The State of Ecommerce

State of E-Commerce: How Shoppers Buy Online

Welcome and Introductions

Dan Buczaczer: Thank you for joining. I'm Dan Buczaczer, the VP of Marketing here at Constructor. Today we're discussing our latest State of E-Commerce report, which is a comprehensive look at how shoppers buy online in 2025.

This report is based on a survey we did just last month in August of 2025 of 1,500 consumers in the United States and Europe. Respondents were balanced across a mix of income levels and age groups. This is the third year that Constructor has done this survey, and this year we actually joined forces with Shopify, which we're extremely excited about.

In the past, we focused the survey exclusively on product search and discovery. But this year, with Shopify on board, we looked at the entire shopping funnel — from finding the right place to shop all the way through to checkout. We saw some significant shifts from even a year ago, so we're really excited to share those with you.

From Constructor, I'd like to introduce the person who has headed up this report since the beginning, Nate Roy.

Nate Roy: Hey, good morning everybody — or good afternoon depending on where you're dialing in from. My name is Nate Roy. I oversee product marketing, brand, and content, and I'm the primary author of the State of E-Commerce report. I've been in commerce technology for most of my career — I was at a company called Salsify if you're familiar with product information management or product data syndication. I've seen this space evolving a lot over the last decade, and I think we're in the middle of a really pivotal time for e-commerce. Super grateful to have Shopify involved this year.

Dan Buczaczer: Speaking of Shopify, we have Solutions Engineer Paiman Parmaei.

Paiman Parmaei: Hello everyone. I'm Paiman. I'm part of the Solution Engineering team focused on the enterprise space at Shopify, specifically on the consumer goods vertical. I've been at Shopify for 5 years. Prior to that, I worked at SAP in a similar role. I couldn't agree more with what Nate said — the industry is changing, AI is changing everything. I think this is a very timely time to talk about this report and its findings.

Dan Buczaczer: Please feel free to ask questions — there's a Q&A button at the bottom of your Zoom. We'll either address them along the way or in a dedicated Q&A section at the end.

How We Structured the Report: Define, Refine, Decide

Dan Buczaczer: We structured the State of E-Commerce report by dividing the shopping journey into three phases: Define, Refine, and Decide. To be clear, this is not a perfectly linear funnel. In fact, this year the data suggests consumers are moving back and forth between these three areas more fluidly than ever before making a purchase. But this is still a useful organizing principle for thinking about the three major phases of buying online.

Phase 1: Define — Where Shoppers Start Their Journey

In the Define stage, consumers have an idea or a goal in mind but may not know exactly what type of product they need yet. They might be actively searching with something specific in mind or just looking for broader inspiration — it's where they first start figuring out what they want or think they need to buy.

Some noteworthy statistics from the survey this year: 68% of shoppers say retailer search needs an upgrade. And very similarly, 66% claim they abandon to Amazon when results disappoint. We're also now seeing Instagram, Facebook, and TikTok all at very similar levels as major discovery sources, with TikTok especially dominant among Gen Z — 58% of Gen Z shoppers say TikTok is a place for product discovery.

68% Say Search Needs an Upgrade — Does That Surprise You?

Dan Buczaczer: Nate, Constructor's in the business of retailer search. Does this finding surprise you?

Nate Roy: It's a little bit of yes and no. What was most surprising, having run this report for a few years in a row, is actually how flat this number is. It was 68% this year and 68% last year as well, which means at least in the perception of the consumer, search hasn't improved in the past year.

Unpacking why that might be: shoppers are getting more comfortable using conversational-style queries in discovery mode, probably because they're getting used to querying things like ChatGPT that way. Especially when their original intent is on the vaguer side. A query like "comfortable shoes for standing all day" could mean sneakers to one shopper, Danskos to a nurse, or insoles to somebody else. Most search engines are going to struggle to bridge that nuance and will probably show the same exact results to all three profiles.

If you dig into the report further, 86% also said they have to reformulate their queries at least once — they type a query, see results they don't like, and go back to the search bar to type something different or more specific. Each time they have to do that is potentially a loss in confidence for that retailer. That's how the perception builds that something like Amazon might just work in a way that retailer sites don't.

The opportunity for retailers is to stop treating search as static keyword match and start implementing AI — especially AI that incorporates in-session shopping behavior to uncover that in-the-moment intent.

Are Shoppers Done with Retailer Websites?

Dan Buczaczer: You mentioned this readiness to flee to Amazon, and we're seeing this increase in social media as a source of inspiration. Are shoppers on their way to being done with retailer websites?

Nate Roy: I want to state as clearly as possible: in my opinion, no, absolutely not. Shoppers are not done with retailer sites. One of the other findings in the report was that 45% said they actually don't care whether a product was recommended to them by a human or by AI, as long as it's the right thing ultimately.

There have been a lot of different points over the last decade or so where a new technology gets introduced and we all start talking about whether the retail site is going to disappear. I remember this happening when social commerce first became a thing with Instagram checkout, Facebook checkout. But retail sites haven't gone anywhere.

Maybe the role has shifted, and maybe shopper expectations for the experience on-site have increased. But every retailer still has loyalists who look to them for guidance, especially in trend-based industries like apparel or beauty. Plenty of first-time visitors will land from a Google search or an LLM search as well. In that moment of Define — when a shopper tells you exactly what they want in their own words in the search bar — that's still one of the most valuable parts of the on-site experience.

The retailers who are most ahead of the game are investing to make their sites true destinations with innovative experiences like their own AI agents or even simply very creatively merchandised collections. Social platforms do a great job of sparking demand, especially with younger consumers, and LLMs will add yet another discovery channel. That doesn't mean retailers lose. It means the websites need to be ready to capture that demand with AI-fueled discovery that feels as seamless and tailored as whatever inspired the shopper in the first place.

Social Media: Friend or Foe?

Dan Buczaczer: Paiman, how should retailers be thinking about social media platforms? Are they their enemy? Are they their friend?

Paiman Parmaei: Shopify's viewpoint has always been that merchants should be where the buyers are. There are some incredible stats around social media usage — 94.5% of internet users use a social media platform on a monthly basis. That's clearly where the buyers are.

That's exactly why we've invested heavily in making integrations with TikTok, Instagram, and YouTube Shopping. Most of the time we've been first whenever these platforms added shopping capabilities. We really see it as a competitive advantage, and we know buyer behavior changes a lot. We saw how TikTok blew up over a short period of time. Now with LLMs, we expect that to happen again. One of the competitive advantages Shopify provides is making it easy for merchants to show up exactly where buyers are, especially as behavior continuously changes.

Phase 2: Refine — Personalization and the Stranger Problem

Dan Buczaczer: In the Refine phase, consumers are thinking, "I know the type of product I want, but now I need help choosing." Some findings: 41% of shoppers say their favorite retailer still treats them like strangers. Only 20% report seeing results that are frequently personalized. Also, 50% of shoppers view sponsored listings with skepticism or frustration, and 25% would pay to avoid ads entirely. On the flip side, 32% have purchased from a sponsored listing.

And in terms of LLMs, 64% of shoppers have reported trying a GenAI tool as part of their shopping journey — that's a 13-percentage-point jump year-over-year. And 58% now say they are very or somewhat comfortable using AI for shopping help.

How Shopify Approaches Personalization

Dan Buczaczer: Paiman, what has Shopify seen retailers do to improve personalization?

Paiman Parmaei: The way we look at this is: in order to provide a personalized experience, you need the right data. Our core belief at Shopify is that a unified data model provides the best opportunity for that personalized experience. Regardless of where your customer is purchasing from — a retail store, a POS system, online through social media, or LLMs — having that single data model is really important. That's exactly what Shopify provides.

So now you have a single view of the customer regardless of where they're coming from. From there, we've invested heavily in customer segmentation. We've invested in Sidekick, which allows you to create customer segmentations by just describing them. You can say, "I want to create a customer segment of everyone who has spent more than $150 on my store," and Sidekick creates that segment for you. The unified data model is the basis for providing that personalized experience — whether it's on the storefront, at checkout, or through personalized loyalty offers.

Why Shoppers Still Feel Like Strangers

Dan Buczaczer: Nate, what's your take on why shoppers are feeling like strangers on the sites they frequent most, and how do we fix that?

Nate Roy: On the technical side, it comes down to two things: how you're tackling relevance and how you're tackling personalization. Most retailers today can deliver pretty decent relevance. If I go on an apparel website and search for black hoodies, I'm more than likely going to see black hoodies. The problem is what shows up first and in what order.

If the order doesn't match what's most interesting to me — if I see women's hoodies first, or hoodies with poor reviews, or ones out of stock in my size — that is not a good result, even though from the eye test it looks relevant. The second issue is personalization. A lot of retailers are still struggling with personalization that feels too simplistic — leaning solely on repeating past purchases or pushing the same brands I've clicked on before.

Two examples from my own life: I recently bought a new iPad and started seeing recommendations for more iPads for three days, even though I'm not buying another for years. Similarly, football season just started. I bought an NFL jersey and started getting recommendations for more jerseys. They could have recommended the new New Era hats or similar accessories that would have made much more sense.

When relevance is that shallow and personalization is just showing your own buyer history back to you like looking in a mirror, even a familiar site feels to the customer like it doesn't know them.

Sponsored Listings: Making Ads Feel Like Service

Dan Buczaczer: What do you make of those numbers around sponsored listings? Is there something retailers can do to improve this?

Nate Roy: The main takeaway is ads do work when they line up with the shopper's intent. They fail when they're just noise layered on top of other noise. What most shoppers want is alignment — product suggestions, search results, both organic and paid, should all be pointing them toward a good result that meets their mission.

The path forward isn't to kill ads. They're a great revenue stream. It's to make them feel less like ads and more like an intentional part of the experience. That requires unifying both on-site and off-site personalization — the search bar, landing pages, and the sponsored slots should all be reading from the same behavioral signals.

Constructor's retail media solution approaches it this way: we hold paid results to the same standard as organic results. Revenue shouldn't compete with customer experience — you should be able to maximize both. Most retail media networks today run with two brains: a shopping brain (your core search engine serving organic results) and a separate paid media brain (your ad platform). Those systems are rarely talking to each other.

The direct result is that shoppers get irrelevant ads. I was literally shopping for work boots a couple of weeks ago and saw an ad for red high heels — completely not even in the same category. That killed trust for me, wasted the ad spend for the brand, and damaged the credibility of that retailer. With Constructor, you can integrate all those signals into one brain where reinforcement learning creates a virtuous cycle — higher relevance on both sides, better performance, greater trust, and of course, revenue.

LLMs and the Shopping Journey

Dan Buczaczer: Paiman, how are you seeing your customers adapt to LLMs?

Paiman Parmaei: When you think about LLMs, it's important to talk about this concept of declarative versus imperative UI. Here's an example we've all experienced: you're about to order something and want to find out the return policy. The imperative experience — what we've been used to — is you scroll down, wonder where it would be, click on buttons, get frustrated, try Ctrl+F, and finally find it. The declarative experience would be: you ask an AI agent, "What is the return policy?" and it replies.

We think we've gotten used to the imperative experience because LLMs didn't exist previously. But it's obviously not a great experience. The declarative experience — where you just declare your intent and the software takes care of it — is going to be more dominant across the board.

At Shopify we're thinking about this both on the operational side and from the buyer side. On the backend, Sidekick is basically an LLM that works across Shopify. It can help with all the tasks you do on the admin side — customer segmentation, reporting. It changes your experience because instead of calling someone with technical knowledge to create a report, you can describe the report you want in plain English. We're seeing a lot of operational efficiency gains from Sidekick. Fun fact: we now allow Sidekick to be full screen, so you can use this LLM experience as the main method you interact with Shopify as the merchant.

On the buyer experience side, we're thinking about it on both a macro level and a store-specific level. On the macro level — Perplexity and similar tools — we are creating an MCP server, the Shopify Catalog. Because of the scale of the platform, we have information across millions of products and millions of merchants. These LLMs can call that directly to get the information they need about any product a buyer is asking about.

Then we've also released the Storefront MCP, which works on a specific store basis. You come to a website, you have the typical experience, but on top of that, you can have an AI agent that helps you choose a specific product. For both of these, we allow you to look at the queries your customers are asking and modify your answers so you show up exactly how you want for your customers. We're giving a lot of flexibility to dictate exactly what these agents are saying.

This space is obviously evolving and Shopify is very much invested in it. We've covered operational efficiency on the admin side with Sidekick, and on the buyer experience with both the massive LLMs and specific storefront agents. We're seeing more and more merchants adopt across the spectrum.

External LLMs: Friend or Enemy?

Dan Buczaczer: Nate, how are retailers thinking about these external LLMs?

Nate Roy: I've heard both approaches, and every retailer has to decide their mindset. If you treat it like an enemy, it will be your enemy. If you treat it like a friend, it'll help you.

If you frame it as an enemy, you might say ChatGPT and these LLMs are siphoning off discovery moments by answering questions directly in their platform and not sending traffic to the site. But if you take a friendly approach, they can become a bridge and actually send back very highly qualified traffic. Usually the traffic coming from the LLM is much more high-intent than it might be from Google — less traffic, but much more qualified.

I don't think external agents and LLMs are stealing customers. It's really just another channel, like apps or marketplaces once were. If you're on Shopify's platform, you can plug in directly. ChatGPT back in April also opened up a waitlist for catalog feeds.

Ultimately, you're probably going to need your own sales agent on your site that can speak to these external agents — so the outside agents know what's in stock, what's trending, what the merchant wants to prioritize. It basically acts as a salesperson, agent to agent.

Inside the retailer's own site, LLMs are even more powerful, and this is something you can implement today. Retailers should be deploying agents on their site that can reason about intent in natural language — taking that query "comfortable shoes for standing all day" and mapping it to sneakers for one shopper, insoles for another, and Danskos for a third.

If traffic originates on another LLM, your site still needs to be the best destination. You don't want a lack of continuity where they have a great experience on ChatGPT and then come to your site and it feels 10 years old. Embrace both strategies: equip your business to communicate seamlessly with external agents (that tech is coming rapidly) and make your own site agent-powered. Constructor offers AI shopping agents and AI product insight agents, and there are other vendors that do this as well. The leaders will be those who treat agents as extensions of their salesforce, not as competitors.

Phase 3: Decide — Trust Signals and the Final Mile

Dan Buczaczer: Now we're at the Decide phase — "I believe I have the right product. Help me confirm it's the right choice." We all know about cart abandonment. Just because they've put something in their cart doesn't mean we're at the end of the road.

Some statistics: 67% say reviews are one of the most important trust signals. We also saw real generational variation — boomers prioritize clear return policies and secure checkout, Gen Z points to influencers and endorsements, and both millennials and Gen Z cite media coverage as important. Additionally, 62% cite high shipping costs as a leading reason for cart abandonment. On the flip side, 64% say one-click or accelerated checkout is somewhat or very important.

Trust Signals and the Power of Shop Pay

Dan Buczaczer: Paiman, I know Shopify has thought a lot about trust signals in this phase.

Paiman Parmaei: This is another area where we can all relate. Sometimes you end up on a website and you're consciously or unconsciously looking for signals: do I trust this merchant?

There's a remarkable stat around Shop Pay, Shopify's accelerated payment method. A third-party consulting firm ran a study across hundreds of thousands of stores and buyers. What they found is that Shop Pay increases lower-funnel conversion rate by 5% even if it's not used. That stat alone tells you that just the familiarity of seeing Shop Pay gives the customer a boost in confidence — "I can trust this website, it has legitimacy."

So Shop is becoming a brand that buyers recognize. We've invested in Shop Promise — showing expected delivery time using Shopify's brand so that builds another trust factor. And making all the information the buyer wants easily accessible throughout the journey is really important. A lot of times when I'm hesitant about buying something, I want to figure out the return policy. If it's hard to find, I lose a bit of trust and might not purchase.

This is another area where we think buyer expectations are going to rapidly change, and a great use of AI is making sure all this information is accessible in the simplest manner possible.

Reviews, Agentic Q&A, and the PDP

Dan Buczaczer: Nate, Paiman talked about return policies and secure checkout. But we also saw that statistic about reviews. What are we seeing at Constructor in terms of how retailers address these final assurances?

Nate Roy: Reviews are obviously a massive piece of the trust puzzle. A lot of that is derived from the fact that they're user-generated — you're getting a regular person's opinion on their experience with the product.

Product detail pages are ultimately where conversion is won or lost. The mechanics you add there — reviews, return policies, transparent shipping information, or even real-time Q&A — have an outsized impact. Shoppers want the assurance that if they have a last-minute doubt, it can be resolved without leaving that page.

At Constructor, one of the things playing a huge role here is agentic Q&A — AI that can answer specific questions on the product detail page about that specific SKU. Sizing (if you're between sizes, do you order the medium or the large?), materials, return rules. It effectively addresses any doubt that might block someone from converting because they couldn't find that information easily. And by the way, it's also pulling in information from user-generated content and reviews, summarizing that to get somebody a quick, clear answer. It turns the PDP into more than a static catalog entry — it becomes interactive and confidence-building.

Accelerated Checkout: The New Expectation

Dan Buczaczer: Paiman, that stat about 64% saying accelerated checkout is important — I don't think I ever would have said I absolutely need this before it existed, but once it did, it drives me crazy when it's not there.

Paiman Parmaei: The majority of checkouts on Shopify now happen through accelerated payment methods. Customers are not going through the experience of entering their credit card information. Across all of Shopify, Shop Pay is the most popular accelerated payment method.

If you think about it, the customer has gone through this whole journey — Define, Refine, and Decide — and they've finally decided to check out. We look at the checkout as a finalizer. The goal should be to make it as seamless as possible and decrease the friction.

An important point: we've talked a lot about loyalty and personalization, and some of those elements obviously happen at checkout. Historically, a lot of accelerated payment methods didn't have access to the customizations you'd make at checkout — you either went through the normal experience with all your configurations and personalizations, or the accelerated option where none of that showed up. Shop Pay actually allows you to show all those personalizations and customizations even through the accelerated experience. We think that's very important.

And a quick note: we now allow Shop Pay to be used off of Shopify too. You don't need to be on Shopify as a platform to use Shop Pay anymore. Given the stats we're seeing with accelerated payment methods, it can be a quick win for improving conversion rate.

Audience Q&A: Reducing Query Reformulation

Dan Buczaczer: Nate, you mentioned shoppers often have to reformulate their query. What can retailers do to reduce that?

Nate Roy: We could spend 45 minutes on this topic alone. What it comes down to is search that understands real shopper intent in the moment. Going beyond keywords is that first initial step — results that match what the person meant, not just what they typed. Then layering in context signals like clickstream data, product attributes, and natural language semantic models and transformers helps ensure that first query gets to the right answer without having to try several different times. Fewer dead ends, faster paths to the right product, which boosts conversion and loyalty.

A good example: we have a piece coming out soon about the query "red lipstick." We looked at that query across 40 million different searches to see what people actually bought, and it was all over the map — colors, shades, and a lot of people didn't even end up buying lipstick. They bought lip gloss, lip balm. A seemingly specific intent like "red lipstick" ended up in the purchase of thousands of different SKUs. That drives the need for both AI combined with in-session clickstream personalization.

Making AI Recommendations Trustworthy

Dan Buczaczer: Nate, you mentioned half of shoppers don't care whether recommendations come from AI or a human. But how do you make sure those AI-driven suggestions are actually relevant or trustworthy?

Nate Roy: They feel most personal when they're adapting in real time to the shopper's actual behavior — what they're clicking on, what they're interacting with — rather than just relying on broad segmentation or generic foundational LLM models. Retailers can build a lot of trust by making suggestions transparent and relevant. Present results — maybe through an AI agent — but also include text explaining why it presented those results and what the user did that made it think this was a good option.

When personalization feels useful, transparent, and respectful, shoppers aren't going to care whether it came from a machine or a human because it mirrors that in-store experience where an associate might say, "I think this is a good fit for you because you said XYZ."

Shopify Sidekick: How Retailers Are Using It

Dan Buczaczer: Paiman, you mentioned Sidekick, the Shopify AI assistant. What are some examples of how retailers are actually using it today?

Paiman Parmaei: I would actually be curious for all of us to reflect on how we're generally using AI day-to-day. Personally, I try to use AI with anything I do and constantly get surprised at its capabilities. That's exactly the trajectory Sidekick is following.

You can have Sidekick full screen — so instead of the Shopify admin, you can be talking to Sidekick through text. We've seen it be very helpful in creating reports, because that's historically very complicated with a lot of clicks needed. Now you can just describe the report and Sidekick creates it for you.

Beyond that, I'd heavily encourage anyone on Shopify to just try using Sidekick as a first pass for whatever task they want to accomplish. It can take action for you and help you achieve goals, but beyond that, it can be your companion. You can ask, "I'm thinking about providing a discount — can you run an analysis to tell me how effective this discount could be?" That was the vision when we released Sidekick: a companion for your day-to-day tasks that you can run ideas by.

When to Intervene Manually vs. Rely on AI

Dan Buczaczer: Another question that just came in: how do we know when to interfere manually versus rely on AI, and what's the best data to help make that decision?

Paiman Parmaei: On the Shopify-specific side, part of that question refers to the worry about buyers asking questions and AI replying with information that isn't truthful or doesn't represent your brand accurately. That's something Shopify has thought about. We released the Knowledge Base app, which creates a report showing you the queries your buyers are asking — whether on Perplexity about your brand or on a chatbot on your website. It constantly shows what type of queries people are asking, what the AI replies, and gives you the ability to customize the answers. We want to empower you to show your brand exactly how you want it to be seen rather than the AI deciding how your brand should be shown.

Nate Roy: I'd love to tack on a couple of thoughts. This is really challenging, and I don't think a lot of platforms today offer good data on this. At Constructor, there are a couple of helpful things. For AI-created merchandising rules, there's a rule performance dashboard that shows the performance of those rules based on whatever KPI is important to you — revenue, conversions, add-to-carts. You can see whether the AI's rule actually improved the metric you care about over a given period.

Conversely, if you decide to intervene and introduce a merchandising rule, there's also a dashboard for merchandised pages to say, "You merchandised this page two weeks ago — have those rules actually helped performance on the KPI you care about most?" This is something all search vendors need to get better at: providing transparency whether it's the AI or the merchant making the rules, showing if it's actually moving you in the direction you want to go so you can make better decisions.

Closing

Dan Buczaczer: We're always available for further questions. I know Shopify is available as well. Feel free to reach out to either of us. Please do download the report — there's a lot more in it than we covered today, including breakouts by generation, by region, and many more statistics. Thank you for joining us, and Nate and Paiman, thank you for being part of this panel. Please join us for the next one.

Nate Roy: Thanks.

Paiman Parmaei: Thanks, everybody.