Product listings now appear directly inside ChatGPT, Google’s AI Mode, and Perplexity, and showing up there requires more than traditional product page optimisation.
Whether you’re building ecommerce sites or running one, the rules are changing. Join Cara Corbett, Filip Ruprich (SUSO’s Head of SEO), and Mack Johnson (Director of Commerce, eHouse Studio) as they break down what ecommerce teams need to build for in 2026.
We cover:
- Why AI visibility needs more than traditional product page optimisation
- How to structure product data for LLMs
- Whether you need to start preparing for agentic commerce
- Measuring your store’s AI readiness
See the recording below.
Click here to access the presentation slides.
Webinar Transcript
00:00:00 –> 00:00:05 Cara Corbett: Welcome, everyone. Really glad you could join us today. I’m Cara Corbett, VP of Partner Growth at SUSO Digital.
00:00:06 –> 00:00:23 Cara Corbett: Thanks for joining us for another webinar — this one’s called Building Discoverable eCommerce Stores for AI-Powered Search. If you joined us on our previous webinar, Building Websites for Humans, Search Engines and AI, you’ll already have some good context as to where this fits in. Today is essentially a spin-off of that one.
00:00:23 –> 00:00:42 Cara Corbett: We covered a lot of ground on web design and development for AI search, and eCommerce really kept coming up as an area that deserves its own dedicated deep dive. Given all the changes and chatter lately around agentic commerce, the challenges are specific enough and the stakes are high enough that we felt it warranted its own session.
00:00:43 –> 00:01:06 Cara Corbett: So today is AI search-friendly web development with a really sharp focus on eCommerce builds. And who better to join me than Filip Ruprich, our Head of SEO, and our VIP guest, Mack Johnson, Director of Commerce at eHouse — a highly regarded Shopify agency. Mack, why don’t you tell us a bit about yourself?
00:01:07 –> 00:01:25 Mack Johnson: Hello, everyone. Great to be included — thank you to the SUSO team for having me. My name is Mack Johnson, I’m the Director of Commerce at eHouse. We’re a Shopify Platinum Partner and we help omni-channel brands move to, build on, and grow on Shopify at scale.
00:01:26 –> 00:01:58 Mack Johnson: I used to be a developer and I’ve been working on the Shopify platform for about fifteen years now, though I’m not allowed to build things anymore — I have a much more capable group of people who do it with me. I spend a lot more time talking about building things than actually building them. But a big focus of ours over the past eighteen to twenty-four months, especially when it comes to customer acquisition, has been AI: preparing brands to navigate the changes while making the best use of Shopify natively as a platform.
00:02:00 –> 00:02:18 Mack Johnson: I’ll preface this by saying I’m very much Canadian, so if you need me to repeat anything in conventional English, just let me know. I’m really excited to share a bit more about how we’re working with brands to embrace the changing nature of how customers are discovering and engaging with them.
00:02:18 –> 00:02:27 Cara Corbett: Awesome. And yeah, Mack, don’t worry — a lot of our partners are based in the US and the UK, so you will be well understood.
00:02:28 –> 00:03:09 Cara Corbett: Here’s what we’re covering today. We’ll start with the big picture — what’s actually changed in eCommerce and why it matters right now. Then we’ll get into the technical blueprint, which is really the bulk of the session: the things that directly determine whether a store shows up in AI-powered search. We’ll look at it from a Google and product schema lens first, and then Mack will get into the nitty gritty from the Shopify perspective. Then we’ll cover measuring AI readiness, and as always, we close with Q&A. Feel free to drop questions in the chat as we go — a couple of you have already been doing that, and we’ll get to those at the end. We’ll also be sharing the slides after the session.
00:03:10 –> 00:03:53 Cara Corbett: Let’s start with why we’re all here. If you’ve been wondering whether the AI shopping shift is real or just hype, these numbers should settle that. Adobe reported 693% year-on-year growth in retail traffic referred from generative AI platforms during the 2025 holiday season, and a 254% increase in revenue from AI-referred traffic in that same period. Shopify also shared that AI-referred shoppers converted at around 50% higher rates than organic search on product pages. We’re dealing with a potentially much more qualified, ready-to-buy set of visitors in the AI era. The shift isn’t coming — it’s here, and the stores set up for it are already pulling ahead.
00:03:54 –> 00:04:39 Cara Corbett: What does this actually look like? Because it looks a little different on every platform — and this is really where it gets real. On Google’s AI Mode, when someone searches for a product now, they’re not just getting a list of blue links anymore — they’re getting AI-generated summaries with product comparisons and top lists embedded right in the answer. On ChatGPT’s Shopping Assistant, users are having full conversational research sessions, comparing products and getting recommendations without leaving the chat. And on Perplexity Shopping, there are product carousels with very clearly sourced results. This is the new shelf — this is where product discovery is happening. And if a client’s products aren’t showing up here when buyers are using these tools for research or purchases, they’re simply not going to be in that consideration set.
00:04:40 –> 00:05:17 Cara Corbett: When a user is ready to buy and types a commercial query into Google, what they get is product grids with advanced filtering front and centre, pushing organic blue links and even AI Overviews way down the page. That makes it clear that traditional SEO alone is not enough at the commercial stage. The product data layer matters as much as the page itself. You can have a perfectly optimised product page and still lose out because the structured product data feeding into Google Shopping surfaces isn’t clean. And that’s exactly what we’re here to talk about today.
00:05:18 –> 00:05:36 Cara Corbett: There are really three things that determine whether an eCommerce store shows up in AI-powered search: AI bot access, structured data, and product feed optimisation. Everything else builds on these. Filip, I’ll pass it over to you to walk us through each one in detail.
00:05:38 –> 00:06:04 Filip Ruprich: Thanks, Cara. Let’s start with Pillar 1: AI bot access. Most AI shopping services — including ChatGPT product carousels — depend on content being publicly crawlable and accessible to AI systems. Crawlability is the foundation of AI product discovery. If AI bots can’t reach a page, they can’t understand it, retrieve it, or surface its products.
00:06:05 –> 00:06:45 Filip Ruprich: The instinct is to go straight to robots.txt — and yes, that’s the starting point — but it’s not the full picture. We often find that robots.txt allows the right bots, but access is still blocked: sometimes through Cloudflare rules, sometimes at the hosting level, and we’ve also found staging restrictions left in production. The crawl path is layered: DNS, hosting provider, CDN, robots.txt, and finally the product page. Any one of those layers can become a silent blocker. That’s why bot access auditing has to go much deeper than simply checking robots.txt.
00:06:47 –> 00:07:38 Filip Ruprich: Now let’s move to Pillar 2: structured data. I want to be upfront here because recent research has changed how we should think about this. Google’s AI Optimisation Guide states that schema is not required to appear in AI results, and Ahrefs found that schema markup doesn’t significantly boost AI citations. So why are we still talking about product schema? Because for eCommerce, it’s not really about citations — it’s about product visibility. Schema supports shopping experiences with product, pricing, and availability data, and those shopping surfaces are increasingly connected to AI-powered discovery. The bottom line: don’t implement product schema expecting more AI citations. Implement it because AI shopping experiences rely on those product signals.
00:07:41 –> 00:08:28 Filip Ruprich: If you’re implementing product schema, there are certain properties worth prioritising. The basics — name, image, SKU, and offers — are your foundation. Ratings and reviews add trust and improve how products appear in search. Then there are properties that are often overlooked: isSimilarTo, hasVariant, and isAccessoryOrSparePartFor. These help AI understand product relationships — what’s similar, what goes together, and what works as an alternative. And finally, FAQ schema or well-structured Q&A content — not because it drives AI citations directly, but because AI can easily understand and reuse it.
00:08:31 –> 00:09:24 Filip Ruprich: This is where things are heading: the Universal Commerce Protocol, or UCP. UCP is Google’s open eCommerce standard that lets AI agents find products, apply discounts, and complete purchases — so they’re not just reading product information, they’re supporting transactions as well. Pricing, availability, fulfilment, and returns all become machine-readable. The key thing to watch is that UCP started in AI Mode but is already expanding into mainstream Google Shopping experiences. And importantly, UCP builds on product schema — getting your structured data right today is really the foundation for UCP readiness. You don’t need to implement UCP yet; the standard is still evolving. But stores with clean, structured eCommerce data will be best positioned as adoption grows.
00:09:27 –> 00:10:06 Filip Ruprich: One thing worth making clear: UCP is not something that automatically applies to your store. It requires concrete steps to become eligible. The high-level path involves getting your Merchant Center in order, setting up Google Pay, building the checkout integration, and syncing order updates back to Google. Access isn’t open yet — you need to join a waitlist and be approved by Google. If you’re thinking about this for your clients, the time to apply is sooner rather than later.
00:10:09 –> 00:11:13 Filip Ruprich: The starting point for UCP — and honestly just for good eCommerce performance generally — is having Merchant Center in good standing: clean feeds, no errors, products approved, everything up to date. That’s the baseline. For UCP specifically, the main addition at the product level is a new attribute called native_commerce. You need to add it to each product you want to make eligible for agentic checkout — without it, that product will be excluded. But UCP isn’t a fit for every store. Many product types — including subscriptions, digital goods, personalised items, and pre-orders — are still excluded. And consumer adoption is still early: many shoppers still prefer to complete purchases on the website. Even ChatGPT rolled back its in-chat purchasing feature after limited adoption. Focus on strong data foundations now — full UCP implementation can come later.
00:11:16 –> 00:12:03 Filip Ruprich: Let’s move to Pillar 3: Google Merchant Center. GMC is no longer just a shopping ads platform — it’s becoming an AI data layer. AI systems are increasingly verifying stock levels, pricing, and shipping speed directly from product feeds, without the user ever visiting the website. That means feed quality matters more than ever. If stock levels are wrong, product data is incomplete, or pricing isn’t up to date, AI may simply stop showing your product. The key optimisation areas are: accurate inventory, competitive pricing, fast shipping, and clean product attributes. All of these help AI better understand and trust your product data.
00:12:04 –> 00:12:17 Filip Ruprich: One feature worth watching is Google’s upcoming AI Performance Insights in Merchant Center. It’s not fully available yet, but it will provide better visibility into product performance across AI-generated answers.
00:12:20 –> 00:13:13 Filip Ruprich: I’m particularly excited about these new conversational feed attributes that Google just announced — most stores haven’t implemented them yet. These are extra feed attributes that sit on top of your standard Merchant Center feed and give AI systems more context about your products. The most valuable is probably question_and_answer, because you’re effectively feeding AI the exact questions your customers may ask and the answers you want it to give — things like “Is this waterproof?” or “How long does the battery last?” That’s incredibly useful for conversational shopping. Other attributes help AI understand product relationships, documentation, variants, and popularity signals. They can be submitted via a supplemental feed or the Merchant API — a small addition that can significantly improve how AI understands your catalogue.
00:13:13 –> 00:13:26 Cara Corbett: Awesome, thanks Filip. So we’ve just covered the three pillars from a theory standpoint. Now I’m turning it over to Mack to show us what this looks like in the Shopify environment specifically. Mack, over to you.
00:13:27 –> 00:13:52 Mack Johnson: Great, thanks Filip. I’ll be honest — I hadn’t heard Filip’s talk track before we got on this call, and I’m very happy with how naturally it segues into what I want to share today. Thanks for the alley-oop, Filip. I work here at eHouse on the client solution side of things and I’ve spent the better part of the past two years almost exclusively talking to people about AI.
00:13:53 –> 00:14:43 Mack Johnson: I’ll say this upfront: I’m a definite AI sceptic, and I truly believe there’s still a lot of noise around it. But one thing I’ve really grown to accept is that AI is absolutely impacting how customers are discovering, engaging with, and ultimately transacting with brands today. It would be borderline malpractice for me, in the commerce space, to simply ignore AI’s impact just because I’m getting tired of talking about it. So at eHouse we’ve made it a principal focus — giving all the brands we work with the tools they need to keep pace with the rapidly changing role AI is playing in the buying journey.
00:14:44 –> 00:15:25 Mack Johnson: I was at NRF earlier this year where Shopify and Google announced UCP — giving brands a framework by which to engage with users in chat in a much more bespoke way. At the same time, Shopify announced its dedicated agentic selling channel. And for the next sixty to ninety days, all the talk was around the death of the conventional website, the inevitability of all future transactions happening in chat between anonymous buying and selling agents, and just how much needed to change from a technology level for brands to keep up.
00:15:25 –> 00:16:27 Mack Johnson: Instead — and I’m not the only one who uses this analogy — AI seems to be going in the direction of the dot-com boom of the nineties. Just as the internet’s role narrowed significantly from those early days when we wanted to install connectivity into absolutely everything, we’re seeing the same thing happen with AI. It’s clearly very good at some things and, at least for now, not so great at others. The impact on commerce may not be as earth-shatteringly fundamental as invisible robots negotiating on your behalf to buy your next pair of shoes — but it is going to be very significant.
00:16:29 –> 00:16:46 Mack Johnson: We’re about six months after that grand announcement, and I’m trying to figure out what remains universally true. As Cara and Filip already alluded to with some very impressive figures — AI is a powerful product discovery tool, and that is not changing.
00:16:47 –> 00:17:17 Mack Johnson: To add some more stats to the conversation: sixty percent of Google searches today are ending without a click, and in Google’s AI Mode that number is reportedly over ninety percent. What that’s indicative of is that the channel a lot of brands are budgeting for is structurally declining, while AI itself becomes search — and what search looks like is based on entirely different queries than it has been over the past fifteen years.
00:17:18 –> 00:17:52 Mack Johnson: Our challenge is: how do we stay part of that discussion? And the very short, seriously unsexy answer is data. Organised data and schemas that reduce the need for AI to infer and do the stuff it hates — guesswork — and follow a repeatable and clear product taxonomy. I want to talk about some tools available to both Shopify and non-Shopify brands to help organise and validate their product data and identify some very common gaps.
00:17:53 –> 00:18:17 Mack Johnson: Getting into what this actually looks like inside Shopify: the gap between “yeah, we’ve got product data” and “our product data is actually working for us in AI, and we know it is” is where most brands are sitting right now — and they don’t know it.
00:18:18 –> 00:19:16 Mack Johnson: Starting with the agentic side: Shopify now offers what they call an agentic plan. You don’t need a full Shopify storefront to participate in it. As long as your products are in the catalogue and you meet the eligibility requirements, you can be listed and recommended across AI channels and in the Shop app without running your entire commerce operation on Shopify. That’s very attractive to a lot of brands. If you’re on other platforms and have been watching the agentic conversation from the sidelines, the barrier to entry is lower than it’s ever been — you can participate in AI-driven product discovery without requiring a full replatforming of your business.
00:19:17 –> 00:20:19 Mack Johnson: If you’re already on Shopify, you’re probably closer to a strong state than you’d think. As of May 1st this year, agentic sales channels are automatically enabled on all live Shopify stores — you don’t have to set anything up. Shopify is treating this like a core sales channel alongside the traditional online store, retail POS, and B2B channels. The agentic dashboard gives you a view into how your products are performing across ChatGPT and Copilot — you can see sessions, revenue attributed to those channels, and run test queries to see how your products actually appear in an AI chat search. It’ll give you a score, quantify it, and proactively highlight deficiencies relative to the top ten results in Shopify’s global catalogue.
00:20:20 –> 00:21:25 Mack Johnson: Shopify has a catalogue query developer tool where you can test specific prompts and see where you rank relative to the global Shopify catalogue. You type in a query the way a shopper would, and Shopify will tell you whether your products rank in the top ten for that specific prompt, and if not, exactly why they don’t. If I were running an apparel brand selling five-panel hats, I could query the catalogue for “five-panel hats under $25” and see the actual JSON output of how my products show up, whether they appeared in the top ten, and what the competing products that did appear look like.
00:21:26 –> 00:21:55 Mack Johnson: Natively, you can also run completeness checks. The four baseline factors for showing up are description length, image count, review volume, and average rating — plus whether your store policies are in place. Review volume and rating are very good examples of the organic, trust-based signals that AI looks for when prioritising results.
00:21:56 –> 00:22:30 Mack Johnson: What we consistently find at eHouse when we run these audits is that most stores are passing on descriptions and policies, but really falling short on images and reviews. A single product image and a handful of reviews — or no reviews at all — will consistently knock products out of the top ten, even when everything else is relatively solid. These are fixable problems, and the agentic dashboard will tell you exactly which products need attention.
00:22:31 –> 00:23:29 Mack Johnson: The dashboard also connects directly back to what we’re about to cover on metadata. You can see in real time what data Shopify is sending to the catalogue for each product, what fields are populated, what’s missing, and whatever AI is inferring — which to me is a very bad word. If your data is thin, the inference is going to be narrow and thin too. The agentic dashboard is where you can see the downstream effects of those gaps against real actual queries. And the most ironic part: when it comes time to think up a whole set of test queries relative to your business, you’ll find yourself turning to AI.
00:23:31 –> 00:24:30 Mack Johnson: All that being said — why does this matter before agentic commerce has fully arrived? The window to get ahead of it is right now. The brands that are going to win in AI-referred traffic are the ones whose product data is already clean, structured, and complete when that traffic volume and the context of the agentic conversation shifts. The numbers Cara opened with — 700% growth in AI-referred retail traffic, 50% higher conversion rate — that’s something that’s already happening, not a projection for the future. I wanted to go just one level deeper and look at what’s actually happening inside the Shopify catalogue itself.
00:24:38 –> 00:25:29 Mack Johnson: I want to talk more about what the Shopify catalogue actually is, because it’s the foundation everything sits on. The Shopify catalogue is essentially a standardised, machine-readable index of every eligible product across Shopify’s merchant base. It’s what AI queries when surfacing product recommendations in real time — pricing, availability, product options and variants, product relationships — all of it gets pulled directly from the Shopify catalogue. If your data is in there and accurate, AI platforms can find you and recommend you. If it’s not, or if it’s messy or disorganised, you’re either going to be invisible, or misrepresented — and misrepresentation can at times be even worse.
00:25:30 –> 00:26:15 Mack Johnson: The important thing to understand is that the Shopify catalogue isn’t a separate system you have to go and build yourself. If you’re on Shopify, your products are automatically included. But when it comes to actual differentiation, the question isn’t whether you’re in the catalogue — it’s whether what’s in there is actually working. And Shopify does give you a head start here. If you’re running a native Shopify store, a lot of the foundational product schema gets generated automatically — your product title, description, and category feed directly into that catalogue representation. But the default mapping is fine. It’s just not great.
00:26:16 –> 00:27:23 Mack Johnson: By default, Shopify takes your standard product title, description, and category and bundles all that up into the catalogue representation. And that’s what an AI agent is going to see. The problem is that most product descriptions were previously written for a human browsing a traditional product page — they weren’t written to answer the conversational queries AI platforms are now fielding. Things like “blue jeans under $100 with free shipping that’ll arrive in two days” or “a cosmetic bag as a gift for a work acquaintance.” Those obviously aren’t the traditional keyword searches we’re accustomed to — they’re intent-driven questions. If your product data doesn’t contain the right signals, you won’t show up in a response to any of those queries. That’s why appropriate data mapping has become so important.
00:27:25 –> 00:28:42 Mack Johnson: When data is incomplete, Shopify doesn’t just leave the fields blank — it attempts to infer using AI. It pulls from whatever product data you provided and makes its best guess at things like your unique selling point, top features, and even technical specifications. And while that sounds helpful, inferred data is a guess. If your inputs are vague, generic, or optimised for a completely different purpose, the inference will be equally vague. The difference between a product that shows up for “cosmetic bag as a gift” and one that doesn’t isn’t the product itself — it’s whether the catalogue entry contains the word “gift,” or signals like “perfect for gifting” or “gifts under $50.” If that language isn’t in your data, it’s not being inferred, and you don’t show up for that query. Thoughtfully mapped data puts you in control, rather than leaving it to chance.
00:28:53 –> 00:29:28 Mack Johnson: Coming back to the catalogue query tool: this lets you run real search queries against your store’s catalogue to see exactly what AI is reading back. What often comes back for a brand starting out is that AI-inferred version of your catalogue entries. If those fields look vague or off when you’re reviewing them, that’s your signal that the underlying product data needs work. It’s the clearest diagnostic tool we have for understanding the gap between what a merchant thinks their catalogue says and what an AI agent is actually seeing.
00:29:30 –> 00:30:25 Mack Johnson: The last thing I want to cover is some of the common pitfalls we see at eHouse when running audits, because these come up consistently. The first is empty category metafields. Shopify generates suggested metafield values automatically when you assign a product to a category taxonomy — colour, materials, style, and other product-specific attributes. Those suggestions just sit there until someone accepts them. What we see most often is that the taxonomy has been assigned but the metafields are all still empty — a significant missed opportunity, because those fields are exactly what help AI understand the specific context around a product. The analogy I always use: the difference between “what is the name of this product?” and “what is this product?”
00:30:26 –> 00:31:11 Mack Johnson: The second is mismatched or vague product descriptions. The description field in an AI context is intended to be a short, clear blurb about the product itself. Most stores are mapping this to their full product page description by default, which is often long and focused on persuasion and branded messaging rather than clarity. If you have a short description metafield available, that’s probably a better mapping. If you don’t, it’s worth creating one specifically for catalogue purposes.
00:31:12 –> 00:32:11 Mack Johnson: Third, for larger marketplace-style sites: inconsistent or non-standard product titles. The catalogue relies on titles to understand what a product is and how it relates to others. If your titles follow an internal naming convention that wouldn’t make intuitive sense to an AI parsing thousands of products, that creates more guesswork on the AI side. Normalised, descriptive titles consistently perform better. And if changing titles would cause conflicts downstream, you can map the catalogue title to a separate metafield for this specific purpose and keep your live website titles untouched.
00:32:13 –> 00:32:57 Mack Johnson: Fourth is fragmented product variant structures. If your products have variants, the way variants are grouped together matters a lot. Stores with non-standard setups — custom grouping logic, unusual option labels, products that should be combined listings but aren’t — see inconsistent catalogue behaviour. If you have individual products for every colour of the same item, you’re going to see problems. Standard product and variant structures have a clear advantage because Shopify’s indexing is optimised to understand them.
00:32:58 –> 00:33:57 Mack Johnson: And the last one — which might seem obvious but is still very rampant — is missing intent keywords in your data. AI catalogue queries are semantic and conversational. If a shopper searches for “gift for a crafter” and your products don’t contain any language around gifting, you won’t show up regardless of how great your product is. Those keywords can live in descriptions, in grouping metafields, in tags — they don’t need to be in titles, but they need to be somewhere in the data the catalogue is reading. What Shopify gives you in its native data mapping tool is the ability to override defaults and point any of these fields at a metafield instead. That means you don’t have to modify anything on the front end of your store or touch content that’s been optimised for SEO — you’re just giving the catalogue a cleaner, more intentional signal specifically for AI.
00:33:58 –> 00:35:12 Mack Johnson: One stat worth mentioning: the overlap between top-ten Google rankings and AI citations went from around 75% in mid-2025 to around 17% by early 2026. High organic rankings on Google now have less and less of a relationship with AI visibility — these are different lists entirely. All that to say: this type of data mapping is still new to a lot of brands, but it isn’t hard to get started. Shopify has already done a lot of the heavy lifting — you just have to configure it. And the exciting part is that many stores and brands haven’t done this yet. You can still very much be a first mover here.
00:35:15 –> 00:35:48 Cara Corbett: I want to drill into the inference piece, because I think it’s really interesting. Something we hear a lot about in AI search is hallucination — especially on platforms where people are asking about brands or products. Is it the same case within Shopify? If someone asks a specific question about a product and that data isn’t in there, will it make things up about your brand if the product specs aren’t there?
00:35:49 –> 00:36:40 Mack Johnson: I think there are two silos to consider. One is how Shopify is massaging your catalogue in advance of you being part of that AI conversation — that’s where it creates inferences around what your description and product taxonomy actually are. And that is going to have an adverse impact on whether you appear and are represented accurately. There’s a real chance that inaccuracies happen in both environments and are exacerbated by each other. And that’s why having a very high degree of specificity in your product data — as it’s ingested by AI in the Shopify catalogue — is so important. It really is a cascading effect.
00:36:42 –> 00:37:45 Mack Johnson: I spoke about this at an event a couple of weeks ago, and an attendee asked a great question: given that how AI understands data seems to be changing every three to four months, what’s the point of trying to keep up? And for me, if you’ve always prioritised clean, clear, and concise data, you’re never going to be behind. This is basic housekeeping for a brand — it’s always been a huge asset. The more specific information you can tie back to your product, your brand, and your product taxonomy, the clearer the results are going to be, and the more AI is going to favour you because it doesn’t have to do all that guesswork.
00:37:51 –> 00:38:20 Filip Ruprich: Before we move into measurement, I also want to mention one area that’s often overlooked: images. For eCommerce, visual search is an important discovery channel. Tools like Google Lens and AI-powered shopping results rely heavily on image signals, and the difference between a poorly optimised image and a well-optimised one can be significant.
00:38:21 –> 00:38:53 Filip Ruprich: The basics are straightforward: use descriptive alt text, use high-quality unique images rather than stock photos, include multiple angles where possible, and make sure your image property is included in product schema. File names matter too — IMG_4832.jpg doesn’t tell AI anything, but a descriptive file name does. Individually these are small changes, but together they’re often the fastest AI wins we find in eCommerce audits.
00:38:54 –> 00:39:20 Cara Corbett: Awesome, thanks Filip. So that’s the technical blueprint covered — Filip and Mack have given you a lot to work with there. Now we’re shifting into the third part of today’s session: measurement. Because once you’ve built for AI search, the obvious next question is how do you know if it’s actually working? We’ll keep this practical: three concrete steps you can take right now, plus a look at what’s coming. Filip.
00:39:21 –> 00:39:51 Filip Ruprich: Let’s start with Step 1: auditing bot access. I want to come back to something we discussed earlier — this is about much more than robots.txt. Yes, that’s the first thing to check, and you should make sure major AI crawlers are allowed. But then you need to look at the layers behind it: Cloudflare and CDN rules, or hosting-level restrictions. Any one of these can silently block access even when robots.txt looks perfectly fine — and we see that quite often.
00:39:52 –> 00:40:09 Filip Ruprich: A useful starting point is SUSO’s free AI Search Visibility Checker. It shows how visible a brand is in AI-generated answers and can highlight accessibility issues before you dive into a deeper technical audit. Think of it as a quick health check before the full investigation.
00:40:12 –> 00:40:35 Filip Ruprich: Step 2 is the shopping reports in Search Console. These act as a health check for your product’s structured data — highlighting errors, warnings, and opportunities to improve your product schema and merchant listings. This isn’t AI-specific reporting, but keeping it clean and error-free improves the chances of your structured data being used in both Google Search and AI experiences.
00:40:38 –> 00:41:13 Filip Ruprich: Filtering performance by search appearance in Search Console shows clicks and impressions for different rich result types. For eCommerce, focus on product snippets, merchant listings, and review snippets. AI traffic can’t be separated from regular search here, but the data is still useful: higher CTRs often indicate structured data is working well, while page and query insights help optimise titles, descriptions, and attributes. Google is also testing a generative AI performance report in Search Console, so keep an eye on future updates.
00:41:13 –> 00:41:47 Filip Ruprich: One broader point on measurement: right now, measuring AI visibility is genuinely difficult. For Google, AI traffic is buried in general performance data. For other tools like ChatGPT, Perplexity, or Claude, you can track referral traffic in Google Analytics — but measuring impressions and visibility is very hard. You can monitor selected prompts manually or with dedicated tools, but AI responses vary by user, location, and context. Visibility tracking is still far from perfect.
00:41:50 –> 00:42:40 Filip Ruprich: Step 3 is the product analytics report in Merchant Center. Ads and organic are mixed together by default, but you can drill down to organic specifically. You can see clicks, impressions, CTR, purchases, and purchase rate individually for specific products and by brand. Products trending up are the ones to keep in stock and error-free. High impressions with low CTR usually means the image, title, or price needs work. Trending down is worth catching early. This data isn’t AI-only yet, but it does include data for Google’s AI Overview and AI Mode — so this report should get significantly more useful once Google finishes rolling out AI Performance Insights.
00:42:41 –> 00:43:32 Filip Ruprich: Speaking of which — Google’s AI Performance Insights in Merchant Center is going to give merchants direct visibility into how products perform across Google’s AI surfaces: AI Overview, AI Mode, and the Gemini app. It’ll report on share of voice, showing where your brand’s visibility sits compared to similar brands; Shopping Funnel Performance across discovery, evaluation, and purchase stages; Product Term Insights, showing which search terms are triggering your products in AI; Product Attribution Insights, showing which attributes are appearing most often; and a completeness score to help identify which products are missing crucial structured details. It seems like Google is taking a lot of notes from Shopify. It’s not fully live yet, but it’s rolling out over the coming months.
00:43:37 –> 00:43:44 Filip Ruprich: And Shopify now has its own agentic readiness assessment — so it’s worth running any Shopify client through it to see where they stand.
00:43:46 –> 00:44:36 Cara Corbett: One last thing before we move to questions. If today’s session was useful and you want to keep this kind of access and education going, that’s exactly what SUSO’s free Partner Club is for. It’s a free resource for agencies and consultants where we give you access to private workshops and training, a dedicated SEO and GEO team for questions, and bespoke analyses and reports for your clients to help start sales conversations. If you’re building or maintaining eCommerce sites and want a specialist partner for SEO, GEO, or help navigating the AI search world, this is a free and easy way to get started.
00:44:38 –> 00:45:14 Cara Corbett: Before we wrap up, a few things worth bookmarking from today. Filip mentioned SUSO’s free AI Search Visibility Checker — a quick way to get an initial read on how visible your clients are in AI-generated answers and to identify potential technical blockers. It’s self-serve, you get a response in about three to five minutes, and we look at over a hundred factors. It’s on our website — you can also scan the QR code on the slide. And second — Mack, eHouse does an AI readiness audit. Do you want to say a quick word about that?
00:45:14 –> 00:45:37 Mack Johnson: Sure. Everything I’ve walked you through today is exactly what we look at when we run an AI readiness audit for a brand. It’s something eHouse built specifically for this moment in time — a baseline assessment of how well your store is set up to be found, understood, and recommended by AI.
00:45:38 –> 00:46:26 Mack Johnson: What you get out of it is two things: an AI visibility scorecard, using a combination of internal tooling and a variety of other sources, which shows where your catalogue stands across the signals we’ve been discussing today — and what we’ve called a hallucination report, which is exactly what it sounds like. It’s proof of how AI platforms are currently misrepresenting your key products based on the data you have in place right now. The second one tends to get people’s attention, because it’s one thing to know your product data could be better — it’s another to actually see an AI agent confidently describing your product incorrectly to a shopper who’s ready to buy. If that’s something you’d like to look at for your store, you can request an audit through the QR code on the slide.
00:46:32 –> 00:47:14 Cara Corbett: SUSO’s Partner Club is also free to join — we’ll send out the recording and the presentation slides so you’ll always have access to those. We’ve got about twelve minutes for questions. We’re also on LinkedIn Live. We’ve got an expert on Shopify on the call, so let’s give everyone a couple of minutes.
Q&A
00:47:23 –> 00:47:32 Cara Corbett: Jamie asks: if a customer completes a purchase inside AI Mode without ever landing on the site, what happens to that customer relationship? Do you still get their data?
00:47:35 –> 00:47:57 Filip Ruprich: Great question. The good news is yes — you will receive a portion of the data: the order and shipping information, and everything needed to process the order.
00:47:58 –> 00:48:24 Filip Ruprich: But at the same time, you won’t have any analytics data from the source — how long the customer spent in ChatGPT or AI Mode, pixel data, or the broader analytics layer we’re typically used to having. So that’s a limitation. But the basic data will be there.
00:48:25 –> 00:48:37 Filip Ruprich: Mack, I know Shopify specifically has partnerships with some of these LLMs — does that give any deeper access?
00:48:38 –> 00:48:52 Mack Johnson: Still the same picture Filip described — you’re getting the order and the basic customer information needed to fulfil it.
00:48:53 –> 00:50:00 Mack Johnson: Buying directly in chat is still very much a developing concept, and it’s not an apples-to-apples transaction. Some product types lend themselves well to transacting within chat. Others are far too complex — subscription-based engagements, product customisation, loyalty programme redemption — these all become much harder in that context.
00:50:01 –> 00:51:09 Mack Johnson: That said, there are verticals that will genuinely benefit. One I’m very excited about is grocery and food. Meal planning is something AI is actually quite good at — if you could say “here’s what’s in my fridge, find me the ingredients for this week’s meals,” that becomes a very smooth kind of transaction. But most research to date indicates people are still very comfortable transacting on the website through conventional checkout, and want all the functionality they’re used to as part of that transaction.
00:51:10 –> 00:51:15 Cara Corbett: Stell asks for Filip: for a store with thousands of SKUs, where do you start with feed optimisation?
00:51:17 –> 00:52:11 Filip Ruprich: Definitely don’t tackle all SKUs at once. Start with the roughly twenty percent that generates most of the revenue or clicks right now. Then work through the attributes one by one — starting with the basics: product titles, descriptions, and the intent keywords Mack mentioned. Focus on image quality too. Set a proper benchmark after a while, and if it works, roll out the same pattern across the rest of the inventory. With the help of AI, that process gets easier with each iteration.
00:52:13 –> 00:53:09 Mack Johnson: I can’t help myself — I’d add: product taxonomy. In larger, marketplace-style catalogues, it cannot be understated. Taxonomy is effectively the filing system for how your products relate to each other, and it helps AI get much more specific in locating products that fit a particular conversation. If your taxonomy isn’t consistent with your vertical — or with Shopify’s clearly defined product taxonomy — you’re putting yourself at a disadvantage. It’s the equivalent of a store where products are just piled everywhere versus one with clearly labelled aisles directing you exactly where to go.
00:53:11 –> 00:53:20 Cara Corbett: Anastasia asks: should eCommerce stores invest in blog content?
00:53:22 –> 00:53:40 Mack Johnson: God, yes. It’s more important than ever. It cannot be overstated — this is not going away. And especially because of the semantic nature of search: product discovery is very much brand discovery a lot of the time.
00:53:42 –> 00:54:34 Mack Johnson: A simple example: I work with a number of jewellery brands, and they have customers buying engagement rings for the very first time. For somebody who doesn’t know where to start and is asking “what is a shank on a ring?” — you want to have the most detailed answer to that question, and you want people engaging with it because that builds trust with your brand as an information provider. The more detailed you can be, and the more closely it’s tied back to your products, the more confident AI is going to be in referring people who are actually looking to make a purchase.
00:54:34 –> 00:55:15 Filip Ruprich: Of course I agree — blog content is an absolute foundation and we should continue expanding it. I would avoid very basic informational topics and focus instead on genuine experience content: product comparisons, expert reviews, use cases, industry expertise, original research. That’s where the real added value is for eCommerce sites in an AI search world.
00:55:17 –> 00:55:31 Cara Corbett: From Claudia: Search Console and Merchant Center don’t tell you if you’re showing up in ChatGPT or Perplexity. Merchant Center will cover the Google side soon, but that’s Google only. How can we measure that right now?
00:55:33 –> 00:55:58 Filip Ruprich: Two main options. First, you can track specific prompts on specific LLMs — your visibility across those prompts, number of mentions, number of citations — and compare against your competitors.
00:55:59 –> 00:56:28 Filip Ruprich: Second, you can track AI referral traffic in Google Analytics from different LLMs. On the SUSO Digital blog there’s a specific article on how to configure this report. It’s easy — five minutes, you do it once, and then you have it to monitor week on week and month on month. Hopefully you’ll see that AI-referred traffic growing over time.
00:56:30 –> 00:56:43 Cara Corbett: Mack, anything to add?
Mack Johnson: No, Filip nailed it — anything I add would only detract from his clear messaging.
00:56:43 –> 00:56:58 Cara Corbett: Last question, from Richard: have you got a go-to example of a brand doing all of this really well? Mack, maybe one you work with, or just a brand you admire?
00:56:58 –> 00:57:53 Mack Johnson: I’ll be frank — I don’t have a single North Star brand right now. But that’s actually where Shopify’s catalogue query tool becomes very useful. If you have a specific store ID, you can check specific shops and see what competitors are doing in the space, and the quality of their data — though that’s very hard to see below the water.
00:57:54 –> 00:58:42 Mack Johnson: Shopify has an AI readiness check you can run any site through, and it will score any Shopify store on the quality of their data and taxonomy. I need to find the link, but it’s a dedicated URL — you can put in Gymshark.com and see how everything looks. That’s probably more useful than a general North Star brand, because you can check how well your actual competitors are performing relative to your own store.
00:58:43 –> 00:59:03 Cara Corbett: Love that. You know, a lot of LLMs are actually going to Google’s index — the index still matters, but maybe less so in this agentic commerce era.
00:59:03 –> 00:59:35 Cara Corbett: OK, we’ll wrap it up there — we’re right on the dot, and we did great with time. Thank you so much Mack and Filip for joining, and thank you to everyone who tuned in. We’ll be in touch soon. Our next webinar is going to be a great one — we’ve got Mike King from iPullRank joining, and it’s going to be very interesting and very polarising. Stay tuned. Have a great rest of your day, everyone.
