AI-Referred Orders on Shopify: What Is Real and What Is Hype

Quick answer: Shopify’s own data shows AI-referred orders growing nearly 13x year on year in Q1 2026, its CEO said on the August earnings call that AI-driven traffic and orders each tripled in Q2, and its President said on the 3 November call that AI traffic grew 2.6x and orders 3.3x in Q3. Shopify has not published the order counts behind those multiples, and it says organic search still sends more sessions than all tracked AI platforms combined. The conversion lift is real in Shopify’s and Adobe’s data, but a lift on a small slice of traffic moves a store’s total orders only a little (by half of one percent in the worked example below). Find the channel in GA4’s AI Assistant channel and Shopify’s referrer reports, treat every count as a minimum, and put your effort into being citable and fixing product data.

You have probably seen the headline numbers: orders from AI assistants up 13 times, traffic up 8 times, shoppers converting better than people from Google. You have probably also seen the opposite take, that AI shopping is a rounding error. Both can be argued from the same data, which is why neither is much use.

This post is a reality check. I went to the primary pages: Shopify’s published figures and earnings-call remarks, Google’s documentation, Adobe’s retail data and the one payments-industry figure people use for how much buying agents do. I record what each says, what it leaves out, and what it means for an Australian founder with a store doing $150K to $1M a year. Everything is as at 4 November 2026 unless a figure carries its own date, and that includes Shopify’s third-quarter call of 3 November.

One limit up front. None of the Shopify or Adobe figures below is an Australian figure. Adobe’s data covers United States retail sites. The Shopify pages I read give no country breakdown. I will not turn a global or US number into an Australian one.

What Shopify has actually disclosed

Shopify has published AI-referral figures in three places: an enterprise blog post based on its Q1 2026 data, and remarks on the Q2 and Q3 2026 earnings calls. The Q2 press release itself, dated 5 August 2026 and filed with the US Securities and Exchange Commission, mentions AI in the CEO’s quote and the risk language but gives no AI figures. The Q3 press release, dated 3 November 2026 and filed the same way, mentions AI only in its forward-looking and risk language and gives no AI figures either. So when someone says “Shopify reported”, it means a blog post and call remarks, not a financial statement.

From Shopify’s enterprise blog, published 11 May 2026 (Q1 2026 data):

From the Q2 2026 earnings call, 5 August 2026 (CEO remarks, as printed in a third-party transcript; I could not open Shopify’s own recording or transcript, and the transcript site warns of possible errors):

From the Q3 2026 earnings call, 3 November 2026 (remarks by Shopify’s President and CFO, as printed in a third-party transcript that says it may contain errors and that parts were produced with AI tools; I could not open Shopify’s own recording or transcript):

Now the gaps, because they are as informative as the numbers. In the pages I read, Shopify gives no absolute order or session counts for AI referrals, no share of total orders or sessions, and no country split. It does say, in the blog post, that organic search still sends more sessions than all tracked AI platforms combined, and that these signals are early and agentic commerce has not reached mainstream adoption. It also says the true share of AI-mediated commerce is almost certainly higher than referral attribution can show, which I come back to below.

Why “13x” and “3x” are not one story

It is tempting to line the multiples up and read a trend: 13x, then 3x, then 3.3x, so growth is slowing, or levelling out. Or to stack them and read 39x or more. Do neither.

The Q1 figure is orders from AI-referred sessions, from a blog post. The Q2 figure is “AI-driven” traffic and orders, from a call, and the CEO moved between “AI-driven” and “agentic” while describing it. The Q3 figures are “AI traffic” and orders, from the next call, and the speakers also used “AI channels”, “AI-referred” and “agentic” in the same call. I could not find a Shopify document saying these are the same measure. If they are not, the multiples tell you nothing about each other. The same goes for conversion: “nearly 80% higher” in Q2 and “64% more often” in Q3 do not show a shrinking premium, because neither came with a stated baseline. Treat each as a separate, dated statement.

The bigger issue is the base. A multiple tells you how fast something grew, not how big it is. Here is an illustrative example, not a real store:

That store’s AI-referred orders grew 13x (2 to 26). Its total orders did not move, and AI is still under 2% of them. Growth from a very small base is real growth and still a small channel. That is the honest reading of the headline: it says the channel is arriving, and it does not say it is large.

What the conversion and order-value numbers do and do not tell you

Higher conversion and a higher average order value are the most useful findings, and also the easiest to over-read. Four things to keep in mind.

The comparison is against organic search, and in the 23-of-25 cut it is product-page sessions only. More than half of AI-referred sessions start on a product page, against about 20% for organic. A visitor who lands directly on a product has usually already been told what it is. My reading, which Shopify does not say, is that part of the lift may come from where people land, not from AI itself. That does not make the lift fake. It means you cannot assume your whole store will convert like the product-page cut.

It is an average across many stores. Shopify gives one number for all its merchants together. Yours depends on your category, your product pages and whether the assistant is sending the right people.

An independent source shows the lift is recent and can flip. Adobe’s retail data (a report dated 16 April 2026 and a Prime Day post dated 29 June 2026) is based on direct online transactions across more than 1 trillion visits to US retail sites. Its figures, which count AI traffic as shoppers clicking a link and compare it with all non-AI traffic (paid search, email, social and so on):

• In March 2026, AI traffic converted 42% better than non-AI traffic. In March 2025 it converted 38% worse.
• Over the Prime Day event, 23 to 26 June 2026, AI traffic converted 40% better. At the same event the year before, it converted 23% worse.
• The Prime Day post also says the overall volume of AI traffic remains modest compared with other major channels.

Adobe and Shopify use different baselines (all non-AI traffic against organic search only), different regions and different dates, so I am not averaging them. What they share is direction: AI-referred visitors have recently converted better, and a year or so earlier they did not. A premium that appeared in the last twelve months is a finding to watch, not a law.

A lift on a small share barely moves your total. Take a store with 20,000 sessions a month converting at 2%, so 400 orders. Say 1% of sessions, 200, come from AI assistants. If they convert like everyone else (2%), that is 4 orders. If they convert 50% better (3%), it is 6 orders. The lift added 2 orders to 400, or half of one percent. The lift only becomes large when the share becomes large. That is illustrative arithmetic, not a forecast for your store, and it is the reason I would use these figures to justify measuring and preparing, not to forecast revenue.

”AI-referred” is not the same as “an agent bought it”

People also ask how much buying is done by agents themselves. I can give you a careful answer, but first the vocabulary, because the three things people call “AI orders” are different and are measured by different tools.

For ChatGPT, Shopify’s help page says customers complete their purchase on your own online store checkout, in a ChatGPT in-app browser or a new tab on the web. So for ChatGPT, there is a person at your checkout. For Google AI Mode and Gemini, Shopify’s help page says direct checkout fires only server-to-server tracking events, and that Google Analytics and custom pixels do not fire there. That is why orders from that channel can be missing from GA4.

The figure people use for “how much is agent-executed” is 3%. It comes from a Checkout.com press release of 9 June 2026, which says that, according to merchants across the UK and US, only 3% of transactions involve AI agents. Before you repeat it:

• It is what merchants told a payments company, not a count of transactions.
• The release says “involve”, not “complete without a human”. It does not define the word, and it does not state the sample size.
• It is about UK and US merchants in general, not Shopify and not Australia.

So I will not tell you that “97% of the time a human still clicks”. That turns a loosely defined merchant estimate into a precise-sounding claim it cannot carry. What I can say, with sources: Shopify’s blog says agentic commerce has not reached mainstream adoption, its CEO calls agentic volume small relative to Shopify’s total gross merchandise volume, and I found no Shopify or Adobe figure for the share of orders completed by an agent with no human step. Agent-completed buying exists and is being built. The best available evidence is that it is small, and that nobody has published a number you can lean on.

The hype in the other direction

Dismissing this is as sloppy as hyping it, for one reason Shopify itself states: referral attribution undercounts. Shopify’s blog says Google AI Overviews send referrals that standard analytics classify as organic search, and that the real share of AI-mediated commerce is almost certainly higher than referral attribution shows. Google’s documentation agrees on the classification: AI Overviews and AI Mode traffic sits inside Organic Search, and Search Console reports it within the Web search type.

Also, a shopper shown your brand in an AI answer who types your name into Google later arrives as branded search or Direct. No analytics tool links the two. The number you can measure is a floor, and nobody has published a trustworthy figure for how far below the truth it sits. Anyone quoting a precise “hidden AI traffic” percentage is estimating.

Why the volumes are still small for most stores

“Small” is well supported by the evidence above:

• Shopify says organic search still refers more sessions than all tracked AI platforms combined, and its CEO described AI as a complement to search, not a substitute.
• Adobe says AI traffic remains modest compared with other major channels.
• The growth multiples start from a low base, as the example shows.
• Whether an assistant sends anyone to you depends on whether it names your products at all.

One figure points the other way for smaller stores: Shopify’s CEO said 75% of AI-attributed orders in Q2 came from outside its top 100 categories, and on the Q3 call its President said 74% of AI-attributed purchases did. That suggests AI orders are not confined to the biggest categories, but it says nothing about how many orders any one store gets. I will not guess a session count for a store in the $150K to $1M band. Your own data is the only thing that settles it.

How to identify AI referral traffic in GA4 and Shopify

You do not need a tool or a developer. For the custom channel group below you need Editor access to GA4 (Google’s documentation says Editor or above). Our full walk-through is in how to track ChatGPT traffic in GA4 for a Shopify store. This is the short version.

In GA4

1. Check the built-in channel first. Google’s “What’s new” entry dated 13 May 2026 describes a new AI Assistant channel in the Default Channel Group. Google’s channel definitions describe it as visits from sources like ChatGPT, Gemini, Deepseek, Copilot or Grok. Open the Traffic acquisition report and look for an AI Assistant row. Google says the channel excludes Google’s AI Overviews and AI Mode, which count as Organic Search.
2. Look at the raw sources. Change the primary dimension to Session source / medium and search for chatgpt, openai, perplexity, gemini, copilot and claude. Record what you find in a spreadsheet, row by row.
3. Build a custom channel group as a cross-check. The default group cannot be edited, but Google lets you create your own and apply it to reports retroactively. A starting pattern is below. I tested it on 31 October 2026 against sample values: it matches chatgpt.com, chat.openai.com, www.perplexity.ai, claude.ai, gemini.google.com and copilot.microsoft.com, and does not match mail.google.com, notchatgpt.com or m.facebook.com. It is lower case and a starting list of hostnames, not an official one from the assistants, so add whatever you see in step 2.

^(.*\.)?(chatgpt\.com|openai\.com|perplexity\.ai|claude\.ai|gemini\.google\.com|copilot\.microsoft\.com)$

4. Expect some visits to land elsewhere. Google says Direct means traffic without a clear referral source, including links that carry no tags and visits where tracking details were stripped. An AI visit can end up there for the same reasons any visit can.

In Shopify

1. Sessions by referrer. In Analytics, Reports, the Sessions by referrer report shows whether visitors arrive directly, from search or by referral. It counts sessions only and applies to the Online Store channel. Shopify’s blog also says you can filter an existing analytics report by AI answer engines under Referrer Channel, and by Referrer Host for gemini.google.com for Gemini.
2. The Agentic section. For each AI channel you sell on, Shopify’s help page lists four measures: sales, orders, online store sessions and online store conversion, with a default range of the last 30 days. It states that sales data combines referral-based sales and direct-checkout sales, and that these analytics are not yet available for headless stores. This is your best view of orders, because it includes purchases that GA4 may never see. See our guide to Shopify Agentic Storefronts for set-up and eligibility.
3. ShopifyQL, if you use it. The sessions table has a field, agentic_referring_channel, described as the AI agent or assistant that referred the session, with values including ChatGPT, Google AI Mode and Gemini, Microsoft Copilot and Shop. It also has referrer_domain, utm_source and a human-or-bot flag. I have not run a query against your store, so build one from those field names and check that it runs.

Two dated cautions from Shopify’s own help page, as at 31 October 2026. Its session measurement rollout ran from 21 to 23 September 2026, and Shopify says to treat it as a measurement change when you compare sessions and conversion across that period. Its page also refers to changes to the marketing channel dimensions on 18 October 2026. If your date range crosses either, comparisons may mislead.

For orders and revenue, treat Shopify as the source of truth and use GA4 for what visitors do. Our guide to why GA4 and Shopify Analytics don’t match explains the usual gap.

How to read a small number

Once the set-up exists, the risk shifts from “I can’t see it” to “I’m reading too much into six orders”. A few rules:

The 30-session line is my rule of thumb, not a statistical threshold. The point is simple: with few orders, one order moves a conversion rate by several percentage points, so report counts and direction until the numbers are bigger. A tool dashboard that gives a “share of AI traffic” to two decimal places from a handful of sessions is giving you precision it does not have. Our post on what AI visibility tools measure and miss covers where those numbers are sampled or unstable.

What an Australian Shopify founder should do now

None of this calls for a new strategy. It calls for three small jobs and a decision to stop worrying about a fourth.

1. Track the channel, so you know your own number. Do the GA4 and Shopify steps above, record a baseline this month and look at it on the same day each month. Tag the links you control, whoever runs them, so your own email, social and ad traffic does not show up as Direct and muddy the AI numbers. We do not run ads or email, but whoever does should tag them. If you can’t yet explain why GA4 and Shopify disagree on orders, fix that first: every channel’s numbers depend on it.

2. Be citable. An assistant can only point people to a store it can read and has a reason to name.

• Check you are not blocking the crawlers you want. OpenAI’s documentation says OAI-SearchBot is used to surface sites in ChatGPT’s search features, and that sites opted out of it will not be shown in ChatGPT search answers, though they can still appear as navigational links.
• For Google, the documentation says a page must be indexed and eligible to be shown with a snippet, and that there are no additional requirements or special optimisations needed to appear in AI Overviews or AI Mode. That is good news if you have already done the basics, and a reminder that the foundation is ordinary SEO.
• Put shipping times, returns terms and policies in plain text on the page, not only in images. Our guide on how ChatGPT shopping picks products separates what OpenAI documents from guesswork.
• Be wary of claims that one file or tag guarantees citation. Seven GEO myths Shopify founders still believe grades the common ones.

3. Fix your product data. Shopify says products reach its AI channels through Shopify Catalog, and non-product information such as returns and shipping goes in a knowledge base. The CEO said that AI searches powered by Catalog converted at twice the rate of those using scraped data. That is Shopify describing its own product, so weigh it accordingly, but the practical point stands: clean titles, variants, categories and attributes are what an assistant works from. Our guide to Shopify product data for AI shopping agents lists the fields to check, and what Shopify’s AI channels setting does by default covers which channels receive your products.

4. Do not abandon what already brings orders. Shopify’s own framing is that AI complements search. The dull work that makes you findable in Google, such as indexing, collection pages and product copy, is also what feeds AI answers. I would not move budget out of anything that works because of these figures.

On Australia specifically: which AI shopping features are available to Australian shoppers differs by assistant and changes quickly. Our guide to AI shopping in Australia is where I keep that list, as at its own date, rather than repeating it here.

What to re-check before you act on this

Several things here have a short shelf life:

• Shopify’s next results. The Q3 call (3 November 2026) is the latest I have read. When the next quarter’s call is out, check what it says about AI. If the figures changed, trust the newer ones and check whether the measure has changed.
• Google’s AI Assistant channel. Added on 13 May 2026. On the pages I read, Google gives examples of the assistants it recognises, not a full list, so a gap between the built-in channel and your own pattern is normal.
• Adobe’s monthly figures. The conversion premium has moved month to month, and Adobe’s baseline is not Shopify’s.
• Shopify’s reporting. Report names and fields have changed repeatedly in 2026, as the September and October notes above show.

Key takeaways

• Shopify reported nearly 13x growth in AI-referred orders and more than 8x in AI referral sessions for Q1 2026, its CEO said AI-driven traffic and orders each tripled in Q2, and its President said AI traffic grew 2.6x and orders 3.3x in Q3. No order counts, shares or country split were given in the pages I read.
• A multiple is speed, not size. Do not stack or compare the Q1, Q2 and Q3 multiples.
• AI-referred visitors converted better than organic search in Shopify’s data, and Adobe’s US data shows a similar recent premium that was negative a year earlier. On a small share of traffic, the lift barely moves total orders.
• “AI-referred” is not “agent-bought”. The 3% figure is a merchant-reported estimate of transactions that “involve” agents, and cannot be turned into “97% human”.
• Your measured number is a floor: AI Overviews sit in Organic Search, and mentions without a click show up later as Direct or branded search.
• Use GA4’s AI Assistant channel, Shopify’s Sessions by referrer and the Agentic section. Report counts and direction until the numbers are larger.
• Be citable, fix product data, track monthly and keep doing the search work. The newest Shopify call I have read is Q3 (3 November 2026); check for a newer one before you quote any figure here.

FAQs

Is Shopify’s “13x” growth in AI-referred orders real?

It is what Shopify published: AI-referred orders grew nearly 13x year on year in Q1 2026, in a blog post dated 11 May 2026. It is a growth rate, not a size. Shopify gave no order counts or share of total orders, so it cannot tell you how many your store should expect.

Do AI-referred visitors convert better than Google visitors on Shopify?

In Shopify’s data, yes: nearly 50% better for product-page sessions in Q1 2026, nearly 80% better in the CEO’s Q2 remarks, and “64% more often” in the President’s Q3 remarks (none of the three has the same baseline, so do not compare them). Adobe’s US data shows a premium over non-AI traffic too, but AI traffic converted worse a year earlier. Treat it as a recent average, and check your own data once you have enough orders.

How much of my traffic will come from ChatGPT and other AI assistants?

I will not give a figure, because one true for another store tells you nothing about yours. Shopify says organic search still refers more sessions than all tracked AI platforms combined, and Adobe calls the volume modest. Measure your own baseline and compare it with itself.

How do I see AI-referred orders in Shopify and GA4?

In GA4, look for the AI Assistant channel in Traffic acquisition and cross-check with a custom channel group. In Shopify, use the Sessions by referrer report and the Agentic section, which shows sales, orders, sessions and conversion for each AI channel. For orders, treat Shopify as the source of truth, because GA4 may not fire for direct checkout inside some AI channels.

Are AI agents actually buying products for people yet?

Some purchases are completed inside AI surfaces where direct checkout is switched on, but the evidence says it is a small share. For ChatGPT, Shopify says customers finish on your store’s own checkout. The 3% figure that circulates is a merchant-reported estimate of transactions that “involve” AI agents, with no stated definition or sample. I found no published number for fully agent-completed orders.

Should I change my marketing budget because of these numbers?

I would not move budget on these figures alone. The channel is small for most stores, the lift applies to a small share of traffic, and the measurement is incomplete. Set up tracking, fix your product data and make your pages readable, then decide with your own numbers.

Is any of this data specific to Australia?

Not that I found. Shopify’s pages give no country breakdown and Adobe’s data covers United States retail sites. Treat the figures as global or US context, not an Australian forecast.

Want help making your store easier for AI assistants to cite?

Making your store easy for AI assistants to read, cite and recommend is part of our GEO service, delivered alongside our Shopify SEO work. We cannot promise citations or any amount of AI traffic, and I would not trust anyone who does. The easiest place to start is to book a free 30-minute diagnostic or tell us about your store: we look at your store, and at what the AI assistants say about your category, before we recommend anything. No pitch.