ChatGPT sends shoppers to a store from two places. One is a link written into an answer. The other is a product listing, drawn from the store’s catalog and shown as a shopping result. We split 23,051 ChatGPT-attributed orders between them. Listings account for 24%, and they behave differently enough to be worth measuring on their own.
Reports usually show ChatGPT as a single line. Underneath it are two mechanisms with different levers. An answer link exists because the model chose to cite a page. A product listing exists because the store’s catalog reached ChatGPT in a form it could show. A store can do very little about the first and a good deal about the second.
| Month | Answer links | Product listings | Ads |
|---|---|---|---|
| May 26* | 822 | 266 | 1 |
| Jun 26 | 1,803 | 392 | 4 |
| Jul 26 | 2,068 | 602 | 8 |
| Aug 26 | 3,113 | 1,159 | 9 |
| Sep 26 | 3,946 | 1,042 | 4 |
| Oct 26 | 3,317 | 1,197 | 12 |
| Nov 26* | 2,393 | 879 | 14 |
Listing orders tripled between June and August, from 392 to 1,159 a month, and have held near that level since. Answer-link orders grew over the same months too, so the listing share moved less than the listing count.
| Month | Product listing share |
|---|---|
| May 26* | 24.4% |
| Jun 26 | 17.8% |
| Jul 26 | 22.5% |
| Aug 26 | 27.1% |
| Sep 26 | 20.9% |
| Oct 26 | 26.4% |
| Nov 26* | 26.7% |
Nearly every listing order (99.9%) began on a product page, as you would expect from a result that represents one product. Answer links are looser: 71.9% landed on a product page and 6.6% on a collection, with the rest on home pages, articles and other pages.
| Answer links | Product listings | |
|---|---|---|
| Orders in the window | 17,462 | 5,537 |
| Landed on a product page | 71.9% | 99.9% |
| Median order, US-dollar orders | $73.23 | $80.13 |
The median listing order was $80.13 against $73.23 for answer links, about 9% higher. The mix of stores differs between the two groups, so this is a description of the orders and not a claim that listings cause larger baskets.
About 840 stores took at least one ChatGPT-attributed order in the window. About 380 took at least one from a listing. The gap is the practical finding: for more than half of the stores already selling through ChatGPT, all of it comes from being cited in an answer and none from the catalog.
Ads are the third source and barely register. 52 orders in the window carried an ad marker, under a quarter of one percent of ChatGPT orders. They are counted separately so that they can be watched as the ad product rolls out.
Every figure comes from the order records of Shopify stores running Aimerce, read as one fleet. An order counts as AI-attributed when Shopify’s own record of it names an assistant in one of four ways: the sales channel that created it, a tag on the landing link, the referring site, or a browser that identifies itself as an assistant’s app or agent. Each order is counted once.
The three sources are separated by what ChatGPT appends to the link. Product listings carry a feed marker in addition to the ChatGPT tag; ads carry a paid marker; everything else with the ChatGPT tag or a ChatGPT referrer is treated as an answer link. A link that lost its tag entirely is not counted in any of the three.
What this does not measure. Shoppers who asked an assistant and later came back another way, clicks that arrive with neither a tag nor a referrer, visits from Google’s AI Mode and AI Overviews (they arrive as ordinary Google traffic), and agents that run in the shopper’s own browser without identifying themselves. All four make the numbers here a floor.
Privacy. Only fleet-level aggregates are published. Store counts are rounded to the nearest ten, and no store, brand, product or order is identified anywhere on this page.