Jordan Cooney (4:18)

Right.

Martha Van Berkel (4:18)

You know, ChatGPT is the one that's scaling, by the way. Bing is behind ChatGPT. Like, you know, they're actually sort of a search kind of component to ChatGPT. How exactly that gets woven in, I'm still exploring and learning about. But then we have, you know, these others kind of perplexities and so forth. And so I really want SEOs to be thinking about data. Right. This isn't about building a beautiful marketing website anymore. This is about making sure that the data on those website is in the language of machines so that it can show up in those different experiences. And that's where structured data plays a role.

Jordan Cooney (4:52)

Let's talk about experience. Because I think there's a huge disconnect between experiences and the experience of a user and understanding your audience and the user expectation and then how this relates to data. I think many of us are still thinking UX design those types of components. Tell us how you're seeing the integration of data to influence the expectation or the connection with users. What are some classic examples that you feel like we're just overlooking? Where's the opportunity here for SEOs to really unlock this opportunity?

Martha Van Berkel (5:31)

So I think right now like SEOs and content writers need to be working hand in hand. So that's one thing I think, like where we used to be like, oh, like the content, like they'll, they'll sort of like sit over here. Like I am seeing an enterprise like those teams coming closer and closer together. And I think that's actually good news. I think we're at a time right now where. And I'm sort of pausing because like we're still, the website is still the container of that data. And so the reason I mentioned content people is like we still right now need to put the content on the website. So we're still thinking about that experience from like who's coming to the website and who's interacting. They want less fluff though. Right. So it's, you know, we're seeing sort of like things convert better. So the whole product led SEO and sort of thinking about conversion optimization on your website, I think is becoming more than an E commerce thing. I'm hearing about it in healthcare, I'm hearing it in like big tech.

Jordan Cooney (6:23)

Yes.

Martha Van Berkel (6:24)

So I think that's sort of a big change. So how do we think about making sure the content is still on the website? So the reason I say that is because the content then goes into a data layer and then that's the data layer. Schema markup is like what's being consumed by these machines to do at scale. At some point those things might deviate. Right. At some point we might be saying, actually we're going to manage this data layer and it's going to have content that's going to be tied to something in the user experience. Maybe they'll split like. But Google right now and sort of where we've historically, historically seeing this is like the user experience because that's where you're going to send people and experience. It has to match what the data layer kind of thinks or what the structured data represents. So. So we'll see how that evolves. I do think we need to be thinking about, and this is something that I'm really excited about at Schema app is that, you know, there's some timing, things that got announced this past spring that I can't think is a coincidence. So the same week that all of a sudden Google stood on stage at sort of, this was at the New York Search Console event and was like, structured data is Enabling Gen AI. Okay, so they, they haven't talked about structured data on stage and made a big deal about pretty much since like November 2023 when we saw FAQ and that go away. There was a ton of structured data announcements. They've been kind of quiet. Like, I watch this stuff, like, structured data is what we do every day. Yeah, Microsoft. Then the same week Fabrice Cannell said the same thing. We're going to be using structured data now. I think it's because it helps them with accuracy, it helps them scale. When you do schema markup, you're actually building a knowledge graph which is like data context different from LLMs txt, which has no context. Right. LLMs Txt is sort of like a really simplified version of your data. It doesn't allow the inferring that large language models need to do. Now that Same week, Google, DeepMind and OpenAI, who by the way uses Bing like as a data feed, announced that they're going to adopt MCP model context protocol. So this is a sort of standard that Anthropic brought out and said we're going to sort of use this as like a standard way for you to kind of plug and play, you know, into our large language model. And OpenAI and DeepMind, Google basically were like, we're going to adopt this standard. Now when we looked at the standard, it's like schema markup. Like, it's actually like context, you know, with data in triple. So if you actually do proper semantic schema markup, this is what we do at Schema app. Like converting that into MCP is easy. Okay. So when we think about how do we get our content right into these experiences at scale with a data layer, I do think we're going to see this movement of structured data is actually going to enable us to then easily translate into the standard that's going to then allow scale. We'll see. The big dogs, the ones who've done this at scale are the ones who are leading the charge here. Whether we see perplexity and others go that way, I think we'll see. I do think content is still key, but the context that is what you have to do in the structured data is that element, I would say, like the, the elements text don't solve and is where the power of inferencing and the effectiveness of inferencing happens. I don't know if I answered your question, Jordan, but that's like where my brain goes.

Jordan Cooney (9:34)

Well, no, not only did you answer my question, you actually gave a lot More context that I was looking for in my next question, which is, like, what specifically is happening with data that's most valuable for generative AI? And I want to keep working off of this because I think MCP is a concept that is largely, like, still underdeveloped, under discovered in terms of what it means in integrating this universal framework to connect LLMs with your data sources and your information. And so I want to just start by asking you the question, like, what is it that makes MCP useful? Like, practically speaking, why will guidelines and frameworks or protocols like MCP be so critical to the SEOs and the web developers and the brands and businesses that want to be seen in LLMs?

Martha Van Berkel (10:27)

So we work with a lot of enterprise clients and we've been talking to them for about five years about reusing their structured data or their knowledge graph in order to enhance AI initiatives, chatbots on site search, you name it. And I'm finding that even. And like, these are big. Like, you know, you can see who our customers are. Like, even the most senior architects, like, aren't used to using Knowledge Graph technology. Knowledge graph technology has been around for a long, long time. In fact, my CTO just got back from a Knowledge Graph conference, and knowledge graphs, like Gartner called out last year, like, knowledge graphs and generative AI are the critical components to scaling and getting kind of accuracy in this AI world. And so I see sort of MCP playing a connector role to structured data and knowledge graphs to allow ease of use and ease of adoption because, like, if they can't consume the data easily. And again, I'm wearing my, like, architect hat now, right? Like, I'm a data archite. Like, their CIO's biggest challenge is having data readiness, like, to accelerate these AI and large language models initiatives, you know, so if we can kind of think of, you know, okay, we have all this structured data, or Google has all this structured data, Microsoft has all this structured data. Brian Levering, who's like the basically is the advocate for structured data at Google. I did follow up with him after the New York search console and asked him, like, to what extent are they using structured data and in what, you know, what pillars of your business? And he was like, I don't know. Now part of me thinks if they can build like one feed of like a common standard way that then all of these services that need to train and infer, you know, from all of a sudden have this data feed that makes so much sense. Like if you. I don't know, I run a software company. Like to me that's just like this is a data pipeline. This makes so much sense and it's standardized that everyone can plug and play. So now all of a sudden I don't care if people know how to do a sparkle query on a knowledge graph, which is typically like what would be required in order to get the value from the structured data that's stored in a knowledge graph. All of a sudden like if there's these automatic. It's like an, it's like the API world like so for those of you that are around in the 2010s, you know, my job at Cisco was to open the APIs to all of our services data. You know, we're sort of doing the same thing now. But SEOs, like think of this as like, you know, what can you unlock? You know, when I talk about like the three kind of value areas of schema markup, like the first is run SEO, right. Bringing context to rich results, you know, helping us get AIOs, helping us drive non branded queries, contents in the middle. The third one is around AI and unlocking that value and vertex on site search from Google. MCP is like they're going to actually use that. So I do like that's why I think it's actually a really exciting moment, you know, and where I think Google's going to do and Microsoft's going to do a lot of this work where you build structured data. The standard going to go into their standard and it's going to power these experiences.

Jordan Cooney (13:21)

I want to keep going on this topic around structured data and schema and then ultimately like how we can use protocols like MCP in our strategy. Because I think there's this fundamental reality that is we're going to need more guidelines, we're going to need more direction in order to help all of these models ingest and consume our data in our web pages for that matter. I mean fundamentally speaking, there isn't enough capital and enough resources in the world for every single one of these AI companies, OpenAI and the likes to go out and build the infrastructure that Google has built. Google has the biggest, most sophisticated infrastructure.

Martha Van Berkel (14:08)

Yeah. Which is why ChatGPT is using Bing.

Jordan Cooney (14:11)

Yes, exactly.

Martha Van Berkel (14:12)

Because it like why the cost of actually doing the inferencing is very, very high. And so in some of their documentation they're talking about like so we'll search Bing. Right. And then kind of do sort of smaller processing. So they're already kind of thinking about those efficiencies.

Jordan Cooney (14:30)

Correct. Not to say that the chat, the new ChatGPT search and the new ChatGPT Chrome extension is useful, but that's a whole different conversation around selection for consumers. Yes, but the reality here is that there's a great need for structure to help enable these platforms to get access to all the things we have or at least a prioritization of what we have. So my question on this is, we're recording a week before Google I O. What do you think's coming? What types of announcements do you think are coming in this week ahead of Google I O And the best platform that Google has to give not only the SEO community but the entire Google community some context as to how they want to consume and access data and information.

Martha Van Berkel (15:21)

Well, I know they're not going to do an API for structured data because I asked Ryan Levering the question and he said they're not ready to do that. So I don't expect that to be what's coming out. I'm going to answer your question from Google I O but like I need to share something else that you got me thinking of in the earlier part of your conversation.

Jordan Cooney (15:36)

Yeah, absolutely.

Martha Van Berkel (15:38)

One of the shifts we have to think about and we've, I feel like I've been talking about this for a decade. So people think about the requirements of like from Google on like how to get a rich result right? And it's like I'm going to optimize these bare minimum things right? On a product, on a service page, et cetera. Like we've got to get out of that bare minimum mvp because in order to articulate our content semantically and holistically, like we aren't just hitting the bare minimum. Like we need to explain what the whole page is about and what all the elements are. What are the topics on it. Like we need to make it as easy and as clear. So I think that's like a shift SEOs need to think about is like rich results are like, yes, they still drive higher click through rate. They're actually performing really, really well. But like if we're going to get the benefit to the large language models to use it for inferencing, we have to go deeper. Like today on, on schema.org there's like 800 different types. Google has like 40 types that they call out and they're only touching on 10% of the properties for each one. So I guess think about depth and accuracy as sort of like, I think that's a shift we need to think about if we're going to truly build that data layer that's going to inform and instruct large language models. Okay, so that's like one piece. So do I think Google is going to tell you to do that? Maybe, like, that would be a way for them to make their lives easier. I think Google IO, I'm hoping they talk about more about how they're going to get Gemini in front of more people. So I think. I don't know. There was something I read recently that they, they're likely going to put it, like, in some elements of search or YouTube. I already, we use, we use Google sort of desktop and like Gmail and so forth. To me, like, that will be really interesting from a search perspective because it's going to be like, there's going to be a ton of learning on, you know, what are people talking about? And this kind of comes to another shift that I think SEOs, we're kind of in the middle of right now where we're still thinking about keywords, right? Remember when I talked about, like, the consumer isn't searching for keywords anymore, right? So, like, we're fooling ourselves if we think how we do audits on keywords and how we're planning content. Like, we need. That needs to change drastically right now. So entities are kind of like, what are the topics? So we need to kind of think broader than keywords. Like, and we need to think like topic coverage on our website. So this is that. Remember I said there were three values, SEO, content and AI. Like, this is that content piece that, like, I think is going to get very, very disrupted. Now I would love Google to be like, hey, we're going to, you know, stop giving you query data and we're going to give you topic data. Like, that would be amazing. That would, like, make my day out of Google. I o. I don't think they're ready for that yet. I think as they put Gemini in front of more people, they'll get a better sense of that. But like, we as SEOs, like, and this is again, schema apps, like, kind of working on this, right? Like, how do we tell you what entities and topics are on your site? And how do we kind of give you an idea and a framework? But I think like, that gap of like, getting us from keywords to entities, like, that's, I would think, like the urgent need for us to do in the SEO community now. But what will Google tell us to do? I don't know. It'd be interesting. It'd be really fun if they came out with requirements like they've done with rich results to Say like, this is what needs to be in your data in order to be almost like a technical SEO requirements, but it's like gen AI, like Gemini requirements. I actually think that would be very helpful. And I think it's going to tie to structured data and it's going to tie to sort of like how you make things super easy to crawl or super easy to actually give your data. Right. So like, if they make MCP connectors, at what point are we just going to be like, great, give us the feed. I think it happened too close to IO. Like they have to plan way far ahead head. So maybe that's next year's IO. But that would, I think that would be interesting if they, if they're thinking like that to say, like, how do we produce data? The piece that I think is interesting is how are they going to replace that old handshake we had where we give you content and you give us clicks? I think like, you know, are they going to make a new contract with all of us with regards to how that's going to evolve? That would be, you know, and how are they going to frame it? They're the big dog, so they're the ones who are going to call the shots. It's not going to be like, we have a say in this, but, you know, media has actually had. Funny story, Google reached out to me directly through our support channel. Love this.

Jordan Cooney (20:08)

Okay.

Martha Van Berkel (20:08)

And was like, could you guys support this if we add it to schema.org, so this was like 20, I don't know, 2020, probably 2019, 2020. And it was around a licensing property for media. So the media teams were looking for a licensing piece. So it's already built into schema.org since 2020. I know that's my pontification on what we might hear and the direction they might need to go.

Jordan Cooney (20:35)

I want to add one more caveat to this topic, which is just this past month in the antitrust lawsuit with Google, one of their engineers was giving testimony and basically stipulated that a overwhelming majority of Google's rankings are handcrafted. And that was the exact word used, handcrafted. Right. And he talked about DeepMind and certain other aspects that are automated and are part of the core algorithm system, but that there's still a significant amount of Google that is kind of handcrafted or yielded by engineers and that engineers can influence the changing in rankings within these parts of the environment. So the question I have for you is this is as recent as just April 2025. How are you feeling about the components of search results being handcrafted and how should we as operators think about implementation of what we put out to Google, whether it be in the traditional sitemap webpage strategy or how we're talking about now the future of concepts like frameworks like MCP and protocols that feed data into LLMs and AI. How do we prioritize?

Martha Van Berkel (21:51)

All right, this is like for sure just my opinion but well I think about how Google makes money. So if we think of there's like my brain goes to like there's people deciding, right? What gets shown, right? That's what that the data point you just shared. And if that's the case, my question is why? Like why are they doing that? And is it because they still want to be thoughtful about how that maps with ads, right? Like are they still being. And so is there a stipulation of how who's spending on ads that's going into that? That's sort of like part of where my brain goes. Now you could automate that because you know how much ad spend people have and so forth. The other piece is like what's driving people to stay in search? Like we've seen, I would say lots of evidence in Google AIO that like they want it to be self reflective. That's my nice way of putting it, right? That we're seeing like such a huge number. I don't know. There was a data point shared. I think it was like over 50% of like links in AIOS in this latest version, kind of April timeframe, May timeframe are pointing back to Google. And so part of me is like, well what's the content then that enables that is sort of like where I'm like, are they trying to prioritize some of that so that they're getting sort of more kind of visibility in the whole question just starting with like why would they kind of be manually ranking that piece? So part of me is like how do we be strategic about. I don't know, this is like anti sort of how business is optimized, right? So I think of like business, right? Like I'm holding these two truths in my mind, mind at the same time.

Jordan Cooney (23:20)

But you have to realize for the better part of 20 years the mission behind Google was more is better, right? Like the mission has always been like feed, feed, feed Google your content, organize it in these things called sitemaps. There are a couple other ways that they organize this too. But like generally speaking it was feed more. But the reality is that we're embarking on a Whole new frontier where Feed More probably isn't going to work effectively if we have a multi channel environment for discovery, if consumer discovery is going to change and it's not just Google and even within Google it's segmented to multiple experiences. Feed More isn't going to work.

Martha Van Berkel (24:00)

No, I think it's like feed quality and feed uniqueness I think is going to be like the key piece and I would say like the organization or the content that I'm hearing perform like in our enterprise is like not generalized, not kind of, you know, even gen AI generated that's like not unique. It's like that unique perspective like all this EA T stuff we've been talking about now for years, right. That with sort of like context of the business and like elements of the business. And so this is again where like what are the topics that you like you really need to be an authority on in order to stand out. And it's not quantity, it's not like I can publish this many blogs. It's like really deep rich content that's unique, that's not out there in the world in any other way. And I think that's the part that's hard. I think the other piece that's really interesting is I think of my brain just went to social and how LinkedIn content is now showing up in search and how individuals versus company kind of articles that are coming up. And I think that's sort of when we think of strategy like thinking of where are all these elements, it's unique content and is it again hosted on the website or is it hosted on other pieces that have their own kind of connected data and kind of elements and linking. Right. Like when you tag people in it, when you refer to other articles, you know, so to me that's like where we're seeing like those kind of goes to the very first point of like how are we thinking of our data and where it's all connected and how it's coming out but unique deep content by experts like that's what I'm seeing perform and that's I think what we need to insert into the machine and then make sure again that data layer like that piece of it's like are they articulating and consuming it easily?

Jordan Cooney (25:44)

All right, my last kind of deep dive in terms of how we're going through these questions is really around building an actionable framework. How can our SEOs, the web community, the web development community really take clear steps to provide clear actionable data to both search engines and generative AI LLM models to perform better in discovery.

Martha Van Berkel (26:10)

So I'll talk about structured data because that's, like, what I see really works. So one is like, structured data starts with structure on the website. So I think let's talk about web developers and people thinking about. So machines love structure. And so if you think of, like a templated site, like those of you that have use AEM and pull random components in all the time and just let marketers go wild, that's hard for the machines to understand. So, you know, having some structure as they crawl your site and understand it. And that goes at the very beginning of, like, web development, right? Your content then is like the piece that goes in that. So, like, think about, you know, really clear content, this whole, like, lots of words to just, like, marketing isms that don't mean anything to anyone and that we all skim through because we're trying to get to, like, what do you actually sell? Or, you know, what is this doctor? Like, what services do they actually pro? What's their expertise? What can they solve? You know, like, we need to get to that really clear, articulate, you know, content that then is in that structured way, because then the machines can get what they need and then they understand it. And then the last piece then is like, that structured data. So how are you thinking about not just optimizing a page, but connecting the dots across your site? And this is where, like, the IT people I talk to, where they're like, well, I can add schema markup on this product page. And I was like, great. But, like, how are you connecting the dots to that rich content we just talked about that's super deep and super specific. Like, you need that content to, you know, be talking about that product and connecting those dots, you know, very clearly. Is it mentioning it? Is it about that product? Is it recommending that product? You know, as an example?

Jordan Cooney (27:44)

One of the things that I really feel is critical, especially for websites that have been around a while, is the cleanup journey. Is the, like, the removal of, like, things that are not useful anymore still in play when it comes to data?

Martha Van Berkel (28:00)

I think so. I think so. And I'm seeing our clients do this, right? Where it's like, let's sit back and say, like, healthcare is a great example because they used to all have the same content. Like, what is melanoma? Like, what is heart disease? Right? So all that's gone, right? And again, like, now they're thinking about their websites and, like, you know, what are the, you know, specific service lines? Like, what is the medical things that we solve and, like, how do we articulate, you know, not just who this physician is, but, like, what surgeries they do and, like, what specialties do they have, like, work on in context of, like, what hospital they practice at? So I am seeing people think about, like, what is core to answering that consumer question, and where does that data live? And we also think about, like, if there's a. I call it a thing, but, like a topic or, you know, an entity, right? Like, if there's a thing that you want to be an expert and known for, there should be an entity home. Like, what is the one webpage that you want people to go to that talk about that thing? And I would say, like, that's when we do a lens of, like, cleanup. It's like, often it's in, like, a million places. And so, again, if you're a robot and you're like, which is the page that is most authoritatively talking about this topic from this brand that I like, because I think they're connected, like, that should be an entity home. There should be, like, one page that talks about that. And so this is sort of like, you know, I joked, I was like, category pages might come back, which is, like, an interesting product kind of piece. But, like, you know, there should be a page that talks about, like, we specialize in, I don't know, I always use, like, I love, like, organic cotton dresses, right? Like, I want. Like, I don't want to. I want to know that from that brand. And it's, like, handmade in Canada, and I can, like, get it and it ship, you know, like all those things. So, yeah, how do we really think differently about. Again, we still have to think about the consumer, but I think we need to think about authoritative kind of content, articulating what we're experts in, and not general stuff like women's clothing or really, really specific. And this, I think, is going to make. Actually, this is a new thought I've just never had, but I think it's going to force businesses to get really clear on what they're good at and what they do and what they don't do.

Jordan Cooney (30:18)

Correct.

Martha Van Berkel (30:18)

Which is actually really exciting. So Canada's oldest department store, okay, this is like, you know, Hudson's Bay. Like, it's been around since, like, we, you know, like, the founding of Canada is going out of business. It's going out of business, but they don't actually.

Jordan Cooney (30:34)

Don't worry.

Martha Van Berkel (30:35)

We're.

Jordan Cooney (30:35)

We're following suit. I think JCPenney's is just trying to come back to life. But you know, right.

Martha Van Berkel (30:40)

Like I like, thank you for giving me a great us example. But like, but if you think about it like the department store is this generic place that we go to buy underwear. Right. Like, let's be honest.

Jordan Cooney (30:50)

Yeah, yeah.

Martha Van Berkel (30:50)

And, and now we can go to specialty stores that build like make really great underwear that's like tailored for like whatever we're looking for.

Jordan Cooney (30:56)

Exactly.

Martha Van Berkel (30:57)

And so I think like that's a great example of how the consumer behavior is changing. We can find super niche very specific things at the, at our using voice while driving the car.

Jordan Cooney (31:08)

Yes.

Martha Van Berkel (31:09)

Right. And this kind of takes us really back to the very first thing I said. Right. So like, like if we need to then figure out how we're going to show up in that voice search, as I'm driving my kids to hockey practice or ring at practice, we better make sure our content is talking about those specifics so that they find us and that they see like our brand as that authority. So I think this all then comes to like web, like this is why I think content and SEO is so tied. And then it will play a role in just like how are you enabling that data to get out to those people? And, and again, I think structured data is what we're seeing the two big players, Microsoft and Google, lean towards.

Jordan Cooney (31:46)

Yeah. And I just want to add this, we don't have to debate it, but I think I want to add this, which is like today where we are within this generative AI and these response based search experiences is that categorically or information wise are really good at providing us some direction and we're only going to get more and more specific. We're going to get down to the individual person entity. So think of the doctor and what that doctor specializes at this particular hospital and his years or her years of experience and capabilities. Like we're going to get to that level of granularity and responses. We're not there yet, at least not effectively. And so I think that's why this whole topic of data is such a critical one for us to get good at now so we can answer that in the future.

Martha Van Berkel (32:33)

Yeah. And you can do that with structured data. Like that's when you do it semantically and connect the dots. I think the other piece is it's going to know us personally too. Right. Because we can give it context. So imagine a personalized search journey on whatever platform I want. And then at what point are we going back to APIs? Like, I'm not even buying it. Like ChatGPT just announced their product search. Google's been doing a ton of work right, where you can, like, literally, like you're just pushing the buy button. You're just converting on the website. So conversion optimization, I think, is maybe the new SEO.

Jordan Cooney (33:04)

Absolutely. And that wraps up this episode of the Voices of Search podcast. A big thank you to Martha Van Berkel from Schema App for joining us. If you'd like to connect with Martha, you can find a link to her LinkedIn profile in our show notes or on thevoicesofsearch.com for more information about schema app, visit schemaapp.com if you haven't subscribed yet and want a daily stream of SEO and content marketing knowledge in your podcast feed, hit the subscribe button in your podcast app or on YouTube and we'll bring you back to our feed. Okay, that's all for today and until next time. Remember, the answers are always in the data.