2026, eTail Palm Springs -SEARCH_ Keynote_ The Shift from Search Engines to AI Assistants – How AI is Reshaping Discovery

Announcer: Welcome back everyone. Uh, I'd like to introduce Ali Harris, who's gonna be talking about how AI is reshaping discovery. So welcome, Ali, and look forward to- Thank you ... look forward to the stage, yeah. I might actually just

Ali Haris, Director, Organic Search & AI Innovation, Men’s Wearhouse at Tailored Brands: I might actually just walk here. Much more comfortable. Good morning everyone. How's everybody doing? Fine. Great. Awesome. Awesome. I hope you all got some snacks. So before we get started, let's, uh, I had nods in your work too. Let's get, uh, show of hands. How many people think that there's too much information, too many reports, too many trend charts out there, right? There are days I'm scared of looking at actually my LinkedIn feeds.

So at Men's Warehouse, what we are trying to do is we are trying to turn this whole search chaos into something that we can execute in execution models. Because I really feel that there's so much specific tactical things that's going on, that looking at an overall picture is very, very critical. And that's what I want to share.

I want to share that model. I don't want to share another data slide with you guys or have another five minutes going through, "Hey, Reddit is where you wanna go. You wanna create videos. You want to go into structured data." But how do you actually operationalize? Again, a quick pulse check. How many of you guys feel, compared to last year, that you guys have moved or started from searching keywords to se- or to asking questions in the search experiences?

How many compared to last year? We're talking about what? Ten, 15, 20%? And that's exactly the shift we are talking about, right? So this shift is happening not only within the behavior, but we are talking about the shift in experience itself, right? You have looked at Google AI Overview, ChatGPT. If you haven't shifted already, Google guys will shove it down your throat with AI Overview.

So that's coming So as an organization, the question becomes, how do we adapt to this shift, right? And I think so a lot of you guys at least signing here also to hear from everybody else within the conference, how are they adapting to that shift? How can we make sure as an organization that we are there where the customers are, right?

Our customers are on Google, we need to be there. If they are on ChatGPT, we need to be there. So that's the big question that actually we need to address. And at Men's Warehouse, we start with a very three basic questions, making it very, very simple as much as possible. What, how, and where What as in What is it that the customers are looking for?

You've heard of it, prompts, right? Hey, what are the questions people are asking for, right? To some extent, what we have done is actually we are trying to reverse engineer. By the way, the model that we are following, is it the perfect model? Absolutely not. Is it going to change? Hundred percent. Is it gonna change by tomorrow?

Probably So going back to the what question, right? So what is it exactly the customers are looking for? The second is what we start looking at is what is the AI answer that is showing up. The-- What is a good AI answer? So for example, the question can be: What should I wear to a semi-formal winter wedding in Chicago, right?

So I'm not saying that the LLM model will always give the same answer, but it's kind of in the same vicinity. Comparison, simple rules, timing, confidence, and trust. The whole idea is, uh, can we templatize it to some form or shape so we can scale it? It's rolling into the scalability piece of it. I'll talk about it a little bit later.

And that also determines the content that we want to create. So rather than just randomly creating a blog content or some other content, we start looking at, hey, what is that actually-- LLM is actually serving, and then build it out and build the content around it So we talked about what, very simple. The next one we go into how.

How do you get picked by assistants? So by a long shot, I'm not clai-claiming that I figured it out, but we have tested this around this, right? So once you create the content, how do you make sure that it-- we are gonna get picked up? The first one is readability. Very, very critical. More further into the engineering realm of the world.

You have to make it readable. Now, some of you over here, some of you guys will probably meet up with, uh, partners and teams who just focus on making your content readable for the LLM models. And also keep in mind, when I go through this thing, you can probably have a dedicated team and have a dedicated partner who can just work on this one specific piece of it.

So going back to it, readability is one. So we have to make sure that... And, and it's kind of the same practice that we do SEO, right? It makes sure it's crawlable, it's readable. But the other piece of it is readable feeds. The feed is very, very important. For example, if you have a product feed, do you have those feed enhanced?

Again, we can have another dedicated session that just talks about the readability part of it. The other thing is consistency, freshness, and actually recency. Consistency is you need to have consistent data across the board because LLM is looking for that content all over. So for example, um, you go on Gemini, I'm looking for a tuxedo suit for a dinner in Palm Springs.

I need it in two days. LLM model in one place will look at it. Uh, on the site it says five days. On your blog it might show seven days. Somebody's feedback has said it, it took seven, uh, ten days. Your trust goes down the drain right there. So that's why consistency is important. And then we have got, uh, freshness and recency.

So, uh, so the-these are two pieces of it. So you can have a very stale content, but very relevant content, so you need to go back and update it. That's what we have started doing. Rather than just focusing on creating new content, recency is important because you can have, uh, for example, a blog content that talks about what to wear for prom is still relevant for twenty twenty-six, but it was published in twenty twenty-four.

There's no way LLM model will pick it up. And then structured format. So there are multiple ways you can structure it. You can have a dashboard. You can have, for example, bulleted, um, format. You can have grid comparison. What we have noticed that in most cases, it's actually a comparison grid or small chunk of content that gets picked up by LLM rather than long form.

Is that true all the time? Absolutely not. You guys have to test it. So this is more to kind of get you an idea. So far what we have covered... I'm gonna test you guys, by the way, at the end of it, so. So what, then we got how, right? So what we kind of figured out, hey, we need to create this content based on the topics that, that people are asking.

We, uh, make sure that it's crawlable. What comes next? Where are we gonna publish it? So that's the th-third part of it. So the way I see it, you need to deploy it pretty much across the board, but this is kind of a high level that you can take a look at. So we look at existing pages. So if you are a retailer, your product grid page, look at your product grid page.

Do you have content? Can we add that prompt answer within that product grid page? Does it apply more to a PDP page? Uh, for example, if it's about summer wedding, we can create a summer wedding journey page. So the thing is, you need to see the scope of where all you can create the content. So if you think of it, that's a lot of work that needs to be done, but unfortunately, that's the nature of it.

And one-man person can't do it. How you do it, that's again, that's an operating model piece of it. We can talk about it. And I had a couple of discussions across about it, and people are trying to figure it out. So the other one is publish editorial blog. Everybody knows about it. The fourth box, I have it in green on purpose because that plays an important role.

That's the voice of the people. Voice of the people, what are people talking about you? And that's where Reddit comes into play. And I can guarantee you, you will hear so much about Reddit, so I'm not gonna spend too much time. How to do Reddit, what to do Rebb- Reddit, uh, where to go into subreddit. There'll be a whole presentation that will probably talk about just how to optimize and what to do within Reddit.

But you, you can get an idea. Being in the community, people talking about wedding suits, right? And that's what the LLM looks at. Hey, that's where the trust and confidence comes into play. We can create content, but if the trust is not there, there is no cons-consistency, even there's consistency. But if people are not talking about it, it's not gonna pick up your, uh, content.

So we talked about what, how, and where. Pretty simple if you think of it, but you'll be surprised because the thing is, again, you have to, and I'm again reiterating that each one of them can be a piece that plays an important role. So the next part of it, so you do all of this thing, the elephant in the room, how do you measure it?

Anybody figured out how to measure it yet? Right? So the most popular one is, that gets floated around is visibility, right? Pretty much any SEO tool you go out there, they will show that, hey, we are measuring visibility for your brand. I'd anything, it's completely bogus because there's no standardization so far.

If, and I can guarantee you, if you are a brand, you go back, go out, look at four different vendors, you will have four different visibility scores. Hundred percent. That I'm confident about. So what do we do? We still look at visibility. We are working with a third-party vendor, but we need to have some kind of a benchmark.

So-- and I'm more than happy to share who we are using. But we use that tool to kind of build out what is the baseline, so when we start optimizing it, we at least know that we are moving above or going below. The other thing we definitely look at it more closely is citations and mentions. And the reason being because you can monitor in some form or shape how many times you are getting cited or how many times you are getting mentioned.

You can go personally, actually go check it out. You can search for that prompt and see if you're showing up or not. Manual, not recommended, but still. Like visibility score, like, who knows? And the thing is, uh, how many times we show up. The other part of it I really like about citation and mention, so for example, if there is a prompt, you create a content, you publish it, you are not getting cited, then you can actually see the performance of it.

Did that piece work and started getting cited or not? So that's a great win actually, because if, for example, my editorial team is creating about twenty different pieces, then at least I know that, hey, these twenty pieces now I-- it's getting cited, and it actually give you the prompts it's starting showing up.

So we use that tool for that. Let's look at a quick example of, uh, what, how, and where. Ah. So this is a real customer ask. As you can see, the prompt is not a fully completed sentence because that's a lot of time people search for like that. Winter wedding guest outfit, classic, not too formal, needed by Friday.

So when you look at the AI answer, this is how it actually looks at. It will give you a primary suggestion. It will give you alternative comparison, uh, charcoal textured sport coat versus turtleneck, Chelsea boots, fit sizing, timing, how long it takes, Friday deadline, confidence and trust. It's present in, uh...

It will give links to Reddit. It will give you, uh, a link to the publisher or the n-not. So if you go back to my earlier point, what do we do with it? We look at the, uh, this answer. We were not showing up for this winter wedding, so we start creating content for it. So again, I will quickly go over this thing, which is where and what.

So we leverage-- we created a guided page, leverage the PGP, lever-leverage PDP, then we also went and participated in subreddit. So for example, we created quick answers, uh, quick answers strip on the pages, so on and so forth. You get the idea

Now the next question becomes: how the hell do you scale this thing? Ali, you got, got, uh, one, one prompt. Woo-hoo! Right? So here's how we go about it. So we start from the top, identify the topics. Pretty basic, right? There are multiple topics. Obviously, you get what is your key topic, that is suits. You can t- use a tool, third-party tools.

But for organization which is medium-sized, there is a whole goldmine of information if you look at your call centers and chat, chat transcripts. Again, there will be a lot of work that needs to be done to extract it. You can extract a lot of questions that people are actually asking. Leverage that. Go into the product reviews.

Go into the customer reviews. Go into the Yelp reviews. If you're using some other third-party vendor to kind of get those reviews, get all of this thing and centralize it, and that's where the AI comes into play. Ask the agentic AI to kind of do the summarization, right? So what we do is, for example, hypothetical number, all suit prompts, there were about thousand.

So then we kind of clustered them into multiple categories. So we have, for example, suit fundamental prompt suits and wedding suits, for example. So then we got a smaller sample set of prompts, then we leveraged that prompt clusters to create that template and then created that content, right? What AI answer it looks like, what content to create, and then where to publish it.

Now, you will see on your right side this whole anti-gravity. I don't know, has anybody used anti-grav-- Google's anti-gravity tool yet? Oh, you should definitely check it out. It's a free tool. It's kind of, oh my-- It's, it's amazing. I'm running out of words, and it's not amazing. So what you do is you go into that tool.

So I'm sure you have t- heard of vibe coding, right? Where you-- even if you're not a coder, you can go into these tools. You can tell them that, "Hey, I want this tool to do A, B, C, X, Y, Z." It will create that software within minutes. It is kind of the same thing, but it's really goes f- good for what we want to do.

So that's an example. So what you see that objective, I told that to AI. You can install it, and it will like literally, it's, it's good to go. So what I ask for, execute a holistic AI visibility audit and content blueprint for men's suits category, focusing on creating reusable content patterns, right? So that's where we are want to.

We want to get a content pattern that is AI showing up to maximize citations. It create-- And you can use actually AI to even come up with this. Then you use it to tell it to anti-gravity to do this task. So task one becomes entity pattern audit. You can tell the competitors, it will tell you, "Hey, competitor one, competitor two, competitor three." The browser will open right in front of you It's, it's absolutely amazing. Definitely check it out. Highly recommended.

That's the next piece of the pu-puzzle. Who owns it and who executes it? So far, what we have seen actually, that AEO performance falls under SEO, right? Because, hey, SEO traffic is going down, so it's probably going to AEO. Tell me what the visibility score is or tell me, uh, how much traffic is going towards it.

But most of the work that I told you about, for example, technical AI foundation, right? On-site content, working on community, multimedia, analytics, they're not owned by SEO. How the hell I'm gonna get that thing done? Because the thing is There are so much interdependency. So with-- when it comes to SEO, even if you look at it, right?

There was some pieces of it, but there was some part of it you could still do it. But in this case, the on-site content can be owned by the editorial team, the community t- uh, community work that is on social media, Reddit, is earned by the social media team. If I go and talk to social media team that, "Hey, I have identified these prompts.

I have created this content. Can you go participate in subreddit?" He said, "Dude, I have my day job." So what we have done is we c-created this layer. So you can see that dark brown layer. So this is a governance squad. So what it's doing is the first step I talked to you about, right? Is say, "Hey, what is that con-- what is that prompt people are looking for, and what is the content that needs to be created?"

So the, that layer or that team, it can be the SEO team, it can be your AO team, whatever is the team is. So they will create that content or recommendations. Then working along with it, the engineering and the content creative team. So the most important part of it, that there has to be an executive sponsor.

Because normally, when most organization, SEO falls under marketing team, but now we are talking about stuff that needs to be done by the product team. In some cases, SEO team is o- under product organization. Then you have to go to the marketing team to go do the social work. So there has to be an executive sponsor for that team to kind of push that initiative. So this is a structure we have put together. We are kind of working towards. So far it has been going well. So before I sum it up Let's see where we are heading

So far what we have talked about is a search experience. For example, user ask, the assistant suggest, right? The ChatGPT will suggest it. User clicks, user confirms The next part of it is where the entire experience is within the AI platform. Has anybody tried Zillow within ChatGPT by any chance? So what it does is, I'll quickly walk through it, is when you go into ChatGPT, you start looking for, "Hey, I want to buy a house in Palm Springs under two hundred thousand," or whatever. You will see a map that will actually pop up

Will pop up within ChatGPT. You're not even leaving the site or the platform. The entire experience is over there. So you go through that experience. The only time it leaves is when you have to book an appointment or something like that, or, uh, get a time to come and visit the house. But that too will eventually go away.

So the whole experience is actually within going into the platform. The next part of it is actually, uh, where the user is not even in that process. So over there at the bottom where you see, right? Agent decides, agent executes, and it's done. You just tell the agent that, "Hey, I want to..." Uh, buying a real estate that quickly won't be a good idea, but, uh, through an agent, for example, "I want to rent a suit for the evening, and these are my requirements."

It will go in, and it will go ahead and rent it out. Now the question becomes-- And that's why it's very, very important that the content that you create is machine-readable because with... I don't know if you guys have done open clause or not, because those agents are already ready. It's just a matter of fact or it's just a matter of time.

So here's a quick example of how Men's Wearhouse potentially looks like. If you look at within ChatGPT, it's not there yet, but sharing it with you. So if you go into ChatGPT, you search for at Men's Wearhouse, what should I wear? The entire product is within the platform, and then you can pretty much go through the whole process

So to sum it up I know there's a lot going on. I would say make your life simpler. Start with those three questions, what, how, and where. We know this shift is happening Let's keep the conversation going. That's all I have

I got about exactly one minute and 45 seconds. I'm open for questions. Say that again. Yes That's it. Let me, can I get- Yeah, I'm more than happy to share the slide too

Audience Member: All right. Thank you. Um, I think there's a lot of AEO and SEO vendors out there, um, you know, claiming they can do, you know, exactly what, what this is doing. So because all the measurement happens pre-post, right? You can't really do a true AB test type of a experiment. So how do you, you know, from the, from the, uh, I guess the retailer perspective, how do we pick the right one, um, if the AB testing is not possible?

Ali Haris, Director, Organic Search & AI Innovation, Men’s Wearhouse at Tailored Brands: Yeah. I would suggest definitely talk to their customers of who they can do, because to be honest, at this point, everybody has put-- thrown a hat in the ring from a vendor and partners perspective. With all due respect, I know some are amazing, but I don't think there has a standard for it. It's very difficult to measure.

The way we are also measuring it, one is we are using Google Analytics to at least know how much the ChatGPT traffic is coming from ChatGPT. The other thing is we are also analyzing our logs because that is also important to kind of see that, hey, if, if ChatGPT, how much it's coming, it's going over the time, what pages are they crawling, what pages they are not crawling. Uh, so, so again, it's very, very limited. I, I totally agree with you. The-- I don't have a very solid answer, but yeah. In that case, thank you everyone