579/mo): you load your tracked products, prompts refresh daily on every paid brand plan, and it records product positions over time along with the top products and top merchant domains for your prompts. Prompts can carry personas and run per country on every plan, with state and city tracking on Business and every agency plan.
One limit to state plainly: Promptwatch's shopping tracking covers ChatGPT Shopping. It tracks your prompts in Google AI Overviews and AI Mode, but not AI Mode shopping cards or Gemini product listings. For the Google side, Merchant Center remains your source.
Job 2: separate sponsored from organic
Agent can do it. Monitor only.
If a competitor is paying for placement on your prompts, that's a media question, not a feed problem. Ads Radar in Promptwatch (Business and agency plans) tracks sponsored placements on the same prompt set, so a paid unit doesn't get counted as an organic loss. Nothing here executes; it just keeps the diagnosis honest.
Job 3: check crawler access
Agent monitors. A human fixes.
OpenAI says to allow OAI-SearchBot if you want to appear in ChatGPT search (publisher FAQ); robots.txt changes take about 24 hours (OpenAI bots). Google's Storebot-Google crawls for Google Shopping surfaces (Google).
Promptwatch's Agent Analytics reads your AI crawler logs from Cloudflare, CloudFront, Fastly, Vercel and other CDNs, shows which product pages AI crawlers fetched, and flags crawl errors. What it can't do is edit your robots.txt or WAF. Someone on the dev side still makes that change. Check Storebot-Google in your own server logs.
Job 4: fix the feed
Mostly human. Some tools execute part of it.
The ACP product feed spec asks for a stable id on every product and at least one variant, each variant with its own stable id and title. Price, media, GTINs, and URLs are optional but improve display. On Google, UCP checkout needs a return policy, a support contact, and the native_commerce(checkout_eligibility) attribute set to true, and products like subscriptions, pre-orders, and refurbished goods are ineligible.
A few tools in our catalog act on catalog data rather than just reporting on it:
- Botify pushes enriched product feeds to Google and OpenAI. It's a sales-led enterprise install with no public price.
- Ranketta enriches titles, descriptions, and attributes with a confidence score, starting on its Store Starter tier.
- Wildcard syncs your catalog from Shopify, BigCommerce, Salesforce, or a PIM, and its actions include PDP optimisation and schema enrichment with publishing to Shopify.
Even with these, stable IDs are a decision for whoever owns the PIM. An agent that regenerates IDs would erase your own history, so keep that one manual. The same goes for making price and availability agree between feed and page: pick one source of truth, and let everything else read from it.
Job 5: turn misses into a work list
Agent recommends. Some of it executes.
Once tracking shows which products drop out, the next question is what to do. Promptwatch's Unified Actions produce an AI-generated to-do list from your visibility data, and content gap analysis shows prompts where you have no page that answers the question. Where the fix is content (a buying guide, a comparison, a category explainer), Content Agents can plan, write, and publish it to Webflow, Framer, or WordPress, with a review inbox in front of anything you don't want going straight live. On Framer, publishing redeploys the whole site, which matters if you schedule it daily. Feed edits are not part of what it publishes.
Job 6: report on a cadence
Agent can do it.
Through the Promptwatch MCP server, an assistant can pull the shopping data on request: listShoppingTrackedProducts, getShoppingProductPositionAnalytics, getShoppingTopProducts, getShoppingTopMerchantDomains. To make it recurring, run a saved prompt in Claude each week, use the Promptwatch Slack agent, connect Looker Studio (Professional, Business, and agency plans), or script pulls against REST API v2. MCP and API access are on every paid plan. Pair it with visitor analytics so the report shows AI-referred visits and conversions, not just positions.
The split
Tracking, ad separation, crawler monitoring, recommendations, and reporting can all run without you. Feed IDs, Merchant Center settings, robots.txt, and approvals stay with a human. Promptwatch takes the most of that list off your plate for ChatGPT, which is why it tops our agentic tool rankings; pair it with a feed tool if your catalog needs rework. For a wider view of commerce tooling, see our ranked list of AI SEO tools for ecommerce stores. Shopping tracking and Ads Radar start on Business at $579/mo at promptwatch.com.