If you sell physical products, you have probably wondered whether people are finding you through ChatGPT. GA4 can tell you, but not on its front page. AI visits are usually a handful of sessions buried under Referral, Direct or Unassigned, so they never show up in the overview.
They may still be worth a look. Writing in Practical Ecommerce, Ann Smarty calls ChatGPT traffic low volume but says it is “often the most engaging source, even more than organic search.” That is her observation across sites, not a number for your shop. The point of this walkthrough is to check it against your own data in about ten minutes, then decide whether any product pages deserve work.
Before you start: what you need
- A GA4 property with at least a few weeks of data.
- Editor access or above if you want a custom channel group (Google’s requirement). If you only have viewer access, use the exploration filter in step 2 instead.
- Your
purchaseevent marked as a key event, so sales show up in the comparison.
Step 1: Give AI assistants their own channel (5 minutes)
A custom channel group lets you add an “AI assistants” channel next to Organic Search, Paid Search and the rest. Google’s help page describes the path:
- Go to Admin → Data display → Channel groups.
- Click Create new channel group. GA4 starts you with a copy of the default group, which is what you want.
- Name it something obvious, like “Channels + AI”.
- Click Add new channel, call it “AI assistants”, and add one condition: Source → matches regex → paste the pattern below.
- Move the AI assistants channel above Referral (top of the list is simplest). Google puts each session in the first channel whose rules it matches, so if Referral sits above it, Referral wins and your new channel stays empty.
- Click Save group.
Here is the pattern:
^(.*\.)?(chatgpt\.com|chat\.openai\.com|perplexity\.ai|gemini\.google\.com|copilot\.microsoft\.com|claude\.ai)$Why it looks like that:
- GA4’s “matches regex” is a full match by default, per Google’s regex help, so the pattern has to describe the whole source value, not just part of it.
\.means a real dot. A bare.means “any character”.(.*\.)?lets subdomains such aswww.through.- It stays specific on purpose. A lazy pattern like
aiwould sweep in every site with those two letters in its name.
Two useful details from Google’s help page: a standard property can have two custom channel groups, and custom channel groups apply to your reports retroactively. You do not have to wait a month for data to build up; your history is regrouped as soon as you save.
Want to add other assistants later? Add them inside the brackets with another |, and check the exact source value in your Traffic acquisition report first.
Step 2: No editor access? Use a filter instead
Practical Ecommerce shows the quickest version: open Reports → Acquisition → Traffic acquisition, switch the dimension to Session source / medium, and type gpt in the search box. That catches ChatGPT only.
For all five assistants at once, build a free-form report in Explore with Session source as the row, add a filter on Session source that matches regex, and paste the same pattern.
Step 3: Put AI next to organic search (3 minutes)
Open Reports → Acquisition → Traffic acquisition and change the primary dimension to your new channel group (it shows up under the name you gave it). Pick the last 90 days so you are not reading tea leaves from a dozen visits. Then compare two rows, AI assistants and Organic Search, on these columns:
- Sessions: how big each source is. Expect AI to be small.
- Engaged sessions and engagement rate: Google counts a session as engaged if it lasts longer than 10 seconds, has a key event, or has 2 or more page views.
- Key events (filtered to
purchaseif you can) and total revenue: whether those visits buy.
Read it with your money hat on:
- AI has a higher engagement rate but few sessions: promising, worth a small amount of page work, not a strategy change.
- AI has purchases: look at which products. That is your shortlist.
- AI has five sessions in 90 days: the comparison means nothing yet. Leave the channel in place and check again next quarter.
The honest bit: you are seeing only part of it
Practical Ecommerce notes that ChatGPT links carry a “noreferrer” attribute, so some visits arrive without telling GA4 where they came from and land in Direct or Unassigned. Your AI assistants channel is a floor, not a total. Do not compare it to organic search and conclude “AI is tiny”. Some of your Direct traffic may be AI too, and GA4 cannot tell you how much.
What you can do: when Direct sessions land on a deep product page rather than your home page, that is a hint someone followed a link, not typed your address. It is a hint, not proof.
Step 4: Pick the product pages to improve (2 minutes)
Go to Reports → Engagement → Landing page, add a filter for your AI assistants channel (or Session source / medium containing gpt, as Practical Ecommerce suggests), and sort by sessions. Now you have a list of the pages assistants actually send people to.
For each of the top three to five pages:
- Compare its engagement rate for AI visitors with the same page for organic visitors.
- Open the page and ask: does it answer what a shopper would have just asked an assistant? Price, size or dimensions, materials, shipping time, returns.
- Fix the gaps in plain text on the page: a clear spec list, a short “who this is for”, shipping and returns stated near the price.
Start with pages that get AI visits but weak engagement. Someone was sent there with a question, and the page did not answer it.
The ten-minute checklist
- Create “Channels + AI” with an AI assistants channel above Referral, using the regex above.
- Traffic acquisition, 90 days: compare AI assistants and Organic Search on engaged sessions, engagement rate, key events and revenue.
- Remember: AI is undercounted because of missing referrers.
- Landing page report filtered to AI: pick three to five product pages and fill in the missing answers.
- Set a reminder to look again in three months.
