AI-ready schema: Feeding Google’s answer engines with structured data

Alex: 

Because maybe it will be and it will be the new Google. But maybe it’s the MySpace—and maybe there’s a Facebook out there that, you know, is just coming up. Maybe Claude is that underdog right now that’s just going to leap through some technology that just, on the mass market, helps everyone.

Andrew:

Hello and welcome to SE Ranking’s DoFollow Podcast. Here we bring you in-depth conversations with SEO, digital marketing, and search experts, diving deep into the latest trends and strategies to help you stay ahead. Be sure to subscribe to keep up with everything happening in search.

Today we’re discussing an exciting topic: AI-ready schema feeding Google’s answer engines with structured data. We’re joined by Alex Moss, a UK-based technical SEO consultant and WordPress developer with 20+ years in search. He serves as Principal SEO at Yoast, the WordPress plugin running for more than 13 million sites. And he’s the co-founder of the boutique agency FireCask. Known for his deep work with structured data, site performance, and EEAT, Alex writes for Search Engine Journal and speaks at conferences worldwide, translating code-level fixes into measurable organic growth.

And today he’ll be sharing valuable insights with us on this evolving area of SEO. Welcome to the DoFollow Podcast, Alex. Thanks for joining us!

Alex: 

Thanks for having me. What an introduction as well. It makes me also realize my age a little bit.

Andrew:

I tried. But yeah, you definitely have a lot behind you, and hopefully still a long road ahead. If you don’t mind, I just want to jump right in and ask you to tell us about the ins and outs of your journey into traditional SEO and then how you made your way into structured data. And when did AI come into the picture for you?

Alex:
Well, traditional SEO, I mean, for me, I think it’s like anyone maybe over the age of 35—we fell into it as a diversion from something else. And it’s something I kind of just understood. But if I go back further, I didn’t realize I was doing SEO when I was a teenager. I used to build websites, had a South Park fan website, and I was doing black-hat SEO without realizing that that was what I was doing.

So I would say I would actually have it as an aggregator site. So I collected content from other websites. Like one site had pictures, another site had scripts, and another site had videos. And I would just pull them all together, and I started ranking for it. And then I did a bit of ad fraud, I guess you could call it. I was doing against the TOS where I did a one-pixel-by-one-pixel iframe on the site, on every single page. And at the time, you got money per impression. So that really worked.

And then I found out I could, like, rank for things that had no relevance at all to South Park. Fake Nike trainers—if I just added content into the site, if I hid it behind a white background with white text, all of those little nifty things—I wasn’t doing it professionally. I was just a kid mucking around with HTML. And then it turns out that I was understanding the first foundations of traditional SEO.

It kind of went from there. I fell into a web design job after doing freelance work through university, and then that just got me into SEO. And then fast forward to when structured data happened. That was around a few years after I got into WordPress, which was about 2009.

Structured data I started getting really interested in because you can have layers of information based on what you’re trying to present on a page. And I thought that was really cool. And I remember I developed a plugin for sharing on Facebook and Twitter using social structured data. Did a few talks on it back in 2013. I remember doing one at Brighton, and that ended up getting sucked into Yoast as well.

Like, Yoast created his own version of it as an addition to Yoast SEO. And obviously, I kind of went, I’m not competing with that guy or that company. So I kind of just left it alone, and then he continued it. But we worked together during that time.

And then fast forward to being at Yoast. And now I get involved with the product development of what gets output in terms of structured data for Yoast, which is, again, someone who I work with now 12 years ago on structured data. So it’s kind of a serendipitous circle that’s gone around—it’s always been around.

It’s always been an interest of mine to understand entities and how they all relate, and things like the knowledge graph, which really spread how the internet and crawlers interpret things—and now LLMs, right?

Andrew:

Fantastic journey that you sort of fell into. Yeah, it took you to places. And I wanted to focus a bit more on the topic of today’s discussion about structured data. So how has it evolved over the years specifically, and what role does it play right now?

Alex:
Well, at the beginning, it was just complimenting SERPs, wasn’t it? It was just, I think, the first ones were for recipe and star ratings on reviews. And that really helped. You could see how that helped people understand a bit more context of what was going to be inside the page before they clicked into it.

And then they brought more stuff into the SERP. And then obviously Google may have monetized off the fact that we were giving them that structured data and then brought it into ads, made it more compelling.

But actually, over the time, you know, from 2011 to 2020, it went from just under 300 types of schema to nearly 800. It’s actually 790 in 2020. And now if we fast forward from 2020 to today, we’ve got—well, the last time I checked, it was 806 when I did my talk in the Zagreb SEO Summit. Since then, which was only six weeks ago, they’ve removed seven types. So now we’re back to 799. In 10 years, it’s nearly tripled. And in the next five years, only nine types have been entered.

So it’s kind of peaked. But does that mean it’s stopping? Or does that mean it’s kind of gone to the ceiling where it needs to be? And maybe that plateau is actually acceptable. That’s been really cool. But if you think of the evolution, it’s not just gone from one layer of structured data, then entities and knowledge graphs made relationships between each other. And then that made everything go just well beyond schema into all types of ways in which crawlers and now LLMs get to interpret different types of content.

But obviously now with LLMs, you go beyond schema and structured data into structured content and formatting, which, you know, has now widened the scope of what we need to be looking at and monitoring over time. And of course, you’ve got LLMs.txt—you know, that didn’t exist about a year ago. It’s less than a year old. And whilst it’s not like the de facto like robots.txt, this is now an added element on how an LLM ingests your site and your content within the site.

Andrew:

Yeah, it gives us enough reasons to be frustrated and to always think about having everything under control. I wanted to get your understanding of—because I realize everybody wants to use AI to their advantage, wants to cut corners, wants to figure out how to speed up their processes—because everybody’s going to be doing it, and you’re just going to be left behind and gone if you’re not using AI just to get a lot of stuff done quickly. So I want to get your thoughts on what you think AI-ready schema or AI-ready structured data is.

What do these words mean to you?

Alex:
Well, AI-ready to me is—is it optimized to that way, and therefore, can it be investigated and monitored and reported on?

But if we say AI-ready schema, maybe that’s like too—would not be broad enough. And in the past, we didn’t need to be broad enough. But now we need to think beyond structured data, which, you know, provides meaning—and that’s fine. But now you need structured content and the way in which that content flows and is presented. And it has to be concise. It has to be structured.

Now we have to think a little bit more—weirdly, as technical SEOs—on how content’s being output. Which, I mean, in my previous experience, had been a little bit split out. You know, you would do keyword research. You would then give that to a creative and content team, and then they would do what they needed to do to produce that content. And beyond, you know, headings and, you know, rich snippets like FAQ snippets and stuff like that—it was kind of hard to do that.

And then, over time, solutions like Yoast in WordPress and Shopify helped with outputting the right schema in the right way, and more importantly, being up to date—because it’s got more complex. Whilst the types haven’t increased so much, Google support for existing types has increased. So we can use products and product variants. Product’s been around for ages, but product variants haven’t been really supported in the way in which I would have wanted it to be—until the last year or two, really.

Actually using that to experiment and investigate things—that’s maybe using Claude or ChatGPT or one of those LLM platforms to see how it actually interprets it. But I guess that’s the way you’re prompting to actually investigate those things.

So, as an SEO, you would now, I would say, have to have a library of prompts per client to understand things that you do as your job. And one of these things is asking it about—is it the right type of schema? Is the schema there? And interpreting it beyond just the page. And that’s just going on diagnosis.

But AI-ready schema and AI-ready structure in general should be the old-school things that we’ve been doing all along—but thinking about it with a new technology in mind that’s receiving it.

Andrew:

Well, that sort of is a bit easier to hear, because we’re just doing what we did before—but just adding something, not just replacing. Hopefully that calms down our listeners.

I wanted to have a bit more focus on the stuff beyond markup. So, like, the writing styles, the heading hierarchy, the lists, other on-page formatting elements—how do they influence the AI’s understanding? Because you said, like, structured content, not just structured data. How do they influence both the AI’s understanding and just the overall visibility of a content piece?

Alex:

The native way that we’ve been doing it—crawlers have been crawling sites. Now, LLMs don’t crawl sites. They ingest content within sites in a different way to crawlers. So now, all of a sudden, structural content is more important because an LLM is ingesting it rather than just reading data. And whilst LLMs can read schema—that’s optional—but content is not optional in general. And the structure of which needs to be concise. And now that’s maybe something to think about a bit more.

And, like, elements include the order of the information, the hierarchy of concepts, the relationship between those concepts. I’m talking about the same stuff that we’ve been doing, kind of as an SEO all that time—is the content there, and there’s internal linking and context all there. But we need to be a bit more concise.

I usually go with one of those examples of, you know, one of those companies who are trying to be futuristic. And you go to the about page, and I’m reading four sentences in, and I still don’t quite know what you do, right? Because you’re trying to be clever with the text and trying to write.

LLMs don’t care about any of that. They just want to get to the point. What do you do? And be clear and concise and honest and transparent—because LLMs don’t care about content marketing fluff around what the actual point and the intent of it is. So make all of that fluff a bit more redundant, and then you’ll find you’ll have less but better quality content. And then you’ll be able to hopefully extend on that as you go on.

Because it’s not just about JSON-LD, for example. It can’t just take it like a crawler can. Read that—it’s very nice—and it’s interpreted in that way. But now LLMs contextualize and ingest that content, and they do it in a tokenized form. So what sometimes I like to do is—ChatGPT have a free tokenized tool, which I’m sure you can share in this podcast—and you can just paste in content, and it tells you exactly how many tokens have been sent. And it highlights areas of the content which use more or less tokens. So you can understand the relationship and context and how complex the LLM will have to work to interpret and ingest that content. So, token optimization, I guess, is partly it—but it helps you understand exactly what the LLM sees when it ingests that content.

Andrew:

Even besides structured data, we’ve had a conversation among our content team—what should a blog look like nowadays? Are we still writing for people? Is somebody going to actually open and read the intro and go to the outro? So right now we made a decision, sort of yesterday, to cut the length of all of our content pieces by maybe two or three times and mostly write for AIs—because everything is going to be pulled out from there.

We actually did consider having, like, a TLDR—but for, like, people or for search engines—to have an article but then condense it or maybe hide it from people and have something more vast available to search engines. And we’re still trying to figure that out. But yeah, I think we’re not gonna go and waste more of our resources on creating even more pieces of content—we’re just gonna focus on serving those rich snippets, specific answers.

So yeah, hopefully that will work. Yeah, we’ll see. We just had an understanding that we added key takeaways to our blog posts like a year ago. And since we added that, our engagement read time—the average read time—was a minute and 10 seconds. So people just log in, open up, they look at the key takeaways, and boom—they go away. So at the same time, we sort of satisfy their demand, but at the same time, they’re not sticking around with us for that long to build a relationship, to read into the whole thing. So it’s a very tricky situation.

Alex:

Yes, it is. And I guess it depends on what you want the summary to do. Do you want that knowledge to be solved? And if you do, then yes—either a human will just know what that summary is and go, “I don’t need to read anymore.” But I guess it’s part of our job now to, in those TLDRs or key takeaways or whatever you want to call them, maybe entice or give incentive for further reading. And from a big content point of view, we now have to think, well, we should just be writing things that AI can’t just answer, right?

And that’s quite a hard conundrum to be in because you want to provide information, but then if you provide solved knowledge, then the LLM will appropriate it. Don’t want to say “steal,” because we’re the open web and we’re giving the information away—and it’s just being appropriated in a different way and owned so that they don’t click you. And now we’re in the zero-click issue that we have. And now we have to think, well, how do you remove that zero-click issue? How do you make that click happen?

And I guess that’s just like, in a marketing way, giving someone something to whet their appetite that might require more investigation in order to come to your site—you know, which is going to be harder and harder as things go. Because there’s going to be, again, a big debate of what is going to happen with the open web. We’ve got people like Cloudflare, for example, who are now taking on AI bots in general by blocking them from crawling all of these sites—which I get the point. And you kind of want to own the content. But by doing that, you are removing any visibility to the open web—which is, again, part of what our job’s been for a generation.

So what happens there? That’s another—you think AI in general is bad? The—it does to the internet and how people intake information—I don’t think we’ve been thinking enough about what that’s going to be like in the next five years. And what happens if, like, websites die off? So anyway, like, I’m sure people think about it. But what if they do in the next five to ten years? And we can’t predict all of what’s going to happen next, or what AI will direct us to do—because it may start telling us the best way in which to, you know, output content and data for everyone’s benefit. And I hope it figures out a way to do it so we all win.

Andrew:

I’m a proponent of the idea that we still have bookstores—people buy vinyls—so I’m hoping that there’s still gonna be maybe, like, a camp of people that are AI-free. And you’re just gonna go to them because you just know for sure that that was all by people—and you just love the idea that it wasn’t just some faceless entity that you’re excited about talking to, you know?

Alex:

Yes! Yes. And it’s, again, the unsolved knowledge aspect. And even take away AI from this—I saw a bit of Steven Bartlett’s Diary of a CEO interview with the Godfather of AI. And his first question was, like, “You’re the Godfather of AI—what’s the prospect for people’s careers in the future? And what’s that going to do to the job map?” That was question number one. And his answer immediately was: learn how to be a plumber.

And that scared me—because, yeah, alright, so we’re going back to basics. Like, having a coup against digital and online—where we’ve been training now for 30 years that technology and digital is the future. And maybe from a human’s point of view, it is in some aspects—but maybe not in others. But again, it’s that risk and opportunity of what to do. And I don’t know how to be a plumber. So I’m hoping, again, there’s a solution for everyone in the long term—which I’m sure there’s always going to be. As long as someone wants to discover something, people like ourselves are going to facilitate that.

Andrew:

Alright, I wanted to sort of go back and ask you to paint us a picture of a situation—like a real-world example—where an AI-optimized structured data actually fed Google’s responses and its answer engines, well, the new ones, the AI. If you could sort of accompany that with a measurable outcome—like data, something that helps us understand how you got there and how you understood that that is a good result—I’d really appreciate that.

Alex:

Yeah, yeah. Well, there’s not many. But one thing that springs to mind is: in AI Overviews, it gives you extracts from text to compile it and synthesize the answer. Now, when it does that, it uses part of that content. So again, if I don’t say structured data, but I say structured content—Google has been able to, with the fan-out queries, find this page, look at the whole content, and they’ll go, “Oh, right, well, based on the chat—based on the query—if you want to just call it a query for a sec—I’ll take this part of the content and I’ll bring it into the AIO and then I’ll serve it there.”

And then when it’s cited on the right-hand side—this is what I’m seeing at the moment—it gives you the source link. You go to the source link and it doesn’t just take you to the top. It takes you to the area that was contextual to the AIO that it gave you. Now it has this, like, hashtag highlight in the cut. If you look at this and you go to the URL, you can see that now you can extract that out, and you can use that—and say GA4 or Tag Manager—set that up, and that’s something that you can monitor over time.

And then, again, it’s not super big at the moment, but it’s going to increase as time goes on. And this will be one of those things to monitor—how much are people coming to which pages where certain elements are highlighted? And then, after a bit of data has been collected, hopefully one can find a pattern in that and then either use it to their advantage or find a weak area in another part of the site or business and then take that and apply it into there. But yeah, there are other tools out there, of course, but from a real-world example from a search, that’s the best way I’m thinking at the moment that you see it.

Andrew:

I wanted to mention that SE Ranking is rolling out an AI Search Toolkit. And in that Search Toolkit, we have the opportunity to track and research AIO results as well as ChatGPT. And the thing that I wanted to add is that sometimes people will get asked a query from ChatGPT where there’s a brand as a result, but they don’t click on it. And the way that we can sort of keep track of that data is by analyzing our server logs—our own server logs—just to see if ChatGPT visited our pages. So that is something that we’re trying to do to fill in the gray zone—exactly all these points of touch that sort of help us connect with these AI-powered search engines.

Alex:

That’s kind of not easy for us to interpret, but it’s much easier than other things—because you’re getting a URL and you’re getting a query back in return, which is kind of similar to Search Console, for example. And our SEO brain should be able to relate to that quite easily—because we’ve been doing that for a long time anyway. And it’s a very good way of monitoring that stuff.

And maybe activate or deactivate something. Maybe you’re finding that G is accessing this page loads and loads of times, and actually they’re getting the solved knowledge. Let’s remove some of that solved knowledge to either let it… and see what happens. Will it become less visible, or will it still be cited? But now it can’t take everything that it knows or knew once before—unless it’s caching that data, which it isn’t—and then it’s providing it with less content. What happens as a result? Does your site or your page have less visibility in AIO? Or does it have better ways of having someone come into the site because of that more limited content production that you do?

And that’s going to be another quandary of what to do. How much content do you provide? Because again, we’ve been trained to provide as much as possible to attract people to come in to know more.

Andrew:

Yeah, we’re going to be doing a lot of research this year and publishing a lot of things—because everybody’s interested in understanding how it works. How do these AI output sourced resources—like, where does the information come from? Can we control this in any way? Or is it always going to be up to just AI and always going to go against us?

One thing that I wanted to ask you—is there anything that sort of gets more weight, more focus, more attention when you optimize structured data for AI search engines? Does everything have the same weight, importance, or is there something that you know for sure gets bigger praise from Google?

Alex:

I wouldn’t say anything’s prioritized, maybe, but if something isn’t populated, it’s not going to be found. My output’s always: be populate as much as you can. Don’t prioritize anything. It’s like schema, for example—isn’t less important, but it’s also not more important as it was before AI Overviews existed. It’s just: keep doing what you’re doing, but make sure it’s all populated and try and do as much as possible where it can be done. Again, product variants, right?

You’ve got shoes in seven different sizes—then that’s seven different product variants. And if you’re in an LLM, which now knows your shoe size and you aren’t specific to it, then maybe it might not find you—because it just sees that you own a shoe. But you, as a searcher, may be very specific to find a size seven. And you can even, again, get that without LLMs.

I’ve been on sites where you tick your size, and then all of these shoes that you wanted before just disappear because it’s not in your size. And whilst that’s annoying, it really refines the process and stops even more frustration—because then I’m now no longer clicking into 10 different shoes, none of which are going to be in my size. So populate as much as possible. Don’t prioritize anything particularly. But yeah, I would say maybe, if anything, just prioritize anything that you would as a business—like your cornerstone content, your really important products.

Andrew:

Is there anything that you notice people doing wrong often—just like a huge mistake or just a minor mistake that sort of has huge repercussions? Anything that you’ve noticed over the years, people getting wrong with structured data?

Alex:

Experimentation beyond what’s there. So I know that, like, the perfectionist in me would love a hugely complex entity structured data on my website—but there’s only so far I should be going, right? First of all, does the internet need to know that much granular detail about me? But on the other hand, like, does it care? Like, does it care about all of my skills? But do I need to attract that just for Google’s benefit—when they’re only now supporting seven less than they were six weeks ago?

It’s one of those—don’t make mistakes, but don’t go too far heavy. That’s why we’ve not, in Yoast, necessarily built a really easy UX to let you do whatever schema you want. We’ve got the Schema API. So if you know what you’re doing, you’ll be able to do it. But what we don’t want people to do is not know enough about schema and think that they can just put 200 different properties on there and think that that’s—well, if I have 200, something’s going to pick one of those things. That’s not necessarily true. That, to me, is spam—and a waste of bytes on source code—and therefore a waste in the internet. It’s like pollution in a way.

Just again—if you’re concise and you just need to do what’s relevant and output that, don’t try and abuse the system, you know? Because now LLMs—they’re going to know what you’re trying to game much quicker than humans.

If you think about Google core updates—kind of over the years like Panda, Penguin—they’ve been reactionary, haven’t they? To bad players in the SEO world and super black hats who’ve been trying to abuse, successfully, holes in the algorithm. And they’ve just been filling holes all this time. LLMs—no regulation, nothing, no teamwork—it’ll just close that hole. And therefore black hat, I think, is going to become redundant as time goes on.

Even link building, right? All of a sudden, unlinked branded mentions are now a positive sign—where black hats have hated that all that time. And now I go back—from an agency point of view, I’ve always said unbranded mentions are good anyway. Like, not from an AI signal—this was, like, 10 years ago. Like, it’s still good to have those things. It’s word of mouth—just like anything. And yes, in a perfect world, you would have either a branded anchor text or unbranded anchor text—no matter, depending on what you’re up to. But now I’m happy that we’ve got those things—because they’re eight years old, and they’re still there.

And now LLMs will consider that as aged authority—which now some agencies might be going, “Crap, I’ve been doing it one way all this time to try and game an algorithm—where now it’s gone beyond anything I can control. And history in content might be really important now all of a sudden.” And it’ll serve as even more important. It’s like a library, right? All of that stuff is just in a massive database—can still be referenced no matter whether it’s linked to another book.

Andrew:

I have a very big question. I get that it might be too difficult to answer, but from a technical standpoint, how does structured data interact with Google’s AI systems under the hood? If you can comment on that in any way, I would be happy.

Alex:

I would say any answer I’d be guessing—because, like I’ve been saying in the past few months on different things, the next two to three years are going to be—that is when the dust is going to settle. But the dust has not settled yet in terms of this ecosystem, and it’s not settling anytime soon. And in fact, I think it’s going to get more dusty as the next year ahead goes—which also is really annoying, because up until now we’ve had great year-on-year data.

And all the things that I’ve been saying now—in a year—may not be as important, or may even be really important, or have changed in a way that you can’t measure on it anymore. So for the next three years, SEOs out there won’t be able to do proper year-on-year data because of—well, it’s been completely changing.

It isn’t just an update to an algorithm that’s changed 10% of sites. This is something that’s touching every single vertical in every single industry—which is very interesting to study around, right, as part of our job.

Because most of my friends don’t even know, like, what’s going on, right? And whilst we can say we can’t guess, we kind of have some inherent knowledge as to what’s happening around us and the way it’s going. But again—in three years—I don’t think I could predict too much that might actually stick, right? I don’t think many people can.

Andrew:

It’s interesting to attend different team meetings and see how different departments sort of react to this. Like, the analytics team talks about one thing, the content team talks about another thing, the video production team is all about Veo 3 and keep that conversation going. So, yeah, it’s extremely overwhelming, and it can be stressful—and, like, there’s no way of knowing what to do. Like, what things should I focus on right now? Maybe I should go to college again.

I’ve heard a couple of weeks ago—nobody’s looking for a content creator anymore. They’re looking for a content updater—who’s somebody who knows about SEO, knows about social media, knows about optimization. So, it’s just basically three people in one, because AI can help you take care of everything else.

So that is something that I’m already thinking about. A lot of us are worried about what to do. So we’re just trying to sort of jump on everything else and just try to become an expert in something else—or, like, a prompt expert at least.

I wanted to ask you about the tools or platforms that you use—or that you rely on—to validate, create, and monitor maybe schema and structure. Anything that you can address?

Alex:

I feel like all the tools I would recommend are, like, the ones that are out already, right? You’ve got schema visualizers, you’ve got Classy Schema, the old-school rich snippet structured data stuff, Search Console.

For structured data, that’s the stuff that I still audit, right? And all the big tools have it as well. But when it comes to how content—and your brand in general—in terms of tools for monitoring, they’re, like, all being born now. I feel like there’s a new platform or tool every month coming out, which is also quite interesting.

You’ve got, like, Profound, Gumshoe—which I tried, which has those free little trial—and Dixon Jones, who used to work for Majestic—he’s now got “What AI Knows About You.” WAIKAY, I think it’s called. And then you’ve got other big tool providers who are now doing their AI toolkit or auditing. I know SISTRIX just introduced theirs as well—I think in beta, or has it just come out of beta? I’m not sure. But there’s a few out there, but again…

The dust is very much not settling, right? There’s going to be a few more coming out. And I guess it also is going to be a weird one—which one’s going to stick? Because even now, can you say that OpenAI in three years is going to be the dominant force? Because maybe it will be and it will be the new Google. But maybe it’s the MySpace—and maybe there’s a Facebook out there that, you know, is just coming up. Maybe Claude is that underdog right now that’s just going to leap through some technology that just, on the mass market, helps everyone.

But at the moment, you know, the only people—all my friends and family who have no technological background—they don’t know what Claude is. Some know about Gemini, but basically it’s all ChatGPT. And they’re now verbing it as well. Like, instead of Googling it, they’re now ChatGPTing it. And I find that interesting—what repercussions that will have.

But I feel like OpenAI should kind of be opening a tool for us as well—because if anyone’s going to have that firsthand data, like Google made Search Console and Bing makes Webmaster Tools—perhaps there should be a solution inside OpenAI. Maybe they’re building it. I don’t know. I don’t actually know. But maybe they are, and it would be a great solution that all of these tool providers should have—their own versions of Search Console or Insights. And then you can layer that with MCP as well, which is now—is it a hot topic or is it a bubble? I don’t know. But MCP is great because—like a USB-C for data and information that you can use and integrate with Claude to ask about anything, really, about different data sources—which is very interesting.

Andrew:

By the time this podcast comes out, rumor has it that ChatGPT-5 is going to drop really soon. So maybe. Because they’re already asking, like, for $200 to do a lot of stuff—so maybe that will also become a part of their package—optimizing even structured data.

Alex:

Interesting, interesting. And they’ve now announced in the last day—or X amount of days or weeks ago now—that they’re now getting into the browser game. And there was a bit of rumor before, but I think they’ve pretty much announced that they’re gonna be developing a browser and therefore they’re in direct competition with Chrome. Again, which is interesting—because it’s beyond just the conversational elements of their platform now.

Now it’s everything—it’s browsing history. And I do believe that if ChatGPT—or OpenAI, sorry—created a browser and they released it tomorrow, I think a lot of people would install it.

Andrew:

For sure. There’s a lot of haters for Chrome. I know a lot of SEO specialists who don’t even use Chrome, or they just use it for that one—specifically just to interact with Chrome in that environment. But they just hate all the stuff that happens behind the scenes and how they just keep track of you. So it’s very interesting to see how that pans out.

Another question that I have is about measuring the results. So what specifically do you look out to to understand that, okay, this site, this page is optimized in terms of structured data—it’s doing well. How do you understand that everything’s right or wrong?

Alex:

Well, talking about MCPs before—you can actually connect to your site as well. And there’s a couple of things out there. One’s called NLWeb and another one’s called Clarity. They’re both by Bing. Think of them as a way in which you can enable search on your site in a natural language manner. So at the moment, if you go on any site, generally it’s the—again, old-school—magnifying glass or search input. You type in a keyword, and it goes in.

But now with things like NLWeb, you can make that conversational. So, forget for a second you’re an end searcher. Now you’re back being the SEO. Get NLWeb connected and start asking questions that you’ve just asked me about its own data. And then it will tell you what it perceives. And then you can then take that information or answers/responses and then do something about it. Or it can suggest—well, if you were interpreting it this way, how have I got it wrong? What can you—what should I do to improve that? And then work on those things.

So that’s kind of on-site. But other than that, I would say the tools out there have things like share of voice—you know, general online visibility compared to that of your competitor, persona visibility, and topic visibility—which I know that a couple of these platforms who’ve got AI insights and tracking and monitoring kind of do try and get, based on—I don’t know—maybe 10 to 20 prompts that it saves and then keeps on tracking what the answer is over time.

But even then, it’s more complex, isn’t it? I think it’s ChatGPT—there was a research that just came out—that desktop and mobile, the exact same prompt, brings out different results. You can’t—it’s not as easy as it used to be, because it makes every scenario different, unique—like a snowflake—where at the moment we’ve not been dealing with snowflake data, where there’s no true connection or correlation between them at the moment. We have to interpret what that is.

Andrew:

Yeah, right now in our AI Search tool—in SE Ranking’s AI Search Toolkit—we have these prompts that produce brands as answers. So this is, like, what we focus on to help companies. But another thing that we do is we run the same prompt three times back to back—but three times right now only. Because we’re thinking, like, is this enough? Maybe we should do it five times, maybe ten times. Sometimes we even simulate, like, a location—even though it sounds weird. But, like, people would always use ChatGPT from Ukraine in English. But I don’t simulate—like, ask something like that—but it sometimes understands already.

So we do the same thing, trying to address the situation as it’s still an evolving thing. But yeah, it’s been interesting to see that results can change on the same day—like, there can be no overlaps with organic results sometimes. Yeah, it’s a very weird situation, but it’s an ongoing developing situation for sure.

And I wanted to ask you if you have any tips and tricks for seasoned structured data enthusiasts. Like, imagine that you have a colleague who is like yourself—is there something that you would be willing to share with that colleague?

Alex:

Make sure—well, I’ve said it a couple of times—product variants. I know that if we go the way in which we take on what Google and the like have done in the past when it comes to monetizing their products—they have to get ads in fast, and generally the first safest ads to put in are products. So if you’re on e-comm—and I was talking about product variants and the size of your shoe—I would say get on top of that right now.

Make sure all the right schema’s being output in the right places. And if you don’t have the ability to do it, just start tapping on C-suite’s door and maybe saying that this is going to be more important—making sure it’s as close to Merchant Center requirements as possible.

Don’t just do what SEOs might do with schema—but go a step further as though you’re doing ads, even if you’re not—and populate all of that and make sure that’s all satisfied. And then that’s more to pick up in LLMs—because that’s what they’re to do. And they have been already doing product recommendations and ads based on it. So it’s all about products here.

Andrew:

Do you think there are any niches that are not using schema or structured data enough—that are just thinking that it’s not for them? Because you said e-commerce a couple of times. I understand that they will make that connection. But maybe there’s, like, a mental health service provider or something—like maybe a SaaS.

Is there some business that may think that it’s not relevant for them, but it is actually relevant for everybody? Maybe you can sort of comment on a distant industry or niche that’s sort of maybe overlooking structured data.

Alex:

Maybe academics. Like, there’s different structured data that’s supported as well for long-form studies and courses and things like that. I know a couple have gone—I forgot which seven have gone from the actual Google support, and I think one of them may have been courses. I’m not sure.

But doing that kind of old-school—the old-school niches that are kind of only becoming digitally aware now—they’re that far behind, things that have been in filing cabinets on paper. Like, when they digitize that and you talk to them about structured data, that’s way—that’s 20 years from now. I’m just guessing—“I’m online. I’ve got email now.” That’s the thing. And they’re obviously not thinking about that stuff at the moment. But they should—if they’re getting digitized now—not just, again, structured data or schema, but how they can be entities within themselves. Legal. Medical.

Andrew:

Brilliant, Alex. Really, thank you for this conversation. And I’m going to do the wrap-up here and say that this is it for today’s episode with Alex Moss on how AI-ready structured data can revolutionize the way we optimize websites for Google’s answer engines and other search engines. I hope you’re leaving with a deeper understanding of how structured data can help you harness the power of AI in search.

Thank you again, Alex, for sharing your expertise with us today. It’s been a pleasure having you on the show. Thank you to all our listeners for tuning in, and be sure to subscribe to the DoFollow Podcast and stay tuned for more updates, insights—and see you next time.