Some AI assistants no longer send the shopper to the store. They take the order themselves, in the conversation, and hand the store a finished sale. Across the stores we measure this went from 5 orders in June 2026 to 208 in October. For most of these orders there is no visit, no session and no pixel event to attribute.
Through July the count was in single digits. It reached 81 in September and 208 in October. Nearly all of the step is one assistant: Meta AI accounts for 87% of the in-assistant orders in the window.
| Month | Meta AI | Google AI Mode and Gemini | Microsoft Copilot | Total |
|---|---|---|---|---|
| Jun 26 | 0 | 4 | 1 | 5 |
| Jul 26 | 1 | 2 | 2 | 5 |
| Aug 26 | 23 | 4 | 6 | 33 |
| Sep 26 | 76 | 5 | 0 | 81 |
| Oct 26 | 196 | 11 | 1 | 208 |
| Nov 26* | 81 | 21 | 0 | 102 |
The growth is not a straight line. The first 16 days of November brought 102 orders, a pace of roughly 191 for the month, which is a little under October. Within that, Google’s share is rising: 21 orders in half of November against 11 in all of October.
| Month | Meta AI | Copilot | Total | |
|---|---|---|---|---|
| Jun 26 | 0 | 4 | 1 | 5 |
| Jul 26 | 1 | 2 | 2 | 5 |
| Aug 26 | 23 | 4 | 6 | 33 |
| Sep 26 | 76 | 5 | 0 | 81 |
| Oct 26 | 196 | 11 | 1 | 208 |
| Nov 26* | 81 | 21 | 0 | 102 |
69.1% of in-assistant orders were created by the assistant’s own server. The record carries no shopper browser, no landing page and no referring site. When we looked individual orders of this kind up through Shopify’s API, the customer journey attached to them listed zero visits.
That removes everything browser-based measurement relies on. A tracking pixel does not fire because no page loads. There is no session to tie the order to an earlier ad click. Analytics that count conversions from page events record nothing, and the sale shows up only in the order list.
The remaining 30.9% did carry a browser. That is consistent with the assistant handing the shopper to the store’s checkout inside its own app to finish paying, in which case a session exists for the last step only.
What does identify these orders is the sales channel Shopify stamps on each one. Every assistant that checks out on a store’s behalf does so through its own channel, and the channel name is on the order whether or not a browser was involved.
This matters for anyone building reports. A method that starts from sessions and works forward to orders cannot find these sales, however good its session data is. A method that starts from orders and works backward finds all of them. We count them from the order side for that reason.
Order values are lower than on other AI paths: a median of $58.50 among US-dollar orders, against $75.95 for orders that arrive by a link from an assistant.
Two things. First, whether October’s level holds through the holiday season or turns out to be a launch spike. Second, how quickly the other assistants follow. At the moment in-assistant checkout is close to a single-assistant story, and the largest source of AI-attributed orders overall, ChatGPT, still sends shoppers to the store to pay. If that changes, the share of AI commerce that is invisible to session-based measurement changes with it.
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.
In-assistant orders are identified by the sales channel recorded on the order. A channel is included only after its name was confirmed on real orders. Orders created by other server-side apps are not counted, which matters because they are far more numerous; see the report on false positives in this series.
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.