Hey 
I have 2 questions.
I’ve found ChatGPT and then blogs, but collections have been getting a surprising amount of traffic.
Shopify put out a number on this recently. In Q2 half of AI referred sessions landed straight on product pages, so platform wide the PDP is the big one. Your blogs result is interesting because it goes the other way.
The collections thing makes sense to me though. Most of what people ask these models is list shaped. Best waterproof boots under 150, that kind of thing. A collection page is the only page on a store that already is a list with prices attached, so it is the easiest thing for a model to cite when the question wants options rather than one answer. Blogs pick up the how do I queries, collections pick up the which one queries, PDPs pick up the already decided queries.
Two things mess up the measurement here and both push in the direction you are seeing. The utm_source=chatgpt.com tag sticks to the link after someone copies it, so if a customer pastes that URL into a group chat, every click after that still counts as ChatGPT. And Google AI Overviews sends referrals as plain google.com, so anything coming out of that lands in your organic bucket and never shows up as AI at all. If you are counting only the named AI referrers you are probably undercounting AI overall while overcounting ChatGPT’s slice of it.
For your own store rather than the platform average, open any sessions report in Analytics, filter referrer channel down to the AI sources, then switch the dimension to landing page. That gives you the actual per page split instead of inferring it from totals.
What do you sell? I would expect the blogs versus collections split to move a lot depending on whether people shop your category by browsing or by asking for a specific spec.
ChatGPT is usually the biggest named AI referrer I see. For revenue, product pages tend to win. Blogs often bring more sessions but lower intent, while collections sit in the middle and can assist a lot of purchases.
A few things you can do today:
Online dtc, plus dozens of retail stores here. I looked at our only meaningful ai sessions (chatgpt) and 90% was getting to home (the majority), and about us, blog, company content, and store finder. But, around half of this (“brand”) traffic was in the top 10% of engagement on site. Not much revenue, but sticky. I’ll keep
Upfront I am founder and app Dev so I cant answer the storefront questions. I can bring up something most store fronts are not thinking about right now. With the AI Agent commerce picking up - the question is how is your stores AI visibility? If an agent reaches your store can it find what its looking for and successfully complete the purchase? Many stores are are blind to this.
Yeah, I’ve seen the same thing with collections. Shopify’s own data actually shows that more than half of AI referred sessions start on product pages, so PDPs seem to be the big one overall.
But I wouldn’t be surprised by collections getting a decent amount either. If someone asks an AI tool for something like “best hiking gear under $200,” a collection page can make more sense than sending them to one random product. You’re basically giving the model a group of relevant options to work with.
I’d probably separate this by revenue rather than just traffic too. AI referrals are still small compared with organic overall, but Shopify found they were converting at nearly 50 percent higher rates than organic on product page sessions, with higher average order values too.
I’d be checking product, collection and blog landing pages separately in Shopify Analytics and seeing which ones actually lead to carts and orders. SiteGuru would be useful alongside that for figuring out whether the pages getting attention from AI are also technically solid and properly connected internally.
The collection traffic is probably the part I’d dig into most though. That seems like it could vary a lot depending on the type of store.
1. ChatGPT + Shop instant checkout. I guess Shop is mentioned as part of Agentic channel revenue is since when Shopify Catalog return listing it’s also sending to checkout in Shop.
2. Product pages get the more traffic - around 75% measured across 700 Shopify stores in our app. We offer per agent and per page breakdown in our reports (see example screenshot)
Didn’t know this, i’ll have a look to find their data.
Yeah for bofu it tends to be listicle style content which gets picked up alot in AI.
Blogs is a tricky one, and honestly for most stores it wouldn’t be first unless they are investing in SEO, but those that do, i’ve been seeing it getting picked up. But for the bast majority it will be products as they have the pages for AI to crawl.
Yeah interesting with tracking i would still count the shared link as ai mentions, and the trust is very high from AI anyways but doubley so from a friend sharing a product so even better, and been guilty of sharing links like this too.
Google is a bit of a pain as you cant track AIO yet, but honestly 90% of the traffic is now AIO so i just lump it to that.
Thanks for the share.
I have an app on shopify and a store but don’t share that publicly. Been doing SEO for 10 years now mainly on the affilaite side before Google killed it.
Agree with all that. I would also factor in most of the AI search is external factors, so not always quoting your site as the source. So it’s a good idea to work out what they are and also be featured in the same spots.
Thanks for sharing.
Yeah the about page gets pulled in a lot. I’ve been testing with adding a couple of products being mentioned in the about page to try and get them featured more in BOFU best products lists. But its hard to say this works on a small datasource.
Yeah ChatGPT is still the biggest for general public, excited to see what the shopping/ads do there.
And the best type of traffic 
Im going to check this study tonight. Yeah, honestly it makes sense as most stores will have products over blogs attached so bound to be picked up more on average.
Yeah for sure, when I ever do listicle content i always make sure to segment each product like; best waterproof, most breathable etc as this works great for google longtail keywords and also AI. But collections are easy to crawl for products and very similar.
50% higher, nice, I’ve never had an exact figure but from my data its been the highest converting as its trusted. AOV is interesting.
Nice! thanks for sharing. Yeah ChatGPT seems to be the consensus
Your small-data caution is the important part here. I wouldn’t judge the About-page change from referral traffic alone—it’s sparse and several steps downstream from the change you made.
I’d treat it as a source experiment:
Save a small set of natural buyer questions where those products genuinely fit.
Record the complete answers and sources, not only whether your brand appears first.
Mention the products on the About page only where the maker story or provenance naturally supports them.
Once the page has had time to be reread, repeat the questions under the same recorded conditions.
Then separate four outcomes: was the product found, included as a plausible option, compared using the right facts, and finally ordered where you expected?
That still wouldn’t prove the About page caused a change, but it’s more informative than waiting for a handful of referral sessions. If nothing changes, I wouldn’t keep adding product mentions—the missing evidence may sit in the PDP, collection, commerce data, or outside sources instead.
Hi @shopmentions,
Hey there! Thank you for sharing your experience. Those are great questions.
For revenue, ChatGPT is the biggest driver for many stores right now. Shoppers using AI usually know exactly what they want. Perplexity and Google AI Overviews are also growing fast.
It makes total sense that your collections get a lot of traffic. When people ask AI for product ideas, the AI often links to category or collection pages. This gives the shopper a variety of options to look at.
Product pages get a lot of AI traffic too, but they need the right setup. AI search engines look for clear answers, deep details, and structured data. They want to know exactly what the product is and who it is for. If your product pages lack detail, AI will just link to your blogs or collections instead.
To get more AI traffic, your store needs solid technical SEO. AI bots read hidden code called schema to understand your pages. If you want to boost your visibility, you should try Searchpie. It fixes schema issues and optimizes tags for you automatically. It makes sure AI bots and search engines can read your store clearly.
Keep up the great work. AI search is the future, and you are already ahead of the curve. Let us know if you have any more questions!
Product pages taking ~75% of AI sessions matches what I see too. One thing worth separating from “which pages get the traffic” though: whether an assistant can actually use the product page once it gets there.
Two cheap things to check on your own store:
1. What the product JSON-LD says. Open a product page, view source, find the `` block with `“@type”:“Product”`. Look for `gtin13`/`gtin12` (or `gtin`), `brand`, `sku`, and an `offers` object with `price`, `priceCurrency` and `availability`. That block is what gets read. What sits in your admin doesn’t count if the theme never renders it — I’ve seen plenty of stores with the barcode filled in on the variant and nothing in the markup.
2. Whether there’s a product identifier at all. For resellers this is the one that matters. An assistant answering “who has the Judy Blue high-rise in a 14 in stock” has to match your listing to the same product on other sites. With a GTIN that match is exact. Without one it’s a fuzzy title match, and fuzzy loses to the store that was explicit.
Rough number from checking a few hundred US fashion Shopify storefronts that resell third-party brands: only about 3 in 10 render a GTIN in the product JSON-LD. It’s not a niche gap.
The fix depends on which kind of product it is, and getting this backwards is the expensive mistake:
- Brands you resell: the GTIN already exists — it’s on the brand’s line sheet, or on the barcode sticker on the polybag. It goes in the variant’s **Barcode** field. Then check the live page to confirm the theme renders it.
- Products you make yourself: don’t invent one. Own-brand items legitimately have no GTIN, and that’s fine. Use `brand` + `mpn`, and set `identifier_exists` to `no` in the feed. Making up numbers is one of the few things that can get a Merchant Center account suspended.
If you’re already segmenting AI traffic by page type, one more cut worth adding: split product-page AI sessions by whether that product has an identifier. That tells you whether the traffic is arriving despite the gap, or only for the products that happen to be matchable.
Formatting note on my post above: the editor stripped the script tag I referenced, so the first point reads oddly. To be concrete: open one of your product pages, use View source, and search the page for ld+json. That block is what crawlers and assistants actually read. Inside it, check that the product declares a type of Product, and that it carries gtin13 or gtin12 (or gtin), brand, sku, and an offers object with price, priceCurrency and availability. If the block is there but the identifier fields are missing, that is the gap worth closing first.
I’ve noticed something similar with ecommerce SEO. Product pages seem to have the strongest commercial intent, while collection pages can work well for broader comparison-type searches.
I’d also pay attention to structured data and how clearly product information is presented. Accurate Product schema, pricing, availability and strong internal linking can make it easier for search engines and AI systems to understand the page.
For measurement, I think it’s useful to separate AI traffic by landing page and then compare conversion rate and revenue rather than looking at sessions alone. That gives a much clearer picture of which pages are actually contributing to the business.
