The Agentic Experience Shift — Fireside Chat with Guy Yalif & Hyoun Park

Julia Nimchinski:
Which is gonna be the agentic experience shift. A fireside chat between Hyun Park, CEO and Principal Analyst at Amalgam Insights, with Gaia Leaf, Chief Evangelist at Webflow. Welcome to the show! How are you doing?
Guy Yalif:
Doing great, thanks for having us.
Julia Nimchinski:
Amazing. We start every, every single session this, time around with a prediction, so I’d love to hear your prediction for AIMGTM for 2026. Guy and young.
Hyoun Park:
I think…
Guy Yalif:
2026 is a year where we have mainstream adoption tied to business goals, where we move from semi-random acts of AI to things that are, directly tied to ROI that we’re looking for day in, day out. I think agents will be a big part of the discussion, and I think it’s more towards the end of the year that we’ll see them in widespread use. Hyun. Over to you.
Hyoun Park:
Yeah, I feel like with agents coming into the mainstream and taking center stage, a lot of businesses are going to find out that this is where data quality, where the rubber hits the road, where it really matters whether you’ve figured out how to clean up your data or not, because the deadline has arrived. This is the time where either your data is clean enough for you to be able to take advantage of AI, or you’re going to have to continue cleaning up. All of the… Warnings over the past several years about turning your big data into real data have now come into fruition.
Guy Yalif:
It’s like, what’s old is new again. I mean, you know, the dirty little secret of machine learning is that 80% of the work is clean data, and how you arrange the data, and what’s old is new.
Hyoun Park:
Yes, absolutely. So, to kick off this chat, obviously, thank you so much to everybody who’s shown up. We’re really excited to talk to you about how agents are changing the role of AI and the work that the enterprise does. We’re gonna start with kind of the idea that the Traditional website-centric model discovery is eroding quickly, perhaps more quickly than you were ready for. We’re seeing agents, personalization, all these things that are shifting control away from the traditional owned channels that you’ve been used to having in the marketing world, and towards… we’re moving towards model-based interpretation. Where signals, data, and trust of content are mattering more than, perhaps, aesthetics or some of the gamified aspects of SEO. And think about it this way. The first thing that is going to read your website in 2026 is going to be an agent, not a person, right? That agent or that AI is going to be reading, evaluating, and providing recommendations on your content as soon as the year starts. So, how are we going to adjust to this world where AI has taken center stage and is controlling the way that we think about the world. So, Guy, of course, you are the chief evangelist at Webflow, one of the absolute premier companies in the world working on the customer experience, the web experience, and you’ve got to be seeing this transformation happen as much as anybody in the world. So, you know, what do you say to that, just as a starting point?
Guy Yalif:
I’ll, a bunch, but I’ll just say, hey, thanks to Julia for creating this whole venue, and Hyun, thanks for sharing the expertise you have as, you know, years as a founder and an analyst. And I’m coming as former CMO and AI website personalization company CEO. And to directly answer your question. I think that websites now have multiple audiences. They have before, but I think we’re talking about it a lot more. And by that, I mean we will need to tell our story our way in emotionally engaging, evocative, visually stunning ways, forever. Folks are going to want to see that. Especially for considered purchases. That having been said, exactly to the whole premise of this conference. machines are going to come and crawl our sites and deliver more info than they have before. We have been serving them for a while, right? We all created SEO metadata. We, to your point about the games, we would play with SEO, we would sometimes create content specific for that. we tagged our sites for Facebook’s open graph, if we were in B2C. And so I think this is… More of that. We’re going further down this direction. This is, in some sense, as you and I were talking in prep, a new channel, right? We had mobile, we had social. We have, other channels that have come up over time, and this is another one. So I… Humbly submitted. Don’t think. Website’s dead. It’s all agents now. I do think. Your website continues to serve a purpose of engaging and driving emotional connection with humans. It also needs to make it really easy for the machines to understand the structure and meaning of your content in a way that’s inexpensive for them to crawl, and through both of those. You differentiate through brand and emotional connection. How’s that land for you?
Hyoun Park:
Yeah, I think that’s a good starting point, that we have to maintain creativity while understanding the new rules. But let’s be honest, in marketing, I feel like we sometimes, in some aspects of marketing, we can get carried away with trying to figure out the latest tweak, the latest.
Guy Yalif:
Never. What do you mean?
Hyoun Park:
You know, SEO has always been about that to some extent, trying to figure out how’s the Google algorithm changing, and then how are we shifting in response? And that… challenge gets even harder in the model world, where it seems like the likes of OpenAI and Google and Anthropic are coming up with new models every week, or every other week, and changing the way that they look at that, so… How much can we really depend on, kind of, that you know, tweaking aspect of SEO, or what we’re starting to call AEO for AI, compared to how can we be creative with our digital presence?
Guy Yalif:
I… to your point about how rapidly things are changing, you and I, and many listening in, lived through… the emergence of search. I was gonna say Search 1.0, but, you know, where there were many players, things changed all the time, right? Like, there’s a reason it makes news when Google updates their algorithm now, because it’s 15, 20 years mature, and it doesn’t change as often, and so I think many of the patterns from back then are instructive, where, to your point, there were games to play then. I observe, empirically, there are fewer games to play this time around. I think part of that is because the LLMs have… more signal than search did. You know, John Patel in 03 famously said Google is the database of intentions. We all saw it, right? Insanely great commercial intent. But as we all learn to speak. Google, our average search query was forwards long. Now, our average query to LLMs is 23 words long, and as we all know, you get a full discussion, and for perplexity and OpenAI, they get to see, what are you doing on the website, what content is useful. So I think they get a lot of signal, and so as we think about our digital brand presence, The core job hasn’t changed. what are you doing that’s going to delight someone so much that they’re going to pay you money? And then how do you get that message pulled through onto other sites the LLM is crawling, into the LLMs themselves? How do you tell that story? And then, when you’re fortunate enough to earn that traffic to your site. How do you convert it? And so, in some sense, it’s, yes, there are some new things to adjust, we could talk about those, but the core job of great storytelling, engagement, and ensuring that’s present everywhere that matters to your prospects, like, that pattern remains true. Does that land for you?
Hyoun Park:
Yeah, and I think it’s just a statistic of moving from 4 words with Google to 23 words with AI. On average, of course, prompts can be much larger than that. We can ask real questions and expect to get coherent answers now, and that really does… change the game, having a general engine that can actually answer questions for us, but it does mean that we have to prepare for these longer questions, and kind of how to answer that and how to make sure that we’re not being left out of the conversation. What really… what is changing from a preparation perspective as we’re trying to answer these? longer questions that can be asked in a general basis, and asked to multiple different models while getting different answers, because these models aren’t perfectly quantitative. They have some qualitative and, you know, differentiation among them. Totally, and they, to your point, are probabilistic, so we don’t always get… we don’t get the same answer 5 minutes later. Yes.

Guy Yalif:
you know, I think the whole thing presents both threat and opportunity to both parts of our jobs as marketers, you know, building brands and driving revenue, and the things that change, I think about them in four categories. One is your content, where, you know, the more you can help LLMs have, to your point. More content that they can match to, the better, and so I think the content machine is as hungry and voracious as it’s ever been. And… for anyone tempted to say, great, I’m gonna use an LLM and spam out a billion pages. joyfully, it seems like we might skip that phase of this new channel, where 85% of the AI content in Google is either entirely human, that’s 69%, or a little bit of AI, 16%, and so I hope we skip the, you know, just, it’s spammed out, so genuinely valuable content. With human involvement, gets rewarded. kind of like good SEO always should have been. That’s content for technical… yeah, yeah. For technicals, the second category, where I think, you know, the single biggest thing we can do is help just like the early days of search, where Google had trouble crawling sites, now obviously it’s the best in the world at that, helping LLMs understand the structure and meaning of your site. Through the content itself, bulleted lists. Summaries, meaningful semantic headlines, and technically. schema as another form of markup that, you know, Google, Bing, Yahoo, Yandex, and others came up with a few years ago. Accessibility metadata that LLMs are using a lot. SEO metadata. That is… it’s hard to overstate how valuable that is. At Webflow, Vivian Huang, who runs our SEO AEO team. She took our top 6 product pages. just for those 6, added FAQs at the bottom that restated some of the content above, but then would make it a little easier for the LLM to match a question to that. Added schema, the metadata saying, hey, here’s what this, you know, means. In 2 weeks, more than half of the incremental citations we got for Webflow.com were to those 6 pages, and we have thousands of pages, and those pages got 20% more traffic. In two weeks. So, we are seeing the opportunity for those that move earlier to get big benefits. Third category is authority, where… backlinks still matter. They don’t not matter, because the LLMs… are searching all the time, so, like, SEO is really valuable for AEO, but, they’re looking for consensus, plain text mentions on a bunch of other sites that… where you have the chance to pull through your message, to your earlier question. You have the chance to influence what the LLM is going to reformulate and cite, and the last is measurement, where… The concept of keyword ranking doesn’t exist anymore, and so now you start looking at, in the reformulated copy. did I show up for the questions I care about? And then to what you said earlier. Was I mentioned positively? Was my message pulled through? What’s the sentiment on it? And so the measurement changes a bunch. So I think, you know, content, tech, authority, and measurement.
Hyoun Park:
Yes, Content Technology Authority Measurement. And I find especially interesting from kind of this measurement perspective, that recency is so much more important from an AI perspective than it was for SEO. To some extent, with SEO, you could create this garden of backlinks, and then it would just kind of stick with your site in perpetuity. And with AEO, that’s not really the case. I think I saw something like three-fourths of all AI references have been brought up in the last, like, 9 or 10 months. Like, it’s a game of what is correct right now, what is interesting right now, and what makes sense… what is most contextually relevant to the topic that you’re asking about. Right now, rather than 5 years ago, 10 years ago, 20 years ago. It’s very different from that perspective.
Guy Yalif:
I couldn’t agree more, and the data, I think, widely supports it. Content with some indicator of freshness, like last updated, gets cited 1.8x more often. 95% of ChatGPT citations were updated in the last 10 months. 95%. at… Webflow, Vivian and team. increase the speed of updating content 5x, over the course of a couple of months using some, some tools. they saw 42% more traffic to those pages, and the traffic converted 14% higher. So it’s like, more traffic, higher. And so, like, as we were talking about, the need to create and update content is as high as ever. I think it creates two interesting dynamics. One is, you know, in SEO, would you ever take down content? It’s kind of like, no, over my dead body. But in AEO, because they’re doing crawls all the time and the consensus matters, if you’re… if you have… I know this doesn’t apply to anybody listening here, but if someone else has, like, out-of-date content on their site, or there’s out-of-date content somewhere else. the concept, if you’re not going to be able to update that anytime soon, of potentially taking it down, it’s a conversation you would at least entertain an AEO or as an SEO. You’d be like, yeah, never, not happening.
Hyoun Park:
Yeah, it really puts much more of a demand on being a factual curator of information, rather than simply putting it out there, you know, shoving more data into the website or into the digital asset and hoping that you are able to pick things up. There seems to be a lot more focus on quality and recency, so… But, you know, so quality and recency, of course, but then you brought up this other aspect of creativity, so I’m interested in how you think about creativity as being relevant to the AEO world. That might be different from the past.
Guy Yalif:
in my humble opinion, it’s as important, if not more important than ever, which feels potentially a little counterintuitive when, you know, all of us are thinking, well, I’m just gonna get some reformulated text, why does this matter? In a world of effectively infinite content. How do you differentiate? I think it is… a lot of it is about emotional connection to the human you are talking to, whether that’s something said repeatedly, so an LLM pulls that message through. Or it’s some visually stunning, emotionally evocative thing on your site, but to be able to Speak to your prospects in ways that is memorable. Stand out. That matters as much or more as it ever has in a world where the LLM may very well confidently, completely ignore your core value prop, or confidently get something wrong. The more you can repeat that creative message, on your site, off your site, through digital PR, through, you know, collaboration elsewhere. The more likely you are to have that message pull through, the more likely you are then to have people Remember it, because, to the whole premise of this discussion, they are able to gather information so broadly, so quickly.
Hyoun Park:
Yeah, so before, when you brought up, content tech authority measurement, of course, this is part of the AEO maturity model that Webflow brought out earlier this year, and that you led the development of. And of course, along with those categories came some maturity levels as well, associated with where your AEO is going, starting from very basic, kind of keyword-based approach, and then getting more mature after that. I was curious, just thinking about how that maturity is taking place. You know, where do you think we are now, and where are we really trying to get to?

Guy Yalif:
I think it’s early days in a new medium. I think everyone feels behind. But there is so much opportunity. You know, We talked about schema before. 78% of Google’s top search results have schema in them, okay? But 88% of sites don’t have schema on them. Like, there’s so much opportunity in the sense of there being bargains of… of… not of literally of price, but rather of, like, energy put in relative to benefit out, in terms of brand building and driving revenue. And so, where are we? I think we are in the very early innings. I think it’s a lot like the early, like the, you know, 00s in… early 00s in search, where… The rules of the game are changing regularly. But the core fundamentals, unlike the early days of search, we know what they are, right? We know the core value you should be adding to your prospects, and the core creativity you want pulled through. You know, those foundationally aren’t changing, and in fact, we have a better read on them than we did back when Google got started, I would argue. I think most teams are coming to grips with the fact that Your SEO team is your EEO team. your SEO agency is your AEO agency, and that the way to help them… it may be a contrarian viewpoint, but, like, the way to help them succeed the most is to create space and permission for them to experiment and fail because it is so early. So, you know, for most of us as CMOs, you know, this is a board-level issue. For a few quarters, probably a few years, you want regular readouts, you want to share experiments and outcomes regularly. probably want to support them with tooling as needed, where they’re going to experiment, right? They’re going to try something, they’re going to try something else. I’ve heard some folks, and we’ve hired two or three, hire for a role of go-to-market engineer, which they, and I would argue everyone else playing in this space, should Come with a lot of curiosity, and ideally some systems thinking to think. Core job remain the same. Vehicle through which I deliver it has changed some. How do these pieces fit together, and how do I get that message pulled through? How do I go drive that traffic? And where we’re going, I think, is, eventually a mature new channel. And just like in the early days of search. would we have really figured out that you’d be doing commerce directly in Google, or having, you know, knowledge graph entries, or snippets of sites? Right. Maybe, maybe not. I’m sure there are new things that are going to come up in answer engines and LLMs that we haven’t thought about yet, and you want a team that can roll with those punches. Does that land for you, Hyun?
Hyoun Park:
Yeah, I’m actually really excited about seeing this new role show up. Just the idea of being able to experiment across different media formats, just even little things like you brought up, bringing FAQs to the bottom of a page. Is it best as an FAQ? Is it best as a video? Is it best as some sort of explanatory document? Maybe it should actually be a bot or an agent that is actually explaining things. Who knows what the exact right format is going to be that is most educational and most, engaging for the user at the end of the day. We really don’t know. And for the agent, like, what is that hybrid experience going to look like? This is an absolutely interesting opportunity. And, you know, you brought up the 00s, like, this is almost like… you know, late 20th century, web, where we’re trying to figure out even what should it look like. Hopefully we don’t get stuck with a bunch of blink tags, but, you know, we’ve got, like, the entire, you know, world is, like. Just a palette and an opportunity to be able to build again on the front end, and that’s a really interesting opportunity.
Guy Yalif:
I am so, so, so with you. It is a privilege to have the learning curve go like this again, and it’s funny you mention the blinking tags. I met my future co-founder running, at the time, the world’s largest web hosting company that included Yahoo GeoCities, which had its very fair share of blinking tags. And, you know, to your point about the things we need to discover, like, one could argue it would be beneficial to put in the metadata of every single page your core value problem. So that the LLMs see it more and more often. in SEO days, that might have been considered black hat, because you’re putting something that the machine sees that the human does not. I don’t think that’s considered black hat right now. I think redirecting would be, and, like, learning how can you be helpful to the LLMs in giving the right message pull through in a way that they are happy with and say is white hat? I have intense curiosity in so many places. That’s one of many examples.
Hyoun Park:
Yes, and at the same time, adding that metadata and some basic schema and mapping is… actually not that big of a load from a work perspective to be able to enhance your current media, library and your portfolio. So I think that might actually be a win-win, instead of having to build, say, a hundred or a thousand or ten thousand new pages of material.
Guy Yalif:
I am with you, and, having the ability to auto-generate a bunch of that is incredibly helpful. Our product team went and built that, and when they did, I was like. This is fantastic! Like, site-wide auto-generated, schema is so valuable! And there are gonna be many more like that that, our tool and many others will do. .
Hyoun Park:
But to your point, it is an interesting time to go explore, and then empirically see.
Guy Yalif:
what impact did that have? I think the feedback measurement loop is immersively, interestingly, faster. Like, you can make a change and see.
Hyoun Park:
an impact this afternoon, tomorrow, because the LLMs are crawling virtually every time. I…
Guy Yalif:
personally, I have curiosity whether that’ll change as they, you know, rebuild a stratified crawl of the web like Google has done better than anybody in the world. Will they crawl less often? I have just so many variables to go explore.
Hyoun Park:
Yeah, it’s funny, the older I get, the more I realize that… how much Yahoo has actually built our internet. You know, even if Yahoo didn’t necessarily get all of the financial gains of building all of that foundation. So, you brought up the idea of the go-to-market engineer already, but I’m curious for CMOs trying to figure out what to do for next year, what are the… biggest organizational changes they need to make. You know, they’re used to thinking about, perhaps, their… their search, their demand, their events, their content strategy. Is there… do they need to have a separate AI? Manager, or a director, or, you know, how do they need to think about this from an organizational perspective?
Guy Yalif:
A few things come to mind, the biggest of which is not a radical shift, because your SEO team is your AEO team, right? If you were to define the ideal AEO person, they would have an SEO background, one. Two, I would change how you report out to the CEO and management and the board, and then how you evaluate the team traffic-wise, because for most folks, traffic is down 10, 15, 20, 25%. But the LLM traffic is converting much better. At Webflow, it’s 6x better, Ahrefs and SEMrush found it 4x to 23x better, and so having this, like, de-averaged thing, oh, traffic’s going down, but conversions are going up. Like, that’s a meaningful discussion to have. To your point about AI Manager, I think AI fluency is a rare thing where, you know. everything from our kids in kindergarten through to experienced CEOs have a bunch of AI fluency to go build. That’s why I’m grateful Julia is putting on something like this. And I think there’s a ton of work to do there. Is that one person? I think every company’s approached it differently. At Webflow. We have AI Council, we have a bunch of tool enablement, we now are pointing that To the initial prediction, more and more towards concrete business outcomes. It’s now part of our hiring approach, it’s part of our evaluation approach. And the last thing I’ll share as input to strategy assess your site on AEO, and I invite folks to go to webflow.com slash AEO. There’s no pitch there, it is purely a, it will go assess your site on all four of those categories, on your site, off your site, and say. Here’s what we can see from the outside, here’s what we suggest you do next year. Not about Webflow at all, more about helping you assess your site on AEO.
Hyoun Park:
That’s a fantastic set of recommendations and some practical tips to be able to start out the new year. Thank you so much for your time, Guy. Thank you so much for your wisdom and for helping us to understand where are we going next with AI and how to prepare for that.
Guy Yalif:
Hugh and Julia, that was a lot of fun. Thanks for the chance to join you.
Julia Nimchinski:
Very inspiring, thank you so much again.