ChatGPT shopping: does it recommend your store?

When ChatGPT recommends products, is your store there?

We ask an OpenAI model, through the API with web search, the shopping questions your customers ask ChatGPT, from your store’s country and in its language. You see whether it recommends you and your products, which merchant gets the purchase link, and whether it states your price correctly. Every number is labelled “API measurement”, with its limits written next to it.

Why this matters now

For shopping questions, ChatGPT shows product cards with prices and purchase links. OpenAI says merchant ranking considers availability, price, quality and whether the seller is the manufacturer or primary seller.
According to data from one vendor (Profound, reported by Search Engine Journal on 9 September 2026), the share of ChatGPT product recommendations coming from product feeds jumped from 8.26% to 61.54% on 10 July 2026. OpenAI has not confirmed this.
Product data, price and stock consistency now decide more than page content alone. A wrong price or a blocked crawler costs recommendations.

Source: Search Engine Journal (Profound data) ↗ · September 9, 2026 · 1,757,723 prompt runs

Third-party figures are one vendor’s observation; sources are listed at the bottom of this page.

What we measure

Two layers, always labelled separately:

API measurement (Basic and Pro): An OpenAI model (default: gpt-5.5) with web search answers your shopping questions in your store’s country and language: category questions (“best running shoes under $100”), product questions (“where can I buy …?”) and comparisons with a competitor you track. Basic asks 10 questions, Pro 20; each question three times, twice a month. We record whether you or your products are recommended, your position, which domains get the purchase links (your site, a marketplace or another store), share of voice against competitors, and whether the price and stock stated match your catalog. Results read “2 of 3 runs”, and the trend is shown over time.
Team observation (Pro): Each month our team asks 10 of your questions in the ChatGPT app, by hand, in a logged-out window, and records the product cards it sees. These are shown separately and dated, with how often they agree with the API measurement.

What we cannot measure

API answers contain no real product cards or carousel, so we cannot measure how often a question triggers ChatGPT’s shopping view.
Feed-based recommendations most likely don’t show up through the API’s web search, so your visibility in the app may be higher than our API measurement shows.
ChatGPT personalises answers (memory, location, history). Our runs are a sample, not what every shopper sees.
We never scrape the ChatGPT app or capture its network traffic; OpenAI’s terms prohibit automated extraction. The only app data is what our team records by hand.

Readiness: free in every plan

An OpenAI product feed (JSONL, plus a Google-compatible TSV) generated from your catalog and validated row by row against OpenAI’s published spec: required fields, price format, GTIN check digits, sale price, ratings.
OAI-SearchBot access: blocking it in robots.txt keeps ChatGPT from showing your products in search; we check it in every site test.
An optional ACP (Agentic Commerce Protocol) checklist for later; it is information only and does not affect your score.

OpenAI’s direct product feed program currently accepts US, CA, MX as target countries. Turkey and TRY-priced feeds are not in it yet. We generate the file so you are ready when it opens; we do not send it to OpenAI.

Tasks it opens

ChatGPT states a different price than your site, if the same difference shows up in at least two answers.
A competitor is recommended in most runs of a category question where you appear in none.
Most purchase links for your own products go to marketplaces instead of your store.
Feed errors (missing description, invalid GTIN …) as catalog tasks.

Price

ChatGPT Shopping Basic

$49/mo on top of Starter: 10 shopping questions × 3 runs, twice a month, feed readiness and tasks.

ChatGPT Shopping Pro

$99/mo on top of Growth: 20 shopping questions × 3 runs, twice a month, feed readiness and tasks.

Plus 10 questions checked by hand in the ChatGPT app each month, shown as team observations.

Added from the panel. Add-ons are paid one period at a time and don’t renew by themselves. Prices exclude VAT and sales tax.

What we don’t claim

We don’t put you first in ChatGPT; nobody can promise what an assistant recommends.
We don’t submit your feed to ChatGPT: direct feeds are open to approved merchants, and you apply at chatgpt.com/merchants yourself.
We don’t track the same cards the app shows; the API measurement is a separate, labelled view.
We don’t set up Instant Checkout: OpenAI pulled it back in March 2026.

Questions

Is there a ChatGPT product feed for Turkey?

Not yet. OpenAI’s feed spec currently lists US, CA, MX as target countries and does not document TRY-priced feeds. In Turkey, ChatGPT can still find you on the open web: OAI-SearchBot access, correct Product/Offer data on your pages and consistent prices matter today.

Why not just read the cards in the ChatGPT app?

Automated collection from the app breaks OpenAI’s terms. We use the official API and, on Pro, a small number of hand-made checks by our team.

Which model do you use?

By default gpt-5.5 with web search, the same path as our AI visibility measurement. OpenAI doesn’t publish which model runs inside the app, so every run records the model it used; if it changes, the trend chart says so.

Will the cost of the measurement come out of my plan?

Adding the add-on raises your project’s monthly measurement budget by a third of the add-on price, and the add-on has its own cap; measurements stop at the cap and never exceed it.

Sources (accessed 8 October 2026)

See plans and add-ons → Start the free test →