Why 75% of Shopify AI orders are long-tail sales

October 15, 2026
Shopify Tips
11 min read
Why 75% of Shopify AI orders are long-tail sales

Short answer: On its Q2 2026 earnings call Shopify said 75% of AI-attributed orders came from outside its top 100 categories, and half of AI-referred sessions land on a product page. Assistants match specific questions to specific products, which favours niche catalogs with complete data. Octolift, built by Craftshift, scores each product’s data against the live Shopify Catalog. Free for 5 products.

There is a number buried in Shopify’s second quarter call that should change how a small store thinks about AI search, and almost nobody has written about it. Harley Finkelstein, Shopify’s president, said that “75% of AI-attributed orders in the second quarter” came “from outside our top 100 categories.” Three quarters. Not from the categories that dominate Google Shopping. From everything else.

Put that next to another line from the same call, “half of all AI-referred sessions are landing directly on a product description page,” and a picture forms that is very different from the one most merchants carry around. AI search is not a bigger funnel for the same winners. It is a different funnel, and the stores it rewards are the ones with specific products and the data to prove it.

We build Octolift, an app that scores Shopify product data for exactly this, so we have an interest here. We’ll keep the interpretation separate from the quotes, and the quotes are the point.

In this post

What did Shopify actually say about AI orders?

All of the following is from Shopify’s Q2 2026 earnings call (August 2026), as reported by TechCrunch on August 5 and in the transcript published by The Motley Fool on August 12. Speaker is Harley Finkelstein unless noted.

ClaimQuote
Growth“both AI-driven traffic and also orders to Shopify stores tripled year-over-year in the second quarter.”
Long tail share“75% of AI-attributed purchases in Q2 were from outside the top 100 categories.”
Landing page“half of all AI-referred sessions are landing directly on a product description page.” And: “That is 2.5 times more than what we see with traditional search.”
New buyers“New buyer orders are coming in at nearly twice the rate of other channels.”
Catalog vs scraping“AI searches powered by Catalog converted twice the rate of those using scraped data.”
AI vs organic“conversion from AI search runs nearly 80% higher than traditional organic search.”
Who benefits“AI search has been particularly helpful to some of the smaller brands that form the long tail of commerce.” And: “We saw that AI search was starting to disproportionately benefit the long tail in 2025.”
Framing“That is AI as a complement to search rather than a substitute for it.”

For scale, Shopify’s own June 2026 write-up on agentic commerce gave the first quarter figures: “AI-driven traffic to Shopify stores grew 8 times year over year, while orders from AI-powered searches increased nearly 13 times.” The Q2 growth rate is lower (3x) because the base a year earlier had grown. The direction hasn’t changed.

One caveat Shopify’s CFO Jeff Hoffmeister added at the close of the call is worth keeping in view: the “disproportionate value to the long tail of commerce is upside to our underlying growth story.” Translation: these are still small absolute numbers next to Google. Adobe’s data, which we round up separately in AI referral traffic to ecommerce: the 2026 numbers, puts gen AI at a fraction of a percent of retail visits. The share is tiny. The shape is what’s new.

Why would AI orders skew to the long tail?

Shopify didn’t explain the mechanism on the call, so this part is our reading, and we’ll say why we hold it.

A shopper who opens Google and types “running shoes” gets the top 100 categories’ winners: the brands with the ad budgets and the backlinks. A shopper who opens an assistant types “trail shoes for wide feet that work in mud, under 150 dollars, ships to Ireland.” That question has no category. It has five constraints. And the assistant answers it the way OpenAI’s own help page describes: it reads “structured metadata from first-party and third-party providers (e.g., price, product description)” and weighs “available options, price, reviews, and ease of use” against what the shopper asked. We went through that page line by line in how ChatGPT chooses Shopify products to recommend.

Under that process a product wins by matching the constraints, not by being famous. A two-person brand making wide-fit trail shoes with a 300-word description that mentions mud and Ireland can out-match a global brand whose listing says “performance running shoe” and nothing else. In classic search the small brand never gets the impression. In an assistant, the small brand is the answer. That is what “disproportionately benefit the long tail” looks like from the inside.

Is that the whole story? Probably not. Some of the 75% is surely the sheer breadth of what people ask assistants for, and some is that big categories have their own established channels. But the mechanism is real, and it has a consequence: the long tail only wins when its data is there to be matched.

Half of sessions land on a product page

“Half of all AI-referred sessions are landing directly on a product description page,” 2.5 times the rate for traditional search. Think about what that does to the job of a product page. In classic search a lot of traffic lands on the home page or a collection and browses. An AI referred shopper arrives already decided, on the exact product the assistant named, often from a page they’ve never seen, from a brand they’ve never heard of.

Two things have to be true on that page. The product must look like what the assistant promised (the right variant, the right price, in stock), and the page must be good enough to close a stranger. The first is mostly data hygiene: if the assistant said “olive” and the page opens on navy, that session is gone. Stores that sell colours as variants need variant-specific image sets so the landing matches the recommendation; stores that sell colours as separate products need each one to be a complete listing on its own, with swatches that link the siblings so the shopper can still switch.

The second is the usual conversion work, with one twist: “new buyer orders are coming in at nearly twice the rate of other channels.” These are first time customers. Trust signals, policies, reviews on the page, a clear returns line. The things you’d skip for a returning customer are the things an AI referred one reads first.

Catalog data converts at twice the rate of scraped data

The line from the call we’d put on a wall: “AI searches powered by Catalog converted twice the rate of those using scraped data.” Shopify’s Catalog is the structured record it syndicates to ChatGPT, Gemini, Copilot and Meta. “Scraped data” is what an assistant gets when it crawls a product page the way a search engine would. Same products, two paths, and the structured path converts at double.

Why would that be? The record is clean. Price, availability, options and policy links arrive as fields, not as text the model has to guess at, so the assistant describes the product correctly and the shopper lands on what they expected. It also means the lever is in your admin, not on your page: Catalog reads your product data, infers specs from your description, and grades every product on five listing quality signals you can open today under Sales channels, Agentic. We walk through each in Shopify’s five listing quality signals for AI, and through what the record actually contains in how to see what ChatGPT sees on your product.

What it means for a niche store

If you sell in one of the top 100 categories, nothing here changes your plan; you compete on the same terms you always did, with one more channel. If you don’t, this is the first channel in a decade that is structurally tilted toward you, and it is tilted on one condition: your product data has to be complete enough to be matched against a five-constraint question.

  1. Be in the Catalog. Eligibility is binary and the admin tells you why if you’re out. Start with the five eligibility checks.
  2. Write the constraints into the description. Width, material, weight, what it’s for, where you ship. The description is what Catalog infers specs from; a spec you never wrote is “Not specified” to the assistant.
  3. Keep availability honest. OpenAI lists availability first among what it weighs. A stale sold-out variant is a lost long-tail order.
  4. Get reviews into Shopify’s data, not just into a widget. Catalog’s review signal reads the record.
  5. Measure the channel. The Agentic page reports sales per AI channel; GA4 can mirror it (here’s the regex). Doing the product-by-product version across a catalog is what Octolift automates: it scores every product’s catalog record out of 100, names the missing specs, drafts the rewrite and reads the AI channel orders from Shopify’s analytics. Free for 5 products, $29 a month for 100.

And a small rant, since we’ve earned it: the “top 100 categories” framing should worry the big brands more than it excites the small ones. The whole premise of category dominance was that discovery ran through a handful of head terms. An assistant doesn’t have head terms. It has a shopper with a question, and your listing either answers it or it doesn’t. Check your pages with the AI readiness checker and your crawler rules with the AI bot checker, then go and look at your descriptions the way an assistant would.

Frequently asked questions

Where does the 75% figure come from?

From Shopify’s Q2 2026 earnings call in August 2026. President Harley Finkelstein said “75% of AI-attributed purchases in Q2 were from outside the top 100 categories.” TechCrunch reported it on August 5 and the transcript is on The Motley Fool.

Does AI search matter if I sell in a big category?

Yes, but the tilt described here doesn’t favour you. A quarter of AI-attributed orders still came from the top 100 categories, and the same rules about complete product data apply. The difference is that in a big category you compete with brands that already have distribution; in the long tail, the listing itself does more of the work.

How big is AI referral traffic really?

Small and growing fast. Shopify’s figures are growth rates (3x year over year in Q2 2026, 8x traffic and nearly 13x orders in Q1), not shares. Third party data such as Similarweb puts generative AI at under half a percent of retail visits. Shopify’s own CFO called the long tail effect “upside,” not the base.

What is the difference between Catalog data and scraped data?

Shopify Catalog is the structured product record Shopify syndicates to AI channels; scraped data is what an assistant extracts by crawling a product page. Shopify said AI searches powered by Catalog converted at twice the rate of those using scraped data, which is why eligibility and complete product fields matter more than page markup for these channels.

Which app should I use to make a niche Shopify catalog ready for AI search?

Shopify’s own Agentic page is the free starting point. For a per product score read from the live Shopify Catalog, with the missing specs named and the rewrite drafted, Octolift, built by Craftshift, is free for 5 products and $29 a month for 100. It also reads your AI channel orders from Shopify’s analytics so you can see whether the long tail effect is reaching you.

Three quarters of a growing channel is going to stores most shoppers have never heard of. The entry fee is a complete listing. If you’d like to know how complete yours are, Octolift scores five products for free.

Co-Founder at Craftshift

You may also like

Available on the Shopify App Store