Should Shopify merchants count AI assistant visits separately from orders placed inside AI shopping tools?
A shopper can find your product in an AI assistant in two very different ways.
They might click from ChatGPT to your product page, browse your store, and check out there. Or they might buy inside an AI shopping tool without ever loading your storefront.
If you put both paths into one number, you lose the useful part of the report.
A visit tells you an assistant sent someone to your site. An order tells you someone bought. A direct in-AI order may have no store visit attached to it at all.
Shopify now separates these events in Sales channels → Agentic. That gives merchants a better way to look at AI commerce than treating every AI-related action as a storefront session.
A visit is not an order
Shopify defines an Online store session as a visit to your online store from an AI channel. This is the familiar path:
1. A shopper asks an assistant for a product. 2. The assistant recommends or links to your store. 3. The shopper opens your product or collection page. 4. Your storefront loads in a browser. 5. They may or may not buy.
ChatGPT currently works this way for Shopify merchants. It sends shoppers to the merchant's online-store checkout, either in ChatGPT's in-app browser or a new browser tab.
That makes visit reporting useful. You can see whether AI assistants are sending people to your store and which landing pages they reach.
But it does not tell you about every order an AI channel may produce.
Google AI Mode and Gemini, Microsoft Copilot, and some Meta surfaces can support direct checkout. In that path, a shopper can buy without leaving the AI channel.
No product page needs to load. No storefront session needs to happen.
The order still matters. It just belongs in a different bucket.
Why one conversion rate can mislead you
Shopify defines online store conversion as:
`completed checkout sessions ÷ sessions`
That works for shoppers who arrive at your storefront and complete checkout there.
It does not describe direct in-AI checkout well. A direct order can increase your AI-attributed sales without adding an online store session first.
Say an assistant sends 20 shoppers to your store. Two buy on your storefront. It also completes three direct checkouts inside its own interface.
Your storefront conversion is based on the 20 sessions and the two storefront checkouts. The three direct orders are real orders, but they did not come from those sessions.
If you divide all five orders by 20 visits, you create a number that mixes two different buying paths.
That does not mean either path is better. It means they answer different questions.
- AI assistant visits answer: did an assistant send shoppers to my store?
- AI-referred online-store orders answer: did those visits turn into storefront orders?
- Direct in-AI orders answer: did shoppers buy without reaching my storefront?
- Total AI-attributed sales answer: what sales value did AI channels contribute across both paths?
Keep these separate before you compare one period with another.
Check Shopify's Agentic reports first
You can verify this in under a minute.
In Shopify admin, open Sales channels, then Agentic. Shopify provides separate reports for:
- Sales
- Orders
- Online store sessions
- Online store conversions
Start with Online store sessions and Orders side by side.
If sessions rise but orders do not, that may point to a storefront issue. Check the landing pages, product details, price, shipping information, and checkout flow.
If orders rise without a matching increase in sessions, look at whether direct checkout could be part of the picture. Do not assume your tracking broke just because the storefront did not record a visit.
Shopify also says AI-channel orders appear in admin with channel or referrer attribution. That gives you a place to inspect orders separately from the traffic that reached your site.
Where Descriva fits
Descriva tracks the storefront side of this picture.
With your permission, its Shopify app pixel counts visits that assistants send to your store and the pages those visitors land on. It can identify visits from ChatGPT, Perplexity, Gemini, Copilot, Claude, and other assistants. It does not record who the shopper is.
That helps answer practical questions such as:
- Are AI assistants sending shoppers to the store at all?
- Which product pages do those shoppers reach?
- Did a recent product-page update coincide with more assistant-referred visits?
- Are shoppers landing on a page that answers the question an assistant likely sent them with?
Descriva does not treat those visits as orders.
Orders placed inside an assistant do not load your store, so a storefront pixel cannot count them as visits. Shopify's own Agentic home reports those AI-channel results, and Descriva points you there.
This distinction is deliberate. A useful report should say what it observed, not fill in the missing path with an assumption.
Count recommendations separately too
Traffic and orders are not the full picture.
Before an assistant can send a shopper to your store, it has to decide whether to name you in an answer.
Descriva's AI Mentions checks this directly. It asks ChatGPT, Perplexity, Claude, and Gemini questions shoppers ask about your products. It shows, question by question, whether an assistant named your store, left it out, or named another brand instead.
Those results are counts, not a score.
For example, you might see that ChatGPT named your store in 3 of 9 answers, while another assistant named it in 1 of 9. That is different from knowing how many visitors arrived on your site afterwards.
The sequence matters:
1. An assistant is asked a shopping question. 2. It may name your store or another brand. 3. A shopper may click through to your storefront. 4. They may place an online-store order. 5. Or, where direct checkout is available, they may buy inside the AI channel.
Each step has a different failure point. If you only watch sales, you cannot tell which one needs work.
Use a four-part AI report
You do not need a complicated dashboard to start. Keep four lines in your regular reporting:
- AI assistant visits or online store sessions
- AI-referred online-store orders
- Direct in-AI orders
- Total AI-attributed sales
Add notes when something changes. For example: a product page was updated, a new collection launched, or an assistant began naming a competitor more often for an important question.
Then look for patterns over time.
If mentions are low, work on the information assistants can read: clear product copy, accurate specs, image alt text, structured data, policies, and answers to common shopping questions.
If mentions are present but visits are low, inspect the recommendation context. Is the assistant naming you for the right product and question?
If visits are healthy but storefront orders are weak, inspect the landing page and checkout experience.
If direct AI orders appear, do not force them into the storefront conversion rate. Record them as their own path.
The next step is simple: open Shopify's Agentic reports and compare Online store sessions with Orders for the same date range. Then keep them separate in your next report.