AI Discovery Revolution: Session Recap: Key Takeaways from Matt Ezyk at eTail Boston 2025
As consumers increasingly turn to AI assistants like ChatGPT and Claude for product recommendations, retail brands face a critical inflection point. At eTail Boston 2025, Matt Ezyk, Senior Director of Engineering and eCommerce at Hanna Andersson, delivered a keynote session titled "The Shift from Search Engines to AI Assistants – How AI is Reshaping Discovery." Ezyk shared real-world insights into how AI is fundamentally changing the way customers discover and evaluate products, and what retailers must do to remain visible in this emerging landscape.
Key Takeaways
1. Search volume is shifting dramatically to AI assistants
According to Gartner research cited in the session, 25% of search volume will migrate from Google to AI assistants by 2026. Ezyk emphasized that this shift is already accelerating faster than initially predicted. Customers are no longer typing keyword queries; instead, they're having natural conversations with AI tools, asking contextual questions like "What's the best sustainable children's clothing for my 4-year-old?" This represents a fundamental change in how product discovery happens, and brands that don't adapt risk becoming invisible to a growing segment of consumers.
2. AI adoption for shopping is already mainstream
The data is striking: 80% of people are using AI for search, and 49% are already using it for shopping guidance. These statistics, drawn from Adobe and Vision Horizon research, underscore how rapidly consumer behavior has shifted. Ezyk noted that this adoption rate surprised even industry observers, demonstrating that AI-assisted shopping is no longer a niche behavior but a mainstream discovery channel. Retailers who treat this as a future concern rather than an immediate priority risk losing market share to competitors who are already optimizing for AI visibility.
80% of people are using AI for search and 49% are using it for shopping guidance. 49% already. This came out this year that that's wild to me that it's changed so rapidly.
— Matt Ezyk, Senior Director, Engineering eCommerce, Hanna Andersson
3. AI agents will reshape commerce, but not replace human experience
Ezyk explored the emerging concept of agentic AI—tools that can autonomously complete shopping tasks. However, he cautioned against the "doom and gloom" narrative that PDPs and websites will disappear. For experience-driven categories like fashion and sunglasses, consumers still want to see, touch, and evaluate products. Agents work best for convenience-driven purchases like groceries, where Ezyk built a proof-of-concept that generates meal plans and automatically sends shopping lists to Instacart. The key insight: AI will augment the shopping experience, not eliminate it entirely.
4. Platform gatekeeping is intensifying
Amazon and Shopify are actively blocking AI crawlers from accessing their sites, requiring official integrations instead. This gatekeeping signals that platforms want to control the AI discovery experience and capture the transaction value themselves. Ezyk highlighted Shopify's "checkouts are for humans" robots.txt directive and Amazon's disallow of Perplexity, Claude, and GPT bots. Brands need to understand these platform dynamics and prepare for a fragmented AI ecosystem where visibility depends on official partnerships and integrations.
5. AI visibility requires SEO-like optimization strategies
Just as brands optimize for Google's algorithm, they must now optimize for AI training and recommendations. Content strategy, Q&A formats, and consistent messaging about brand values are critical. Ezyk shared how Hanna Andersson audits its AI visibility using tools like Scrunch AI and ensures sustainability—a core brand pillar—appears across PDPs, Q&A sections, PR materials, and even an AI-friendly robots.txt file (llms.txt). The approach mirrors SEO best practices but with an emphasis on conversational queries and product information that AI models can easily consume and reference.
6. Direct relationships and first-party data are more important than ever
As AI intermediaries gain power in discovery, building direct customer relationships and owning first-party data becomes a competitive advantage. Ezyk emphasized that brands should not rely solely on search visibility—whether traditional or AI-driven. Investing in owned channels, email, loyalty programs, and direct engagement helps retailers maintain customer relationships independent of platform changes. This diversification reduces vulnerability to algorithm shifts or platform gatekeeping.
7. Continuous experimentation is non-negotiable
Ezyk stressed that no one is an expert in AI discovery yet, and the landscape changes weekly. ChatGPT 5 launched during his keynote preparation, forcing updates to his presentation. Brands must adopt a culture of continuous testing—auditing how they appear in different AI tools, monitoring platform changes, and iterating on content strategies. Ezyk encourages retailers to test prompts, track visibility changes, and learn from peers about what's working in their categories.
Why It Matters
The shift from search engines to AI assistants represents one of the most significant changes in retail discovery since the rise of Google. For brands, the stakes are high: if your products don't appear in AI recommendations, you're invisible to a rapidly growing segment of shoppers. Unlike traditional SEO, where brands have had years to optimize, the AI discovery landscape is nascent and evolving at breakneck speed. Retailers who move quickly to audit visibility, optimize content, and build direct relationships will capture customers that competitors can't even reach. Conversely, brands that delay risk ceding market share to more agile competitors and losing relevance in a post-search world.
Actionable Insights
- Audit your AI visibility today: Test how your brand appears in ChatGPT, Claude, and Perplexity using your core brand keywords and customer questions. Use tools like Scrunch AI to monitor visibility across AI platforms.
- Optimize content for conversational queries: Rewrite product descriptions, FAQs, and brand content in Q&A format. Emphasize your brand's core values and differentiators consistently across all touchpoints—PDPs, PR, About Us pages, and boilerplate content.
- Create an AI-friendly robots.txt file: Implement an llms.txt file that clearly communicates your brand's key attributes, values, and product information to AI crawlers, similar to how you optimize for search engines.
- Build direct customer channels and track attribution: Invest in owned media, email, and loyalty programs to reduce dependence on AI intermediaries. Implement server-side tracking to measure traffic and conversions from AI sources like ChatGPT and Claude.
Want more insights from eTail Boston? Explore the full agenda to discover additional sessions on AI, personalization, and the future of retail.
Click to View Full Session Transcript ▼
2025, eTail Boston. Keynote: The Shift from Search Engines to AI Assistants – How AI is Reshaping Discovery
Matt Ezyk, Senior Director, Engineering eCommerce, Hanna Andersson: Everyone. Hope everyone had a great travel in. I came in yesterday and I got to play golf at Fenway Park. Like how cool is that? Like I, I grew up in Boston. I don't live here anymore, but that was incredible. That was the highlight of my trip. I can go home now. So my name is Matt Ick. I'm from a brand called Hannah Anderson.
If you're not familiar with the brand, we sell premium, sustainable children's clothes. For infant to age 12, and we sell family match pajamas and stuff for cats and dogs, and it's a super cool brand. If you're never familiar with it, check it out. I lead our technology group there and when I pitched this idea to eTail in September of last year, I didn't really feel like a lot of people were talking about this topic.
And it's the shift of consumers going from traditional search to large language models. I'm gonna make this pretty conversational, not really have a ton of slides, just talk through it how I'm seeing it, a couple stats and then a couple takeaways that you can take to your brands and implement.
Is there anyone in the audience that is from a brand that sells sunglasses? Okay. What's your brand? Front. Okay, cool, because I have I do all of my testing with large language models around sunglasses. I don't know why. I just, that was the thing that I was looking at that day and I ran with it.
You probably find this more interesting than anyone else, but anyone else who is interested in doing some of this evaluation I'd be happy to do it for you. It's just something I do on the side. But I do work for brands a couple things before we get started with this too, is that I met a professor at a conference over the summer.
He's a, at a prestigious school. He teaches artificial intelligence and he said something really interesting to me. He said, no one is an expert in this space right now. And I thought that was really profound because it's changing so much. As you can imagine, it's very difficult to put slides together for something like this because it changes, right?
Chat. GBT five just came out on Thursday. I already built this. I, it's it's almost impossible to keep up, but you have to continue to learn, have to do something with it. So with that being said let's see if my clicker works here. All right, if you just wanna hit play in this video again, just wanted to set the stage in the context for what we're talking about here.
Again, I'm using sunglasses here and when I first did this testing, when large language models came out, I was like, I wanna go on a beach vacation. What's the best pair of sunglasses I can buy for that? And it showed me RayBan, it showed me Oakley, it showed me Maui Jim. And my interest in this topic was like, if I'm that brand that's a third, what do I do about this?
This thing that people are moving to are recommending my competitors, this is crazy, right? And it's gotten so much better. And I'm sure everyone in this room has experimented with this, but I opened a fresh session chat, GPT, and I type that in. With a little bit more context to it, right?
I said, Hey, I have a budget of $300 and oh, by the way, my flight's on Sunday. I need this by Saturday. So it started to gimme some results, but then what it started doing is asking me more questions, contextual questions okay, what's your zip code? I put in a zip code and then it's now gonna give me results based on that.
And the results are a little bit different. And again, customers are using this as and they're not even going on your site at all. So great. I put in my zip code. It gave me different results and if you notice, it's from Dick's Sporting Goods and Shady Rays because they are showing up contextually better for chat GPT than some of these other brands.
But it's still showing me Maui Gym as a premium brand. It's given me a recommendation and then it knows that I need it by Saturday 'cause my flight's on Sunday. So would you like me to check availability? Yes. I'm not really gonna go too crazy with this. I'm just gonna say yes and see what it does.
So what it's gonna start doing now, and this is in its infancy, I think a couple brands pilot this right now. I have not been able to do this yet, but eventually you'll be able to send a product feed to chat GPT. I'm gonna use chat GPT in this just for context instead of all the other ones, just for simplicity.
But now it's giving me slightly different results, different models, because it knows it's available at Dick's Sporting Goods. It knows I can go to Sunglass Hut. And then it's just gonna keep narrowing down and keep asking questions. It's looking at polarized lenses, it's looking at all these different things with this, and there's so many different other areas that you can get to with this, right?
Like for me, I have a large head. I need something with large frames and a large nose bridge. I could ask questions like that, right? You can't get that from A-A-P-O-P and A PDP traditionally. So I'm making a decision as a consumer without even going to your brand, and this is influencing my decision.
So it's becoming more and more important to understand where this is coming from. Now, it's gonna take it a step further with agent. So agent is. Basically being able to build something on your own or use a brand's agent to do the shopping for you. So it's gonna automate tasks, it's going to go into the browser and look up information real time.
But I think there's a lot of doom and gloom around this. Like I hear a lot of people saying, the PDP is dead. Or you're, you won't have a website anymore because these things are just gonna buy everything for you. And I don't think that's the case because everyone in this room loves to shop for things, right?
If you look at sunglasses I don't want an agent buying a pair of sunglasses. I wanna look and see what it looks like, right? What does it look like on my face? Is, am I gonna look cool or am I gonna not look cool? I don't want something that I don't wanna look cool. So I'm not gonna trust, at least at this point, something like that.
But this is something to definitely be aware of and it's all contextual based on how you appear in these searches. So what is the industry saying? This first stat was really interesting. This came out from Gartner in January of 24 that said that in 20 26, 20 5% of search volume is gonna drop and shift from Google over to this.
I thought that was accurate at the time. And I just saw Brendan from Forrester who wrote that around here somewhere. I think he's in the AI session actually. I thought it was accurate at the time. I think it's gonna be more now because now customers are getting used to having this conversation with chat GPT and discovering products.
I even caught myself doing it right. I was in, an auto zone or something like that. And there was two products on the shelf and I didn't know the difference between them. So I opened chat, GPT, and I turned on the camera and just had a conversation, said, Hey, what's the difference between these?
This is the type of car I have, this is what I'm looking to do, and it recommended I buy one of them over the other, and I did. So this is really interesting and scary being a brand, seeing this because there's just so many brands out there that. Are not doing anything with it. And you always have to experiment no matter what it is.
Lemme go back one actually. So this Adobe report, 80% of people are using AI for search and 49% are using it for shopping guidance. 49% already. This came out this year that that's wild to me that it's changed so rapidly. This is from my friend at Vision Horizon. She did some some consumer research found the same stat that 49% of people are using AI guided shopping.
33% are using it for brand discovery and 80% using chat. GPT is their primary search interface. So the shift is from traditional search like keywords. Everyone is familiar with that flow. You follow the keywords, the results. You look and you make your decision, and now we can have natural conversations with these tools and it's getting better and better.
How many people have tested GPT five yet? It's incredible, right? It's for someone who uses this tool every day of my life, I could see the difference in how intuitive it is and how much faster it is in doing product recommendations. I wish it came out two weeks ago, so I could have put it in here, but but that's the point, right?
You have to just be continuously watching what's going on. And that's one of the takeaways in here. With agents, if an agent controls the buy button, it decides where the order lands and who collects the fee, right? But again at least now in our lifetimes, I don't think we're gonna ever replace shopping for.
Certain products with an agent, right? We talk about sunglasses shoes is another one. But like for me personally, in my research and in my discovery I don't personally like to shop for groceries. I know some people love that. I personally don't, so I experimented building an agent where I could type in a meal plan.
I can say, Hey I'm getting back from eTail on Thursday, and I want. To create a meal for Thursday and Friday night. I want to cook, I want high protein, low carb, whatever it is, and it builds me a meal plan. And then I built it so it makes an API call to Instacart and then sends the grocery list and then I can just click it and buy it.
Now when I eventually have time, I want to build it so it will do price shopping too. That's a use case, at least for me. Where ag agentic commerce is great, but buying, us at Hannah Anderson. We're a fashion retailer. If you have kids, you know you wanna look at the clothes. Especially with stores, you wanna go in and touch and feel the clothes.
This is not gonna be the thing right now for that, right? But it's all the same challenge. It's all the same problem. Where does it get this information? Now this is really interesting. This happened, or at least I caught wind of this a few weeks ago. Again, right in time before I built the deck suite.
Amazon and Shopify are blocking a chat, GPT or a perplexity or Claude from accessing their sites in, in, in a couple different ways. Why? Because they want you as the customers to use their tools and purchase their tools. They don't want. Other tools coming in. So it's really interesting. And this is, these are some screenshots I took from this.
So the one on the left, if you're a Shopify store here, they put in a checkouts are for humans text in the robots file. So basically what they're doing is saying, Hey, we want you to. If you're a legitimate integrator, you have to use the official checkout kit, right? That's a news that just came out pretty recently.
The one on the right is Amazon, right? They have a disallow of, pretty much everything perplexity, Google Claude, GPT bot, because they want you to use their own tool. So preparing for the post search error, right? It's really a couple different things. It's optimizing for AI training. Making sure that your content is relevant to these tools so they can consume it.
Again, very similar to SEO it's scary how similar this is. Building direct relationships, so not re relying just on search visibility, but developing own channels and first party data, relationships with customers and experimenting. You have to experiment. I experiment all the time. I encourage everyone to do it.
I'm hoping to learn. Hope just as much from the people in this room and the people I see at this show of what they're doing, as much as I want to give that information as well. So what can you do? All right and I'll leave this, I don't have a closing side, I'll leave this up here. But this is the way I look at this.
The first thing you do from experimentation is to audit your AI visibility. How does your brand appear, right? If you're working with SEO or you're just looking at onsite search. What are customers searching with and how are you showing up in a chat, JPT or a quad? For us at Hannah Anderson, we have a product, if you're familiar with it.
That's actually all of our products are known for being sustainable. We're a 43-year-old brand that have always built high quality clothes. That'll never change. That's the hallmark of our brand. So if you open chat GPT and say, what's the best place to buy sustainable children's clothes for my child?
We will be number one, or we'll be number two or three. We're in the conversation there because all of our content says that it shows. Interesting enough, when I tested this the other day, it was showing a Wikipedia, right? I never would've thought that we'd see that there. We. Wikipedia all the different publications we have on like New York Times, Wirecutter all of our PDPs, everything that we do from q and a content from SEO perspective is picked up here.
The scariest thing is Reddit, right? That is the Wild West. People can say whatever they want on there, but I really think if you stick to the laurels of your brand. And for us, it's making good high quality products that kids can can be fashionable and also can play in, and they're durable.
That just permeates into everything that we do, including this. So we look for those different prompts. We use a tool called it's called Scrunch ai. And that audits our visibility in the ether with all these different tools. And that's, we found some pretty good success with that, adapting the content strategy.
Just like SEO right? You wanna optimize for this. So these tools love conversational queries and having product information I, at least my brand, I can't send product data to it yet. I've signed up for the. Release of that with OpenAI when I can. But we want to, from an SEO perspective, build our content and q and a format continuing referencing terms like sustainability.
So that's a term we hit on everywhere. Strengthening direct channels, investing in own media, first party data, like I mentioned. One of the things that really helped us is I built a AI friendly robots file. So if you type in our domain and then you forward slash LLMs txt and you can Google that and see what that is or go on chat GPT and learn about it.
That's basically a robots file for AI bots when they crawl our. And the schema markup I wrote for that says, we're sustainable. And we have high quality and we have fun colored prints, and we do collaborations with Star Wars and Bluey all the things that are really the hallmarks of our brand.
And finally monitoring the platform changes. This changes so much. GPT five came out Thursday. I spent a lot of time with it because I got some early access to it. But customers are using that now and I could have this talk a month from now and it'll still change. Being curious and experiment, experimenting is really the key of how to win with this.
And the question isn't whether AI will shape rediscovery. It's whether your brand will be discoverable in this new landscape. Thank you.
I do have how is that for you, for a brand that sells sunglasses? What do you think about that? That's
pretty helpful.
Are you seeing like some the same things or?
Moderator: Yeah we've had some discussions about our optimizing our text files. Yep. Oh, sorry. And, sunglasses are a very small part of our SKU mix.
We're mostly focused on work boots. And so I think footwear is, that's probably
Matt Ezyk, Senior Director, Engineering eCommerce, Hanna Andersson: why you, that brand never showed up in any of my shirts. Yeah. Yeah. I'm not
Audience: surprised there. Yeah. We're very specialized in that. But, footwear is a really similar vertical with Yeah. Millions of skews, so many different people for sure.
And so we're constantly digging in. A lot of the ring's true. It really aligns with our SEO strategy and I think the web optimization we're looking at both from a technical side. Yeah. Like you have a number three here and a content side, like number two.
Matt Ezyk, Senior Director, Engineering eCommerce, Hanna Andersson: What's a just outta curiosity, what's like a.
For us it's like sustainable clothing. What's a search term that you have for your brand and did you test it in any of these and what were the results?
Audience: Yeah. We turn up pretty good in a lot. We focus on, balance of comfort and durability. Our exactly.
Are made for real workers.
Matt Ezyk, Senior Director, Engineering eCommerce, Hanna Andersson: Yeah. And that's probably because you have them in all your PDPs and all the things that are written about you. Or like Joe Blow who was on Reddit writing about you. So it's good. But, not so good for other people who are like, not monitoring this. So it's good that you're doing that.
I got a couple minutes if anyone has any questions and happy to learn from everyone else here, what you're doing. 'cause I want to pick that up as well. Yeah. What advice you have for product descriptions on your pages? Yeah, it's I think I mentioned it here too, somewhere, but it's really taking the same approach as SEO, these tools, love q and a formats, uhhuh, because it's, it, these customers are typing in questions mostly. So for example, I, one of the terms we look at is what, what's the most what's the best place to buy sustainable clothing for children's? And in our PDPs, we have a q and a section. Where we talk about sustainability we should have a whole page about sustainability.
If we know our customers are looking for that, and that's a hallmark of our brand, we're putting it in q and a format on the PDP we're having in About Us we even have it in our like boilerplate information when we do pr. So you're putting that at like the bottom of the page so it picks it up?
Yeah. It's not necessarily in your description. It depends like, it, like if it's something like we have, we sell a one of our most popular products is a dress for girls that, it's, it looks good, it's fashionable, but it's also really durable. 'cause when they go out and play, they're not gonna rip it.
It's got pockets, like super cool product. And that's part of the why someone would buy that. But yeah, like everything that we do, we put sustainability in it. It's in our pr, it's in everything. So that is out in the ether. And so when you do an audit of our brand and I act, I actually built my own tool to do brand audits and score them.
So I'd be happy to run that for anybody here. It mentions it everywhere 'cause that's our intent to have it there. Okay, cool. And I buy Hannah Anderson love it. We'd love to hear it.
Audience: Hey Matt. Hey. Was curious with the techniques you guys have put in place, is there anything that you're doing today to track that down to the commercial sales side of things? So what techniques are you utilizing to show that this was paying off?
Matt Ezyk, Senior Director, Engineering eCommerce, Hanna Andersson: Yeah. When I first, I've only been at the brand nine months.
When we first started, I realized that we were. Our security systems were blocking this traffic, which, you know, that, that's why, you have to audit your visibility because if you're blocking this, then you know it doesn't matter anymore, right? So we unblocked it, and then we're, we use we're seeing the traffic come in from chat.
GPT. And it's really difficult to tie the attribution to it, but trying to build out something server side where we can do that. So it's not, we don't have it ready yet, but the goal is to get attribution from chat, GPT and Claude. That's the two sources we saw the most traffic from. We see it from perplexity as well.
And tie it to that just like any other channel, right? This is an emerging channel that we just, we wanna see attribution, so it's not great yet. And I think if you have GA four and you don't have something server side, it's a little tricky, but. Hopefully we'll get there before holiday. Cool.
So
you
said you, you ordered the visibility. And how do you do that? Do you open up the tool and open up a fresh session that is not tied to your own personal searches? And start with what I did was I I use Salesforce Commerce Cloud for e-commerce, and we have a console that shows all of the top 100 terms that people are searching for.
And then it shows the top 100 terms that people search for that have no results. And we take those and formulate questions into the tools and we just did it manually. Now there's vendors out there that have ways you can do that, and then you can have alerts, right? If something changes. So we've been pretty new with that tool I mentioned, but that's typically how you do it.
So for example, again, for us, like I know the hallmark of a brand is sustainability, all the questions that we see people asking in Reddit or anywhere else, just from our own knowledge, we're testing it ourselves. And then the results were pretty interesting at first. Like they were they recommend competitors.
Sometimes they recommended us over a competitor. Sometimes it changed week to week, month to month. But the more consistent we've been with content, the better it's been. Yep. Alright. Thank you.