Industry Intelligence
What Shopify's data tells us about the new buyer journey, made in AI's image.
Earlier this week, Kyle Risley and the data team at Shopify published something rare in the AI-search discourse: first-party, aggregated, behavioral data from millions of merchants.
Everyone is excited about AI-referred shoppers converting on Shopify storefronts nearly 50% higher than organic search in Q1 2026, with average order values 14% higher, and AI referral volume growing roughly 13x year-over-year.
For me, and our team at Yolando, the real story lies underneath. The mechanism that explains why we’re seeing this lift in Shopify’s data, and its story has implications well beyond ecommerce. The journey of how customers discover all the things they need to make their lives easier and better. When they do, they're all in as seen in higher average order value.
I joined Yolando as Head of Marketing earlier this year. Consider this post an introduction 👋 come with me as I break down what I believe is the real lead Shopify buried.
Before Yolando, I spent 2015-2018 on the product team at Shopify, where I was the first Product Growth Manager on Channels. These integrations brought Shopify merchants onto Facebook, Instagram, Pinterest, and Google Shopping with more features than any other platform at the time. This work also contributed to Shopify’s marketing-tech backbone, which provides merchants with marketing attribution for orders from Google Analytics and Google Ads.
Now, just as I did then, I have a front-row seat to the way the buyer's digital discovery is shifting. I’m excited to share what I believe this means for marketing and growth teams today.
The mechanism: buyer discovery journey compression
Buried in Risley’s post is the catalyst that’s doing all the work. He noted that the industry is starting to refer to it as "journey compression."
The idea is simple: AI assistants collapse the discovery, research, and consideration phases of the buying journey into a single conversation. The shopper describes what they want, and the model narrows down options, compares, and recommends. The buyer clicks (or, in many cases, zero-clicks) only at the end, when they have already effectively decided.
In Shopify’s data, we can see this by looking at where sessions start. It shared that more than half of AI-referred sessions land directly on a product detail page. Compared to organic search, it has only about 20% of its traffic landing on the product page. AI doesn’t send people to your homepage to browse; it sends them to the product it’s suggesting shoppers buy.
That’s the mechanism. The conversion lift is simply the mechanism as measured at the cash register.
Why the journey shift is more important than conversion
As a former Shopify employee, I understand why conversion rates are the north star metric. For my role as head of marketing at Yolando and more broadly for marketers, it’s how this data shows us that buyers are skipping the traditional funnel journey. This shift is where we should focus.
Conversion rates can fluctuate with creative, pricing, and macro conditions. How LLMs are fundamentally changing the buyer journey, and understanding this is what can determine success for years to come.
Three things we see from journey compression that don’t follow from the conversion lift alone:
Buyer research is happening off websites. When discovery, comparison, and consideration move into ChatGPT, the website stops being where buyers first learn about us. Now it’s largely the surface where they confirm a decision they've already made. Our website analytics will increasingly under-represent how buyers actually evaluate us. The funnel didn't get shorter; the top of it just moved somewhere we don’t own or cannot see.
The homepage isn't the only important surface used for credible evaluation anymore. Evaluation is now done based on what the LLM model retrieves when the buyer asks their question. That includes product pages, but may also third-party reviews, comparison content, Redditor statements, LinkedIn comments, editorial coverage, and other useful structured data the model can parse cleanly (think schema). Most of us have been investing in our owned surfaces, which are now a smaller part of the larger picture that shapes buyer evaluations.
The gap between "considered" and "shortlisted" is collapsing. A traditional buyer searches for "best X," gets 10 blue links, opens 6 tabs, and gradually narrows their selection. In a compressed journey, the LLM does the narrowing before the click. By the time the buyer hits your site, they're not comparing, they're confirming. If you're not in the model's recommendation set, you're likely not in the buyer's consideration set either.
Most directors, VPs, and heads of marketing we talk to right now are sitting with the same questions:
Are we being recommended by AI?
Are we being cited by AI?
Are competitors showing up instead of us?
Are we being described accurately, or is the model getting our value proposition, feature set, and or price wrong?
Underneath all of these questions is the fact that we all have to get comfortable with the idea that we don’t have the same brand of precision on the narrative anymore, and in some cases, it’s hard to know what’s being said at all. The Shopify data is a measured, behavioral answer to why that question feels urgent now, even if the AI-referred traffic on your own dashboard is still small.
The supply side is measuring this. The demand side is building for it.
If you want a check on whether journey compression is real or a Shopify-specific artifact, look at what OpenAI shipped in late November 2025. Its Shopping Research tool generates multi-product buyer’s guides with top products, key differences, and current retailer information all inside a ChatGPT conversation.
The buyer journey compression is productized. LLM platforms, where the journey is happening now, are building specifically for the behavior Shopify is measuring on the other side. Supply side and demand side of the same shift, agree. This is huge, and will very likely shape how we work as marketers for at least the next 5 years.
When one of the largest SaaS companies is measuring AI-referred traffic (Shopify), and the company producing AI-referred traffic is publicly investing in UI/UX for the same behavior change (ChatGPT), we’re not looking at a data anomaly. We’re looking at a fundamental new structure.
How does this look different in B2C and B2B?
Shopify’s data is largely B2C ecommerce, and the mechanism is most visible here because the journey is short, the purchase is discrete, and the conversion event happens on an owned, measurable digital surface.
For instance, a buyer asks Claude for running shoes under $150 with good arch support, gets three options, clicks through, and buys. Journey compression is right there in the session data. The same pattern shows up cleanly in B2C services with high digital discovery—telemedicine, mental health, women’s health, wellness. When a prospective patient asks ChatGPT for a recommendation and arrives on a provider’s site, it is already 80% sold.
In B2B, the same mechanism applies, but it shows up as absence rather than presence. It’s harder to see and more dangerous if marketers ignore it.
Consider a B2B SaaS company in fintech, cybersecurity, legal tech, or adjacent. These are the categories in which buyers do most of their evaluation before requesting a demo. The 2022 version of that journey involved Gartner Magic Quadrants, G2 reviews, three to five vendor websites, demo requests, and weeks of back-and-forth.
Today, that version increasingly starts with “what are the best platforms for X, what are the tradeoffs between the top three, and which one fits a team like ours?” asked on ChatGPT, Claude, or Gemini. The LLM produces a shortlist. The buyer might request a demo from two of those three vendors (traffic buried in direct channel reporting). The third vendor never knew the recommendation existed.
The Shopify conversion lift, translated into B2B, isn’t higher conversion rates on AI-referred traffic. It’s more like retrieval systems are now deciding the shortlist inclusion rate we don’t control, before any session is found in Google Analytics.
The lift exists, but B2B unfortunately struggles to see it using traditional tools like Google Analytics or sophisticated marketing attribution stacks enabled by CDPs. This is what AI search visibility actually means for B2B: it’s not a traffic metric but a presence-or-absence question that comes before any funnel measurement tool we currently trust.
The same shift, a different decade. What the last channel shift teaches us about this one.
In his piece, Risley makes an important point: “Web didn’t kill physical retail. Mobile didn’t kill desktop. Social didn’t kill search. Each one grew alongside what came before it.”
Shopify President Harley Finkelstein made the same argument from a different angle on Shopify’s Q1 2026 earnings call earlier this month, telling investors that AI channels like ChatGPT will be “a tailwind to driving e-commerce growth and penetration over time.” Essentially an additive, not a replacement.
Sometimes when trying to figure out where to put the next dollar, the temptation is to treat the new channel as a zero-sum threat to existing ones, and either over invest defensively or wait for it to “really matter.” Neither move is right. The right move is to recognize early that the channel is additive, measure it as its own thing, and strategically build for it before it forces you to.
I know this because I’ve watched it happen.
When I joined Shopify in 2015 to lead Channels, social commerce was something most merchants (and market analysts) treated as a rounding error. A robust Meta shops integration didn’t exist, and the question was always “how much revenue can it really drive?” But for Shopify merchants, we built early product feeds for Facebook, then Instagram, Pinterest, and Google Shopping. Shopify didn’t invest then because the channels weren't necessarily delivering high order volumes. We did it because the infrastructure was ready and the early data showed that the conversion quality, intent signals, and behavior in beta cohorts indicated it was going to be massive. And it was.
The pattern from that period was clear: Shopify merchants could treat each new channel as a measurable surface from day one, even when volume was small. Those who waited (using other platforms to power their stores) for the volume to justify the work were always one cycle behind, with no data or experience needed for rapid optimization to succeed.
Treating AI referral traffic today as a rounding error, rather than organic search, is exactly how many merchants treated mobile traffic in 2011 and social commerce in 2017. The economics at the per-session level already favor early GEO movers.
What marketing leaders should do now with GEO?
I won’t repeat Risley’s five-point list; it’s good and ecommerce-specific. Here’s the version that applies more broadly to directors, VPs, and heads of marketing, growth, demand-gen, and SEO who own pipeline outcomes—whether you sell sneakers or software.
Measure AI referral as its own channel today, not when the volume justifies it. Most analytics tools now break out referrals from ChatGPT, Perplexity, Claude, Copilot, and Gemini. Pull them out of “other” or “direct” and look at conversion quality on a per-session basis. AI overview tracking is becoming a discipline of its own. Monitoring not just whether AI sends you traffic, but also whether it cites you, recommends you, and frames you accurately, can be done easily with a tool like Yolando. The volume will look small. The per-session economics will not.
Audit how you’re cited, not just how you rank. Blue-link SEO rankings are still an input, and LLMs still retrieve from web search indexes. But the more important question is what the model actually says about you when a buyer in your category asks. Do you appear in the shortlist? Are your differentiators represented accurately? Are competitors getting credit for things you do better?
This is the discipline now called generative engine optimization, alongside the adjacent practice of answer engine optimization (AEO), which affects how you appear in AI Overviews and direct-answer surfaces. It’s the category Yolando works in, and it wasn’t a discipline two years ago. It exists now.
Invest off-domain. Reviews, third-party comparison content, community discussion, and editorial coverage are no longer “PR and reputation” line items. They’re retrieval inputs. The model is reading them on the buyer’s behalf. Brands that have genuine community presence are more likely to have LLMs with the most to say about them.
Structure your owned content for retrieval. Clean HTML, server-side rendering, structured data, comprehensive product and feature pages, and documentation that answers real buyer questions rather than performing brand voice. The principles that make a page useful to a thoughtful reader are increasingly the same principles that make it useful to most LLM retrieval systems.
Keep doing classic SEO. Organic search still drives more sessions than every AI platform combined, and it’s a direct input into AI citation. The fundamentals haven’t changed; they’ve given most sites leverage. This is the additive part Harley and Risley are both pointing at: AI isn’t displacing your organic work, it’s compounding on top of it.
Now is the time to build your GEO advantage.
Shopify’s data indicates a clear, measured, behavioral signal that the buying journey is changing shape, not just changing traffic channels. The conversion lift is exciting for them, but what’s more important is understanding the change today, while we can still see our buyers. In many categories, we increasingly won’t.
The brands that build for AI search visibility today will be in the recommendation set when asked. The brands waiting for the volume to justify it will spend 2027 trying to figure out why their pipeline shrank. There won't be a clean answer in Google Analytics, because the loss happened upstream of the session start.
I’ve watched this exact shape play out once before. Early movers won, and the gap never really closed. That's the part worth taking from this Shopify data.
Sellers who moved early on the last channel shift weren't reckless. They were on time. So is anyone putting GEO infrastructure into their stack now, measuring how AI describes them, monitoring where competitors are positioned, and building the citation surface that decides whether they're shortlisted?
This is why I love working at Yolando. We’re creating a platform that puts GEO infrastructure into any digital surface—quickly and easily. It helps measure what LLMs say about you, and monitors how your competitors might be better positioned to win. It sets teams up with strategic insights and brand-ready content to close gaps. Yolando is your partner in improving the odds you’ll surface on LLMs, which are increasingly where your buyer decides.
The pattern I saw then is the pattern I'm seeing now, and it doesn't reward waiting. It didn't then. It doesn’t now.
