Key takeaways
- ChatGPT attaches shopping features (product cards, price comparisons, shopping recommendations) to only a low single-digit percentage of all web-search responses, but that rate spikes hard on commercial and transactional prompts, and OpenAI appears to turn it on and off for categories rather than letting it run constantly.
- Getting listed requires three separate things working together: a product feed that matches OpenAI's spec, crawler access for OAI-SearchBot (not GPTBot), and either a Shopify/Etsy connection or an approved merchant application at chatgpt.com/merchants.
- Product pages and landing pages already account for close to 40% of everything ChatGPT cites, and that share climbs even higher on commercial "best X" style prompts, which tells you exactly what kind of content to have ready.
- Manual spot-checking in the ChatGPT app works for a weekly gut check, but it misses the ad placements, shopping-card triggers, and crawler behavior that only show up with consistent, logged tracking over time.
- Tools built for this, like Promptwatch, scrape the real ChatGPT UI daily and log shopping cards, ads, and crawler visits so you're not relying on memory or screenshots.
Why this suddenly matters
A year ago "ChatGPT Shopping" barely existed as a thing you could optimize for. Now it's a genuine discovery channel sitting in front of a few hundred million people a week. OpenAI's shopping research feature, which rolled out in late November 2025, turns a vague request like "find the quietest cordless stick vacuum for a small apartment" into a researched buyer's guide with specific products, pricing, and tradeoffs pulled from merchant feeds and crawled pages. Add Instant Checkout (launched via the Agentic Commerce Protocol in September 2025) and you've got a system where someone can discover, compare, and buy a product without ever leaving the chat window.
Here's the part that should actually worry brand and ecommerce teams: shopping features don't fire on every query. Promptwatch's data on ChatGPT shopping usage shows the feature attaches to a low single-digit percentage of web-search responses overall, but that rate has moved in sudden steps rather than a gentle trend, roughly doubling overnight in late May 2026 before falling back weeks later. That's not noise. That's OpenAI flipping a switch for certain categories. If you're not tracking when that switch is on for your category, you have no idea whether your absence from the results is a feed problem, a content problem, or just bad timing.
Step 1: confirm your feed meets OpenAI's product feed spec
Before you track anything, make sure there's something to track. OpenAI's official feed reference lives at developers.openai.com/commerce/specs/file-upload/products, and the required fields are stricter than most people expect:
- item_id, title, description (factual, plain text, no keyword stuffing), url, brand, seller_name, image_url
- availability using exact enum values (in_stock, out_of_stock, pre_order, backorder, unknown) -- anything outside this list gets the row rejected
- price as a proper money string, e.g. "79.99 USD"
A couple of things trip people up. The Google-compatible feed variant caps title at 150 characters and description at 5,000, and critically doesn't accept "unknown" as an availability value the way OpenAI's own spec does. If you're repurposing an existing Google Shopping feed, check that discrepancy carefully. GTINs also need to be 8, 12, 13, or 14 digits with a valid check digit, and certain prefixes (02, 04, 2, 05, 98, 99) get dropped entirely. If you don't have a valid GTIN, mpn becomes required.
Two boolean flags decide whether a product can even show up: one controls eligibility for search results, the other controls eligibility for in-chat checkout (and that one requires search eligibility first). Field names differ slightly between OpenAI's docs and third-party feed tools like Lengow, so verify the exact current names before you submit.
Accepted delivery methods are SFTP, API, or a supported commerce platform. If you're on Shopify or Etsy, this is mostly handled for you through the native integration. Everyone else needs to either build a feed pipeline or get the merchant application approved at chatgpt.com/merchants.
Step 2: check that OAI-SearchBot can actually reach you
This is the step people skip, and it quietly breaks everything downstream. OAI-SearchBot is a separate crawler from GPTBot. GPTBot is for model training; OAI-SearchBot is specifically what surfaces pages in ChatGPT's search and shopping answers. If your robots.txt blocks OAI-SearchBot, or your WAF/CDN is silently dropping its requests, your products and pages are invisible to ChatGPT shopping results no matter how clean your feed is.
Check your robots.txt for the string "OAI-SearchBot." If it's not explicitly disallowed, you're visible by default, but that doesn't mean you're safe from an overzealous bot-blocking rule elsewhere. OpenAI publishes its crawler IP ranges at openai.com/searchbot.json, which is the only reliable way to confirm a request claiming to be OAI-SearchBot is actually OpenAI and not a spoofed user agent scraping your catalog.
Worth noting: Promptwatch's crawler traffic data shows OpenAI's share of verified AI crawler requests dropped from 94.8% in early June 2026 to 79.8% by early September, as Anthropic, Google, Perplexity, and Mistral crawlers grew their share. If you're only checking logs for OpenAI's bots, you're missing a growing slice of AI traffic.
Step 3: build the content ChatGPT actually cites for shopping queries
Promptwatch's citation-type data for the Sep 28-Oct 4, 2026 window found product pages and landing pages together make up about 39% of everything ChatGPT cites. Narrow that to commercial, comparison-shopping prompts ("best X under $Y") and product pages alone jump to 26% of citations, with product plus landing pages plus listicles plus comparisons covering roughly 70% of what gets cited. For branded prompts, where someone names your company directly, reviews jump to triple their normal share and social posts to roughly four times theirs, meaning review volume on third-party sites matters a lot more once someone's asking about you specifically rather than browsing a category.
Practically, that means:
- Write product descriptions in plain, specific language rather than keyword-stuffed SEO copy. Shopping research performs best in detail-heavy categories like electronics, beauty, home and garden, kitchen and appliances, and sports and outdoor, so lean into specs and comparisons there first.
- Keep a comparison or "vs" style page for your top products. These convert well in commercial-intent citations.
- Don't neglect Trustpilot, Reddit threads, and niche review sites. ChatGPT pulls third-party trust signals into its shopping answers, and this is one of the few levers brands can control today that isn't paid placement.
Step 4: set up ongoing tracking, not a one-time check
A single manual search in ChatGPT tells you almost nothing useful, because shopping cards, ads, and even citation mix shift by the day. Promptwatch's ads data shows this clearly: ChatGPT served zero ads until May 27, 2026, then climbed to a 90-day average of 20.1% of web-search responses by mid-August, with daily spikes as high as 43.9%. Checking once won't catch that kind of movement.
To track this properly you want three things running on a schedule:
- Shopping card appearance by prompt. Run the actual commercial queries your buyers type ("best running shoes for flat feet," not just your brand name) daily or at minimum weekly, and log whether a shopping card appears, which brands fill it, and whether that's changed since last week.
- Ad placement tracking. Ads on ChatGPT responses concentrate differently by prompt type, with organic prompts carrying the bulk of ad-bearing responses, brand-specific prompts next, and competitor comparisons smallest. Know who's buying the slot when someone searches your category or your competitor's name.
- Crawler log monitoring. Confirm OAI-SearchBot (and increasingly other providers) is actually hitting your product pages, not just allowed to in theory.
Doing all three by hand with manual searches and spreadsheet logging is possible for a small catalog but doesn't scale, and it's easy to mistake a one-day outlier for a trend when ad and shopping rates move in sudden steps rather than smoothly.
Tools built for this
Promptwatch has a dedicated ChatGPT Shopping feature that tracks visibility score, sentiment, shopping card data, and an entity tracker for which brands and products get mentioned alongside yours, plus which of your product pages and reviews actually get cited. It also shows real-time AI crawler visits to your product pages, broken into discovered, indexed, and blocked, which closes the loop between steps 2 and 4 above. It scrapes the actual ChatGPT UI rather than relying only on API sampling, which matters because what a user sees in the app and what the API returns aren't always identical.
If you want something lighter for a first look, Otterly.AI tracks ChatGPT ads and prompt visibility starting at a lower price point, though it covers four engines across all tiers and reserves extra models like Claude or Gemini for paid add-ons.
Profound and Peec AI are also common choices for general AI visibility tracking, though neither has a shopping-specific feature set the way Promptwatch does.
Comparing the shopping-focused options
| Tool | Dedicated shopping/ads tracking | Crawler log visibility | Entry price | Best for |
|---|---|---|---|---|
| Promptwatch | Yes, Shopping Insights and Ads Radar | Yes, real-time crawler visits per page | $95/mo (Shopping Insights at $579/mo tier) | Brands wanting the full feed-to-citation picture |
| Otterly.AI | ChatGPT ads tracking included from entry tier | No | $29/mo | Budget-conscious teams checking ads and basic prompts |
| Profound | General AI visibility, no shopping-specific view | Limited | Custom/sales-assisted | Enterprise teams already in a Profound contract |
| Peec AI | General visibility, report-only | No | $95/mo | Teams that only need prompt-level monitoring |
Step 5: don't ignore the checkout and feed-freshness angle
Once a product is visible, Instant Checkout changes what "visible" has to mean. Merchants can refresh feeds as often as every 15 minutes, and a stale price or an out-of-stock item that still shows as available will frustrate the exact shoppers who were close to buying. If you're submitting through the merchant application rather than Shopify or Etsy, build the feed refresh into your pipeline from day one rather than treating it as a later optimization.
A realistic weekly checklist
- Run 10-15 of your actual buyer queries in ChatGPT (or your tracking tool) and note whether a shopping card appears, and who fills it.
- Check whether ad placements changed on your brand-name and category prompts.
- Pull crawler logs and confirm OAI-SearchBot visited your updated product pages, not just your homepage.
- Spot-check your feed for rejected rows (bad GTIN, mismatched availability enum, sale_price not strictly below price).
- Skim new reviews on Trustpilot and relevant Reddit threads, since these feed branded-query citations more than people expect.
If you want a broader directory of GEO and AI-visibility tools beyond the shopping-specific ones mentioned here, the GEO software directory at bestgeosoftware.com is a reasonable place to compare options side by side.