If you looked at your analytics this year and decided AI assistants do not matter for your shop because the traffic is tiny and not growing, a new benchmark suggests you may be counting the wrong thing.
Shero Commerce, an ecommerce agency, looked at 73 online stores (42 on Shopify, 14 on Magento, the rest on BigCommerce, WooCommerce and other platforms) between 1 September 2025 and 26 September 2026: 96.4 million sessions and 976,668 orders. The headline: visits from AI assistants were almost flat, going from 896 a day across the group to 917. Orders from those visits went from 5.9 a day to 16.1, which is 2.73 times as many. The conversion rate of AI visitors went from 0.63% to 1.75%.
So the people arriving from ChatGPT and the rest are not more numerous. They are further along. They have already asked an assistant to compare options and they click through to buy.
Two other findings are useful for a small shop. ChatGPT sent 88.1% of the AI visits in the study, with Gemini at 5.1%, Perplexity at 3.7%, Copilot at 1.9% and Claude at 1.2%. And the landing page made a large difference: about half of AI visits (51%) landed on product pages, which converted at 1.77%, while visits landing on the homepage converted at 0.57% and on blog posts or guides at 0.29%.
The limits matter too. This is 73 stores chosen by one agency, not a sample of all shops, and the study itself says GA4 probably undercounts AI traffic: visits from the ChatGPT mobile app and checkouts completed by an agent often arrive with no referrer and land in Direct. Treat your AI numbers as a floor, not the full picture.
Here is how to see the same thing in your own store in about half an hour.
Step 1: give AI assistants their own channel in GA4
By default GA4 drops ChatGPT, Perplexity and the others into Referral, mixed with every blog and directory that links to you. A custom channel group pulls them out.
- In GA4 go to Admin → Data display → Channel groups and click Create new channel group. GA4 starts you with a copy of the default channels, which is what you want.
- Name it something like Channels with AI, then click Add new channel and call the channel AI assistants.
- Set the condition to Source → matches regex and paste the pattern below. It is the list Shero Commerce published, wrapped in
.*on both sides so it works whether GA4 treats the regex as a full or a partial match. - Save, then drag the AI assistants channel above Referral. GA4 assigns each session to the first channel that matches, so if Referral sits above it, your AI channel will stay empty.
.*(chatgpt|openai|perplexity|gemini\.google|bard\.google|copilot|edgeservices|claude\.ai|anthropic|meta\.ai|deepseek|grok|you\.com|poe\.com|mistral).*Custom channel groups also apply to past data, so you do not have to wait a month to see results. One limit: you can use this new group in Explorations and some standard reports, but the default channel group stays in place everywhere else.
Step 2: keep 14 months of data
The study's most useful comparison is this September against last September. GA4 keeps detailed event data for only two months unless you change it, and Explorations can only reach back as far as that setting allows. Go to Admin → Data collection and modification → Data retention and set event data retention to 14 months, the longest a standard property allows. It does not bring back data already deleted, so do it now even if you will not look for a while.
Step 3: build one exploration that counts orders, not visits
Go to Explore → Free form and set it up like this:
- Rows: Session source, then Landing page.
- Values: Sessions, Transactions, Purchase revenue.
- Filter: Session source matches regex, with the same pattern as above.
- Date range: last 28 days, compared with the same 28 days a year ago if you have the data.
Then read it in this order:
- Transactions first. If sessions are flat and transactions are rising, you are seeing the same pattern as the study. If both are near zero, you have learned something too: assistants are not sending you buyers yet.
- Which landing pages get AI visits. Sort by sessions. If most land on your homepage, the assistant knows your shop exists but not which product answers the question. The study found homepage visits converted at about a third of the rate of product page visits.
- Which source. Expect ChatGPT to dominate. A few visits from Perplexity or Copilot are still worth a look, because they tell you which product pages those assistants cite.
Step 4: fix the pages the assistants actually send people to
The landing page split in the study is the practical part. Product pages and collection pages turned AI visits into orders; homepages and blog posts mostly did not.
For a small shop that suggests a short to-do list, in order:
- Take the top five AI landing pages from your exploration and open each on a phone. Someone arriving from a chat answer wants to check the price, the size or variant and the delivery time, then buy. If any of those are below the fold or behind a tab, move them up.
- Write the facts an assistant compares as plain text on the page: dimensions, materials, what is in the box, delivery time, return terms. Text in images cannot be quoted.
- If AI visits land on a blog post, add a clear link from that post to the product it talks about. The study's blog visits converted at 0.29%, so do not expect sales from the post itself.
- Check the month again. Put a reminder in four weeks to compare transactions from the AI channel. Visits are the wrong scoreboard; orders and revenue per visit are the right one.
What this does not tell you
The study compares AI visitors with organic search visitors too: on 41 of the 61 stores with at least 200 AI visits, AI visitors converted better, and the median store saw a 25% higher conversion rate and 23% more revenue per visit than organic, but a 15% smaller average order. That is encouraging, but 917 AI visits a day across 73 stores works out to roughly 12 a day, under 400 a month, for the average store in the study. For most small shops AI is a small channel with good buyers, not a replacement for search. The point of the setup above is to know which one you have, with your own numbers, before you spend money on it.
