* The following transcript is computer generated and may contain errors.

00:00:01:22 - 00:00:35:20
Speaker 1
Hello, everyone. Stuart. Gandalf here with my colleague Brandon. And today we are going to talk about how to adapt your SEO in the age of Google AI overviews and ChatGPT. This is a, webinar that we are rerecording. We recently did this in conjunction with our friends over at e-Health Care Strategy and Trends, also published by Plain English, but we thought we've had so many requests to do a, tighter version of the webinar we're recording it, with the same content, but just, the two of us leading today.

00:00:35:22 - 00:00:44:02
Speaker 1
So, Brandon, I'm excited to work with you today. Brandon Scola. Brandon, you want to do this yourself? Just for a moment, and we'll get started.

00:00:44:04 - 00:00:58:09
Speaker 2
I'm Brandon Skogland, senior director of digital services here. So I oversee our SEO practice, pay media practice in our development practice here at Healthcare Success. Ex Overstock, ex eBay, ex the search agency. So.

00:00:58:11 - 00:01:23:09
Speaker 1
So Brandon is, one of my favorite people anywhere, not just within the company. And, we were successfully recruited, after years of following your brand. So, I'm glad you're on board with our team. So let's just go ahead and get started. If we can go to the next slide. So one of the things that it's really important to note is that the world has changed a lot.

00:01:23:09 - 00:01:50:22
Speaker 1
And, as we talk to clients, as we talk to prospects, as we talk to other vendors we have partnerships with, everybody has noticed a disturbance in the force, so to speak, and the, right around March is when this became a long standing trend, really accelerated and became sort of omnipresent. So the last slide there just was that at zero, click search is not new.

00:01:50:24 - 00:02:13:22
Speaker 1
And we'll talk about that in a moment because we've been doing zero click for quite a while or on that trend. But starting back in about March this year, health care websites in particular were hit with a very big change. And essentially what happened was Google changed its algorithm to begin favoring AI overviews in the search results in a much bigger way.

00:02:14:03 - 00:02:39:04
Speaker 1
And again, this has been happening in smaller ways for quite a while. But they, made much bigger change there. Now in some industries, you know, this impacted 10% of traffic, 20% of traffic. But now in health care, huge numbers of searches are being directed towards AI overviews. And so what we're seeing is that is a trend which is impacting search traffic.

00:02:39:06 - 00:03:05:12
Speaker 1
And click through to websites in a big way. And at the same time, a confounding variable is, the increasing use of tools like ChatGPT, perplexity, also AI mode on Google, which is a related to a little bit different. So all of these things are beginning to change, search behavior. And a lot of people are mistakenly attributing this big loss only to ChatGPT.

00:03:05:18 - 00:03:41:06
Speaker 1
ChatGPT is a relatively small market share, but its impact is increasing at an exponential rate. And those people that are searching with ChatGPT tend to be much more qualified. And, the, this is being called now, we've source over here the a Ahrefs decoupling. Oh, the great decoupling. What we're seeing is that impressions are no longer necessarily resulting in clicks and so and here we have a little quote from Google saying that's the CEO of Google saying that search will continually change profoundly this year.

00:03:41:08 - 00:04:04:02
Speaker 1
And I think we've seen that, very, very big so far already. On the next slide, Brendan, so one of the things that we talked about, it's just a moment ago was that zero click is not new. It's been around for a long time. Brandon, when will this first start happening again? What year was that? About 2010.

00:04:04:04 - 00:04:11:24
Speaker 2
Well, it got more advanced. I would say probably around 2013, 2014 was when it really started rearing its ugly head.

00:04:12:01 - 00:04:34:23
Speaker 1
Yeah. So you can see that people also ask, you're probably familiar with this, map pack or Google profiles became a more important, you can see discussions and forums featured snippets. Google is really in the now today of course, the AI overviews. And what's happening today is Google has always wanted to keep you with in its ecosystem.

00:04:34:23 - 00:04:56:05
Speaker 1
Right. So that's why more and more, rather than taking you to a website elsewhere that they can answer your question within the Google Serp feature search engine results page, they would. And so today, adding the AI overviews is just another part of that, although it is accelerating because people are drilling down now and being satisfied much further.

00:04:56:07 - 00:05:14:07
Speaker 1
Another thing is to think about as this becomes the case and brand is going to talk about AI overviews and also AI mode, which is a little different. As I mentioned a minute ago. Google my opinion on this brand. And and friends, is Google's brought to this party sort of unwillingly. They've been working on AI for a long time.

00:05:14:07 - 00:05:34:20
Speaker 1
I remember, going to Google headquarters back around 2015. I've been doing this for a while, all the way since 2006. But 20 1516, they were showing machine learning and AI, and they were working on this a long time ago versions of this, but they had many billions of dollars set up on things just the way they were.

00:05:34:22 - 00:05:55:13
Speaker 1
But with the advent of ChatGPT, everything accelerated because Google, of course, doesn't want to be, come disambiguate it. They want to continue to be relevant. So they've made these changes. So you're just seeing the landscape change really fast. So again, the point I'm making here is that zero click search is not new but it's just an acceleration.

00:05:55:13 - 00:06:06:16
Speaker 1
It's happening now. And next slide Brennan and Brian is going to talk. Now I don't I guess why don't you pick this up Brennan I had a lot of color commentary and questions along the way.

00:06:06:18 - 00:06:25:05
Speaker 2
Sure. Absolutely. So let's get right to the heart of the matter. You guys can copy the tax office next slide. You can try it yourselves. Just head on over to AI mode and you'll will see that this is not your grandpa's search. So this is your query. And this is actually what's happening now, across all sorts of these new AI services.

00:06:25:05 - 00:06:47:18
Speaker 2
People are writing these long prompts, and those are those are the keywords that we grew up with. Right. And these things can take action. In this case, we're telling it to go out and search for, like the five most talked about stories, bring that, generate some, generate some LinkedIn post ideas for particular audiences. This is not search in the standard fashion.

00:06:47:18 - 00:07:11:13
Speaker 2
And this is the new world that we're getting used to, that it's not a world of, necessarily like ten Blue Links or even rankings. It's it's of of influence and visibility. And so it's changing the way that we need to think about what the role is of, of us CEOs. And some would say we're becoming GEOs or generative engine optimization specialists.

00:07:11:15 - 00:07:29:24
Speaker 2
It's, it's sort of a mixture of both. Because this is a layer on top of search. So I look at it as an, an extension of it. So this is what we're coming to expect, and this is what was really to said is going to eventually become the default for Google search. I think we're going to see that relatively soon.

00:07:30:01 - 00:07:51:03
Speaker 2
And so it's a world that all of us need to adjust to. So the biggest lesson you're going to learn today is your healthcare brand as a data set. If you put garbage in, you're going to get garbage right back out the other side. So you need to be taking care of your fundamentals. You need to be looking at things the way an AI agent may look at them, as well as a human being.

00:07:51:05 - 00:08:18:17
Speaker 2
So we're going to do a deep dive into the shift from traditional AI search. We're going to talk about the zero click data challenge. And what that means for organic attribution. It is a case study that we kind of previously put together that will walk a little bit through, three different hospitals providing and services. We'll look at visibility and large language models such as Gemini, which is what drives AI overviews and and is behind, AI mode, GPT perplexity, Claud.

00:08:18:20 - 00:08:46:20
Speaker 2
And then what you can do about it. So the biggest question we get asked in SEO is what happened to my traffic? And, you know, the answer is is is not new. So in the great Decoupling, and there's been some, some, some new news about this, over the past two weeks since we, we produced this, what we were seeing as, as AI overviews rolled out over time was impressions continuing to climb and clicks kind of going down.

00:08:46:22 - 00:09:08:10
Speaker 2
And the reason the clicks will be going down is because more and more people are staying within this new surface, of search like AI overviews. And they're getting the majority of their questions asked there. So the interesting piece is they're they're typically kind of squashing that search funnel where you would go out and do maybe 40 different searches before kind of coming to conclusions about which path you better take.

00:09:08:12 - 00:09:37:14
Speaker 2
In this case, you know, we're getting pushed further down the funnel, before someone takes an action. The impressions are rising because as more and more, AI bots are kind of coming in, essentially those are kicking off additional searches. So the example that we just came from AI mode actually probably ran about upwards of like a hundred different background searches before coming back with, that, that nicely formatted answer for us.

00:09:37:16 - 00:10:00:21
Speaker 2
So it's search is still under the hood. It's just changing the way that the patient journey works. And so what we're seeing is the old metric of traffic is kind of almost becoming more of a vanity metric. And, and visibility, is, is really the key. It's, it's about being there and being part of the influence of that answer the brand.

00:10:00:22 - 00:10:10:10
Speaker 1
I was going to bring up one comment there. We talk about Be Everywhere. Oh, what does that mean now in this context, this idea of increased visibility.

00:10:10:12 - 00:10:34:21
Speaker 2
So in this case, being everywhere, it means being in all the places that these AI models are either training their data or looking to ground their data out. So if you're asking a question that the large language model doesn't have in its training data set, it then triggers it to go out and do this grounding exercise where it looks for sources that might give it the right information to to produce the answer.

00:10:34:23 - 00:10:56:19
Speaker 2
And so that happens in real time. Some of those are cached. But that's that's essentially part of it. So it's not so much just about the, the onsite piece, which is what a lot of us work on. It's also thinking about what are the important sources out there, the, the large language models are paying attention to as sources of their information, like common crawl.

00:10:56:21 - 00:11:20:11
Speaker 2
We know that there's influence from Reddit and user forums and things like that. There may be specialty niche directories or forums that that need to be mentioned in. So in the offsite world, it's really more about the brand citation and the brand citation within a specific context that can get chopped up and spit back out within these, it also means especially for Google.

00:11:20:13 - 00:11:46:24
Speaker 2
You know, Google's models are multimodal, meaning they can process audio, video, text all in real time. And so it actually provides more, more types of media for you to be producing to get into these overviews. Excellent. So how we got here, search has always been changing. You know, pretty much it was a little static right up until around 2010.

00:11:47:01 - 00:12:15:07
Speaker 2
Which is when, right around when Universal Search was was introduced, that kind of blends multiple content verticals. And that's where you started getting images, video, map tags and so on and so forth. What really changed was in 2012, they released the Knowledge Graph. And this is where it could go, went out. They purchased a company, called freebase, which was starting to markup all the different kinds of entities around, around the web and around the world.

00:12:15:07 - 00:12:36:13
Speaker 2
So being able to determine people, places, things, all of that got mapped up. And as of 2012, they basically had more than, like 500 million objects and, more than 3.5 billion facts. And that's what drives some of that right hand rule behavior. Like, let's say you search for Nike and you get all the facts about Nike, including including like the Nasdaq and all that stuff.

00:12:36:15 - 00:13:02:19
Speaker 2
So, built on top of that is then what became what's known as the topic layer. And this is where they started really deepening that. And so if you would search for, you know, like a famous football player at that point, you weren't just getting a standard set of template links, you were also getting tabs where you get, statistics and history and, you know, games that and all of that.

00:13:02:21 - 00:13:27:24
Speaker 2
And those are things that that the, you know, the third party tools weren't even able to track, was just kind of like, boom, you know, have like actually hundreds of results and, and no one knows what the rankings are or how they're put together. Google that introduced the shopping graph. And then in 2023, in the background, if you were a member of Google Labs, we began to start seeing these generative experiments.

00:13:28:01 - 00:13:47:24
Speaker 2
Kind of hot on the heels of the release of, about gpt2, I think. So last year, in May of 2024, AI overviews rolled out to the world. And it was, you know, hilarity ensues. They were telling us to do things like put glue on our pizza for Memorial Day recipes. And, and it was a bit of a funny launch.

00:13:48:01 - 00:14:14:24
Speaker 2
But they really dug their heels in and started making something way more interesting. As of May this past year, they launched AI mode, that had some agency abilities within it. And there's some pretty interesting, prompts that they show you that you could basically, within your search, have have Google. I may go off and do you know, a bunch of actions like search for an apartment and start booking appointments for you to go view them?

00:14:15:01 - 00:14:40:05
Speaker 2
So all of that is driven from this transition from what I call traditional search into the era of semantic search, where we're talking about things. And so I think this plays a really big, important role in how we look, at this transition into, AI. Now, AI and search are complementary. They're not replacements. People are going, oh, I see instead, I was going to replace everything.

00:14:40:06 - 00:15:00:17
Speaker 2
Actually, it's not, because of that, but that part of grounding. So what they find is, is somebody is using GPT and they're asking these complex questions, or GPT is going out to a search API to figure out all the stuff it doesn't have in its training model. And the same thing is true with perplexity and some of these others.

00:15:00:19 - 00:15:22:13
Speaker 2
So part of our goal is to figure out, well which things are going, are going to give us a grounded response versus which things are going to, sorry, a training response versus a grounding response. But in most cases, this is part of what's inflating all of the impressions. And Google recently tried to correct this by, removing, what's what's called the num equals 100 parameter.

00:15:22:15 - 00:15:41:23
Speaker 2
That's been in use for decades by also all sorts of scrapers and bots and, third party rank tracking tools and basically put the kibosh on it. So over the past two weeks, there's a lot of people in tons of verticals, across the web and across the world have seen their impressions drop, pretty drastically.

00:15:41:23 - 00:16:05:23
Speaker 2
And it's basically in response to this. But the interesting piece about that is it's amazing how much those bots were actually inflating the impressions. In some cases, upwards of like 50 to 70% is what we've seen. What comes out of that, though, is you're actually getting cleaner data about real human impressions. And so that's a key takeaway.

00:16:05:23 - 00:16:29:24
Speaker 2
And we'll have more on that in a in a follow up blog post. So we know that that AI depends on search. And there's been lots of chatter about this. Dan Petrovich, had a really nice piece here where he showed, that OpenAI's model is trained to be intelligent, not knowledgeable. So they didn't add a ton into this.

00:16:29:24 - 00:16:48:18
Speaker 2
This latest set of training data. They're like, yeah, you know it. Well, let's search and go around and take care of the rest of it. So I in a lot of ways is just kind of a new wrapper or a new surface for search experiences that accelerates the the patient and the consumer experience. But tracking this just got harder.

00:16:48:20 - 00:17:11:11
Speaker 2
Because if we were to look at the referrer strings coming into your Google Analytics, for instance, you might say, see like 102 views for GPT. Yeah, the engagements up, which is great. There's more more events happening. But this is at a scale the that's pretty, it's still pretty small in terms of what we're seeing as traffic coming into the site.

00:17:11:13 - 00:17:31:03
Speaker 2
There's a reason for that, and that's because it's coming in and getting filtered out as as an AI bot already from the get go. So if you look at Google Bot and we look at your log files, Google bot coming in maybe about just under 3000 times a month, and we look at the AI bots, it's like ten x, which is really, really crazy.

00:17:31:03 - 00:17:52:05
Speaker 2
And so I think that's why there are companies like Cloudflare trying to take action against this. And, and Google kind of fighting back. What's parameter is, is trying to get control over, over the web again, control over real traffic. Well, when we actually go in and look at your log files and this is something all of you are going to have to get used to doing from here on in.

00:17:52:07 - 00:18:11:21
Speaker 2
Is that same site that that showed us 102 views from GPT, says nothing about GPT users. So on the bottom of the slide, you can see I bought visits. So this is OpenAI search for GPT bot perplexity and so on. Those are the bot bots where they're going out, you know, and, and trying to gather information to train on.

00:18:11:23 - 00:18:34:06
Speaker 2
Whereas GPT two user visits is an actually user sitting inside of GPT, you typing away, having a chat and a conversation. GPT doesn't know something, so it goes out to the search API, grabs those results, tries to find ones that it can produce an answer on chunks. All that information up ranks it spits it back out. You in some sort of coherent or semi coherent form sometimes.

00:18:34:08 - 00:18:54:14
Speaker 2
And so the users never actually got to your site. That's like a true visit as a true hit, but they're still getting your information and they're still getting the interaction with you. So this is what I refer to as dark traffic. It's something we're all going to have to be watching because basically you have no idea of of that the information you spent so hard.

00:18:54:14 - 00:19:18:19
Speaker 2
Right, and all this great content on your site was even getting viewed as the only place you're going to find it, as in your log files. So let's get into a case example. There's some common patterns. If you're looking, you know, tools like ACM rust, for instance. So this is a hospital site, and, we can see that there's, there's desktop versus mobile.

00:19:18:24 - 00:19:36:11
Speaker 2
Mobile is usually way less keywords in terms of what we're going to see. And what we do know typically in the healthcare space is the majority of your conversions come from mobile. So people do most of research on desktop. And then when they're really in a crisis mode, that's when they like, you know, kind of dive into the mobile.

00:19:36:11 - 00:19:55:18
Speaker 2
They reach for their phone to try to solve that problem. So mobile tend to have a higher sense of overall urgency. But if you look at the stat bars in the chart, where most of the zero click is happening, is is right there on mobile. And so that's an important point to keep in mind. So that's where the majority of the growth has come.

00:19:55:18 - 00:20:25:02
Speaker 2
So it's not true growth but it's more visibility in the long run. And so what this looks like, these are the search features for this particular brand. You can see people also asking about 7000, things to know, which is sort of like an evolution of the Knowledge Graph. I overviews at about 2000 and you can see some adjustments, coming out of the big June to July, core algorithm update and some readjustment, from the August to September spam update.

00:20:25:04 - 00:20:44:06
Speaker 2
Lots of shifting around of what's available there. Something to note as well is, eye overviews are also not just showing up at the top of the screen. They're actually getting tucked into people also ask and they're actually like answering the questions in there. So it's a little bit of sneakiness and recursion. And what's, what's appearing there.

00:20:44:08 - 00:21:04:01
Speaker 2
So how do we move from threat to revenue? It's going to take a much more technical approach. We have to understand how I search differs from traditional search. And to do that, we need to learn to think like AI, which is going to be fun. Most of us. So how Google indexes content currently is, it's out there with its bots.

00:21:04:01 - 00:21:25:22
Speaker 2
It is discovering lists of URLs on the web. It throws those in a queue that gets put into the crawler, the crawler goes out, it tries to render and process the pages. If you're pages are, are using a ton of JavaScript, for instance. It then has to go into the separate queue to then render the fully rendered page with the JavaScript, and then it gets passed back into processing.

00:21:26:03 - 00:21:55:00
Speaker 2
And then from there, if you're deemed relevant enough for certain queries, then you get thrown into the index for the ranking part of the procedure. So we, use a tool here, for crawling websites. Kind of mimicking Googlebot along the way. And things that we look for are, are there common crawl traps or are there traps or things that could be keeping you from getting pages on your site index that you really want in there?

00:21:55:02 - 00:22:19:03
Speaker 2
In this case, things that I'm looking at is this site has about 36% of their their overall URLs closed for indexing or somehow blocked, sometimes through no fault of their own. They just didn't know that this was happening. Another thing I'm looking at is the page load time. Really, with, today's, eager, phone users, you really want your page loading in under a second?

00:22:19:05 - 00:22:40:02
Speaker 2
That can also slow down things like the rendering part of the process. And so that can be another trap. A third is clarity. And so we do a lot of work here with, structured data, schema markup. This is these are the pieces that feed the things to the knowledge graph. And so you want to make sure that this is organized in the proper way.

00:22:40:02 - 00:23:03:07
Speaker 2
And whereas this site has basic organization and web page markup, they have nothing about what the page is about, which is the precise reason to use this kind of markup. There's nothing about medical facilities, hours of operation, phone numbers to call, specialties offer. And those are the things that provide the semantic layer out into the search engines for processing.

00:23:03:09 - 00:23:20:17
Speaker 2
Carlo's site number two. They're actually doing block blocking pretty actively. And a lot of sites are doing this now. It does not bode well for your future if you try to it into large language models and you're not even getting into search. This is a really dangerous procedure. I understand why certain people are doing it.

00:23:20:19 - 00:23:48:09
Speaker 2
But in reality, trying to play whac-a-mole is then just going to be an infinite problem for you. There are ways of controlling them. The large language model bots are not quite adhering to what we know as the normal robots.txt protocols. There are some folks experimenting with an LMS dot txt, but, few of us have actually seen it fully working in the live because it's a it's a proposal.

00:23:48:09 - 00:24:12:15
Speaker 2
It's not an actual specification used by the W3C. So we're really in like the wild, wild West, mode of these, these new search services, site number three better doing much better on the index ability status with about 4000 pages open for indexing, but only 281 pages with structured data. So maybe they're probably in flight with with implementing it.

00:24:12:15 - 00:24:28:12
Speaker 2
Who knows. But you can see at least they're they're tapping into things like medical clinic, physicians, which is in a much better way. These are things that are going to really help you, when people are searching for particular kinds of specialists. So.

00:24:28:14 - 00:24:35:21
Speaker 1
Brian, can you explain? Just take a second there. Okay? You're talking about structured data here. Yeah. This is a good place to go and explain what that means.

00:24:35:23 - 00:24:58:00
Speaker 2
Sure. So, back around the same time that Google bought freebase, which is kind of like the large database of all of these things that are marked up. They partnered with, all of the other search engines. So like, Yahoo, Bing, Yandex, etc. and funny, like this, this structured markup actually originated in the healthcare space.

00:24:58:02 - 00:25:20:22
Speaker 2
So it just it baffles me that they're so bad at using it. This is how you get structured patient records, people, is via things like structured data, ontologies and taxonomies. So there are hundreds of these floating out on the web. And schema was a way of trying to make it simpler for people to want to adopt those, so are ways of extending schema and, or like really, really specialty spaces.

00:25:20:24 - 00:25:44:01
Speaker 2
But this would be enough for just normal human folks to help kickstart this project. So, you can browse the full hierarchy. So medical, you can see health and medical types. They go under a special type called medical entity rather than the standard, you know, organization markup or something that you might use for like a nonprofit or a local business or whatnot.

00:25:44:03 - 00:26:07:20
Speaker 2
And so within each of these, there are rules with each of these, these kinds of objects. So if you take like a simple one like, I don't know, like a genealogy, there are rules baked into these ontologies such that, like, a daughter can't be the parent of herself or something silly like that. Right. So it provides some logic, about the context of what these things are.

00:26:07:22 - 00:26:33:10
Speaker 2
But, you see, the thing that gets marked up, but then the things have properties or relationships to other things, and that's the more important part that almost everyone misses here. You have to think in sentences subject, predicate, object. So location has specialty, for instance. And so this is something I see a lot of organizations, particularly in the health care space, really, really struggle with.

00:26:33:12 - 00:27:00:10
Speaker 2
So how would you do this? Well, what you want to do is sit down and and grab a map and start mapping out all the different kinds of entities that you could possibly have on your site. So you have your brand. The brand has a medical page. It's about a medical condition that has a diagnosis, that, you know, to treat possible symptoms and that particular service is only really, offered a location, versus a different service located, location B, for instance.

00:27:00:10 - 00:27:31:15
Speaker 2
And then you have this relationship with health insurance plans and particular doctors and types of therapy. So you really want to be marking up these pieces and then using that as a way to structure your site to build as much clarity about your offerings as possible. So when I looked at these three sites, the site number one, pretty disconnected in terms of its graph, you know, wasn't really showing up as a clear hospital entity, really a tiny version of the graph was broken.

00:27:31:15 - 00:27:51:12
Speaker 2
Site number two, the same thing that was blocking the bots. Doesn't have any schema at all. Site number three. It's a bit fragmented, but at least they kind of have departments marked a lot, not necessarily granular down to the size of like individual services or treatments. But I would suspect that's that's likely where they're gone.

00:27:51:14 - 00:28:19:08
Speaker 2
So how does this look when you, when you're doing this. Right. Well, essentially you get this kind of like pretty gigantic graph, all of your things marked up and how they're connected to each other. And, you know, this is what you're feeding into, these, these, these systems to process the knowledge graph. So this is part of the path of, of really giving, the search engines an easier way of understanding what your entire site is about.

00:28:19:10 - 00:28:41:04
Speaker 2
So here's an example of how you could do this. You would, you know, build out your entity mapping, and then you would tie everything into like a real world entity. And you have your proposed schema types, what the descriptions are for those schema types. And then you might provide details for like the treatment name, the medical procedure properties, for instance.

00:28:41:06 - 00:29:10:00
Speaker 2
And you want to feed all of that into this particular format called Json-ld. That's an example to the right. And so from there, you would continue just marking up all the things on your site and kind of building out your own kind of personal dictionary of everything to really help the large language models and bots and, and processing systems that these large search companies, to give them as much context and clarity as possible in a machine readable format.

00:29:10:02 - 00:29:35:15
Speaker 2
So once you're there, it's really about how people understand your content and then presents results. So, a while back, Google introduced these models like Bert and that understand what questions and context, they understand things like a topical level, they know when words don't belong inside of that topic, just due to the other words around it.

00:29:35:17 - 00:29:58:17
Speaker 2
So they're not just matching, you know, this, this exact match world where you where we used to use one word searches like tonsillectomy, or anti doctor near me. They're surfacing real patient questions, behind those searches. And so as you're thinking through the content and the flow of that content, think through the actual patient questions person first, not keyword first.

00:29:58:19 - 00:30:18:23
Speaker 2
And if you approach the content that way, you won't sound like a 1970s television commercial where you're just repeating the brand name, what, 15 different times? You really want to answer their questions, and build trust with them along the way. So we looked at one of these sites, this is the second site, the one that was blocking the bots.

00:30:19:00 - 00:30:58:13
Speaker 2
They actually had a really good, way of handling this with, with their content. They balanced out the technical and the layman speak. So on the left, what we're looking at is, is some top level clustering, of their anti web page. And it's really looking at pediatric auto er and geology and to the right, if we kind of skim off that top layer and kind of dig in to see what's underneath of it, you know, at least they're kind of looking at things like treatments, they're talking specifically to common ailments like hearing loss, sinus disease, some more complex, conditions such as airway problems or swallowing disorders, for instance.

00:30:58:15 - 00:31:20:10
Speaker 2
And so these are the things you're going to want to pay attention to, making sure that their services are marked up in a way that include management of, noisy breathing in infants or air tube surgery. These are a little more common to the audience. And, this is really what's going to to make you tune your right audiences.

00:31:20:12 - 00:31:40:19
Speaker 2
So what Google and these other engines are doing, when they get to a page on the internet, is something called passage level indexing. This is from a particular patent, on, on a, one of these kinds of systems, that Google put out called Ginger. But it's showing you how we get to I own views from the content that's written on the page.

00:31:40:19 - 00:32:02:24
Speaker 2
And so even in Chrome, you, as you load a page, Gemma is kind of sitting there in the background and it's already embedded in Chrome these days, but it's breaking down that page into 30 different clusters, of content. And then within that it's hunting for facts, or information nuggets that are salient enough and have enough signal to them.

00:32:03:01 - 00:32:28:04
Speaker 2
Then it's ranking those in terms of importance based on the kinds of question that may be being asked in the large language model, it's summarizing and faceting those clusters, and then that gets assembled into this nice fluent response that you see come out the other side. So in the old world of SEO, where we're really optimizing at the page level for page level relevance, now we're optimizing also for passage level relevance.

00:32:28:06 - 00:32:50:05
Speaker 2
So if we look at this traditional versus search versus AI search, I look at it as in three phases. I kind of took some of the borrowed some of, elitist Alice's amazing work where she was trying to compare traditional and I and I kind of felt that there was this missing layer of semantic search. So traditional search of this short keyword based, one off queries with high navigational intent.

00:32:50:07 - 00:33:09:13
Speaker 2
When we got to the semantic search layer after 2012, we're really looking at mid cell queries, voice queries, chain queries. And those are oftentimes based more around entities or topic spaces. You know, who was the president of the United States? Who was his wife? Where did she go to college? Did she play basketball, that sort of thing.

00:33:09:15 - 00:33:37:13
Speaker 2
And so they're you're really looking at entity level relevance in this AI search world. We're looking at long conversational based multi-target queries, with high task oriented intent. And underneath these prompts or of these like 30 word queries that people are asking, there's all of it's breaking that down into some queries and spitting those out, into what's called a query fan out to then do the clustering and the chunking to then bring that back to you.

00:33:37:15 - 00:33:46:02
Speaker 2
So you're getting a single synthesized answer here, and some secondary links to sources at the grounding worked out properly.

00:33:46:04 - 00:34:13:02
Speaker 2
So we use we use it for variety of tools. We're actually testing a number of them, here in health care success. And so in this case, we're using a tool called K. Hey, at least for this example, and we're using these to figure out how visible are different brands inside of, large language models. So at the top you'll see, a domain, two of them that are kind of there for comparison in terms of competition.

00:34:13:04 - 00:34:36:02
Speaker 2
And you can see we can see the percentage in which they're visible for these kinds of searches in sonar, which is perplexity, gpt Gemini and then clod. And if you notice in general, you'll see a little TD that means that they're known within the training data set versus GR means they're they're known within the ground. Meaning when it gets triggered to go off and do a search.

00:34:36:04 - 00:34:59:04
Speaker 2
So from here, some challenges we're going to find, because we do know that these, these language models can, hallucinate is I've actually caught it where it's like hallucinating URLs that don't exist. And so then that provides, you know, this whole new challenge of not just crap, my site's broken, but no, this thing is lying about me, and I need to go smack them.

00:34:59:06 - 00:35:23:19
Speaker 2
So you're going to have to be monitoring what's what's getting sped up by these and, making that part of your generic, SEO hygiene strategy. Moving forward, you're going to have to be paying much more attention to the branding side of the work. Within that, that means you have to kind of come in and look at the answers that these large language models are providing.

00:35:23:21 - 00:35:41:22
Speaker 2
Obviously, obviously, the sample data, if you try to look at this like a minute later or a week later, you could get all sorts of other things, because these tools are probabilistic in the ways that they provide responses. This particular tool does offer a way to kind of flag facts if it's just straight out a lie.

00:35:41:24 - 00:36:00:17
Speaker 2
And that signals, should be funneling back into, into those models to suppress that. So outside of your brand, then you want to be looking at specific topics. So this is why we kind of looked at ear, nose and throat, just to get the same level of data, look at what the large language models are saying.

00:36:00:23 - 00:36:22:02
Speaker 2
Look to see, are the other URLs, right? Are they even seeing the URLs with the right response codes? So if you see things like a 4 or 4, you might want to go check and make sure that nothing's up with your server. The server load isn't too high or whatnot, because these bots, as you've seen, crawl really aggressively and I have seen them take sides down.

00:36:22:04 - 00:36:38:13
Speaker 2
Then what you might want to do is get a sense of, what is what is the topic graph look like in this space? And are there areas in the same way that Google, you know, you might have this great, gorgeous website, but Google only sees a portion of it because you messed up on the indexing. And you have to go fix that first.

00:36:38:15 - 00:36:59:16
Speaker 2
You may also run into the same things here with large language models. They just may have gaps in their knowledge because they're chunking information in these bite sized chunks, and they may not make the connection to from a smaller topic to a larger one or topic. That's that's really close by. So in this case, otitis media, and swallowing problems.

00:36:59:18 - 00:37:18:19
Speaker 2
Or is is the EMT condition congenital or not. And so these actually provide really interesting insights for how you might want to readjust your content strategy or things you might want to lean in on a bit more. And you'll have to weave that into your process. So we've been finding that this breaks out into a couple different areas.

00:37:18:19 - 00:37:39:17
Speaker 2
There's, there's the structural gaps where, hey, it just hasn't discovered this whole section of your site. You might have this beautiful hub, around these, specialty EMT programs. It's just, for some reason, maybe part of that site that's broken and not getting indexed yet. We are also seeing some really interesting, themes that tend to come out.

00:37:39:19 - 00:38:08:07
Speaker 2
The large language models, especially in terms of patients, do seem to be a bit more interested in that particular audiences journey. Like, what does the course of treatment look like? What is that full spectrum? And I think for for hospital systems, one thing that would be really interesting is to pull out those experiences and, and begin to, to, to bring them in, because if they're interested in it, we know that our audiences are interested in it.

00:38:08:09 - 00:38:34:12
Speaker 2
This next layer of search, which is a gigantic, you know, the agents are basically acting on behalf, you know, and as a specific persona coming through the system. And they are becoming the mediator more, more so than your website. So in this case, we really want to look at competitive differentiation. What is your unique value proposition. And we can figure out, you know, are you lagging behind your competitors in terms of presenting yourself?

00:38:34:14 - 00:39:04:03
Speaker 2
And I would really look at, this, you know, you know, your unique kind of selling proposition, which is your physicians or your authorities, and, you know, get their hot takes on, what's happening on the ground. You know, how fast do people recover things like that that will really, set your subspecialties up to match, the kinds of thought leadership that these large language models are looking for.

00:39:04:05 - 00:39:33:03
Speaker 2
So, kind of a quick executive summary you really want you focusing on on rich Serp features, things like knowledge panels, answer cards, top stories, making sure that you're visible. AI and search engines cannot easily extract or attribute medical authority provider credentials. Patient focused content for 0.0 click experiences. Your experience, expertise, authority and trust is oftentimes going to be undervalued.

00:39:33:05 - 00:40:06:24
Speaker 2
So the actions you can take, you know, add to provider details and specialties like your organization departments, mark up your service and condition pages, structure, common Q&A content like go feed those large language models as much as you can. Link post to, expert bios and certifications. And then if you do have, some sort of reputation management system for ratings and reviews, making sure you're getting as much of that validating, trust information, into, into your site.

00:40:07:01 - 00:40:30:24
Speaker 2
So kind of moving from, from the base level all the way through, you're going to work on your indexing and crawl foundations. You're really going to look at your content, your entity relevance making sure you don't have any gaps in those, you technical and on page, you know, look at how you're structuring content. Is it clear with clear headings, subheadings, sections, call out quotes.

00:40:31:01 - 00:40:57:03
Speaker 2
All of those pieces try to make it as snackable, or add some people say as chunk more to get in included in the passage indexing. Leverage your expertise, your case studies, go beyond blog content, for crying out loud. Also look at fostering digital PR, you know, increase your engagement in the community and the communities around you, both digital and offline.

00:40:57:05 - 00:41:24:22
Speaker 2
Promote your patient stories. They're your successes, too. And, monitor your hallucinated live URLs. Redirect those as needed. Keep track of any cases where there could be hallucinations. You might also see patterns that start to arrive because of the content that you've written in the way that you've done it. And that I think is going to be a huge challenge for, for the medical community.

00:41:24:24 - 00:41:33:17
Speaker 2
So your healthcare brand is a data set. Garbage in, garbage out.

00:41:33:19 - 00:41:38:23
Speaker 1
All right. On that note, excellent. So I can.

00:41:40:11 - 00:42:00:16
Speaker 1
Empathize with some of our listeners, slash viewers. There's got to be an overwhelm factor. So, Brandon, I thought just for a couple minutes here, I would wrap up some of the points and underscore some of the things that, you just said and, hopefully bring this to, make it a little easier for some ordinary humans to follow.

00:42:00:16 - 00:42:24:02
Speaker 1
The one of the things is that I just got this from our client in Houston. We called me up yesterday and said, you know, this just seems like it has become so technical. It's just like ordinary people can't do this. And I'd have to agree. And you kind of alluded to that earlier, just the increasing complexity part. Technical SEO used to be, you know, meta tags, description tags, you know, fast site speed.

00:42:24:02 - 00:42:28:04
Speaker 1
But would you like to comment on that just for a moment?

00:42:28:06 - 00:42:55:13
Speaker 2
Yeah. And I think that's because of a misconception of what SEO is and was and has been like, you know, a lot of executives still think about it as keywords. Even at some of the large companies I was coming from, like just had complete misunderstandings about, you know, that and the interplay of, topics. Right? So, for instance, you know, there was someone at one point going off about, you know, I want to rank for running shoes.

00:42:55:15 - 00:43:25:09
Speaker 2
And I was like, yeah, but people aren't searching that anymore. They're searching things like which running shoes are vegan friendly. So if you think about the experience part of VAT, you know, do you have an experience to solve that question on your site? And so if not, then you have to go figure out how to build it. So I think that's part of it is understanding even with content and creative, there is a technical aspect to it that we have to pay attention to.

00:43:25:11 - 00:43:47:04
Speaker 2
But it's really shifting, shifting to the audience focus, on the, you know, on the technical side, you know, the crawling part has always been there, right? The log files have always been there. It's just that we're having to dig deeper and look, look harder at what's, what these systems are interested in and what they're going after.

00:43:47:06 - 00:44:09:13
Speaker 1
It's really funny you say that because I was about to ask you about keywords versus answering questions, because you didn't talk about that on the contact gap analysis for people that aren't as, versed in the stuff that you and I do every day with our team, like, that's a big issue, right? Like where? So spend just a moment explaining, again, just what content gap analysis means, where the opportunity might be for a health care provider.

00:44:09:15 - 00:44:29:09
Speaker 2
Yeah. So in some cases, you know, you're looking in areas to focus. Like in this case we were talking about EMT. I really tend to look at at least like 2 to 3 brands at a time. Take a sense of, size of organization. And I think through, like, their audiences, the communities in which they're living.

00:44:29:11 - 00:44:52:22
Speaker 2
And you would go through this, the site section by section and go, okay, well, how many specialties are there in EMT? Okay. Do we have pages for each of those? Do we offer those at all? You know, all of our locations, do we, do we mentioning that in our local business pages. Right. I mean, you kind of look at the entire skeleton and figure out which bones are missing.

00:44:52:24 - 00:45:15:12
Speaker 2
And then from there, you develop out your content strategy, what you're hitting first. There's different theories about that, whether you start with all of the small pages and then put the page in, or you start with the current page and go over some pages. There's lots of fun you can have with experimentation. There. But in this case, we were we're definitely looking at, topical overlap.

00:45:15:14 - 00:45:36:15
Speaker 2
And, and are you missing specific topics in general. And then what do you need to do. Do you need to build a page out, or is it just that the large language model, isn't isn't seeing everything about your brand or your site in the same way that a search engine is? And so then just as we do that for SEO optimization, we try to make Google look up being see as much of the site as we can.

00:45:36:17 - 00:46:01:00
Speaker 2
We have to do the same thing, but at a content layer. In terms of, in terms of individual passages. Right. So, are we talking about a page for, I don't know, dual diagnosis. And in that case, I have to make connections to, you know, drug abuse or alcohol abuse, and I have to make references into, you know, complex PTSD, for instance.

00:46:01:02 - 00:46:22:22
Speaker 2
And then within all of those, there may be a completely different path to recovery. And so that gives you more content, opportunity to really speak to those pieces as well. You know, what's the difference between someone who's going in for dual diagnosis, just or just for alcoholism? Those are probably different kinds. And they're gonna have completely different recovery journeys.

00:46:22:24 - 00:46:26:15
Speaker 2
So you need to speak to that and make it as clear as possible.

00:46:26:17 - 00:46:54:23
Speaker 1
Yep. Makes sense. So we talked about the everywhere. And one of the things that I've been talking to clients and prospects about and others is this idea that here's my oversimplification. And you can certainly add on to this brand, but you know, SEO in recent years has been, you know, mostly what we call technical SEO and content. And for a while there, there's then there's still our providers out there that say, are SEO just how many blogs do you want, you know, via blog, by another blog?

00:46:54:23 - 00:47:15:18
Speaker 1
More blogs are all great. So that's very simplistic. And then, you know, I think most people acknowledge that, you know, reputation over time, reputation became more important. The Google local became more important. You know, links have been around since the beginning of Google. Right? So that's kind of important. But now it kind of mutated this online reputation and citations.

00:47:15:20 - 00:47:34:07
Speaker 1
And today it's, you know, all those things just matter more. And then the brand and then the as well, you know, reputation plus the brand plus sort of the digital PR aspect of it. So you know, as that's, you know, again, we're like, what does that mean? We're trying to be everywhere, right? We want to look good everywhere.

00:47:34:07 - 00:47:56:12
Speaker 1
We want to be an authority everywhere. So are there any like, you know, we're almost at the end here 1 minute or 2 minute tips on that. Like so for example, should be thinking about different content types like for example blogs, webinars, podcast. So should be thinking about, you know, mostly digital PR should be thinking about different distribution channels like Reddit or whatever.

00:47:56:12 - 00:48:02:18
Speaker 1
Like what does that all mean? Oh my god, I feel tired just talking about this, let alone do it.

00:48:02:20 - 00:48:37:01
Speaker 2
Well. And this also ties in with, you know, there are ways of, of of using an AI as well to think through all the kinds of things, all the kinds of media types you may need to, spit out. Right. That there is this old concept in content marketing called the turkey dinner method, where, you know, you have this big waffling piece of content and then it's like, okay, well, this outlet's just going to get the leg and this outlet's just going to get part of the time, this outlet's going to get some stuffing, you know, and and you can kind of take it that way where you're feeling bits and pieces

00:48:37:01 - 00:48:59:10
Speaker 2
of it out just to maintain some visibility in those. Because again, the citation is no longer just text. Well, that's an interesting piece. I think people aren't quite picking up. So yes, being a part of podcast, being a part of local radio, being a part of local news, being tied into local blogger communities and there's tons of it.

00:48:59:11 - 00:49:22:10
Speaker 2
Not just like a Reddit thing, right? There's a there's there's meetup communities, there's LinkedIn communities. There's all of these things happening. It's a matter of like, where would your audience be? And and then figuring out the media types, they're going to resonate the most with them. So some of the work that we're doing behind the scenes is to actually think through those personas and scale.

00:49:22:12 - 00:49:29:07
Speaker 2
And, and that's a big piece, I think, of this future that we're heading into that is multi-mode will and be everywhere.

00:49:29:09 - 00:49:49:24
Speaker 1
Yeah. Last couple of comments and then, we'll close here the, without going over the whole webinar again, I just would like to just maybe underscore the part that if you get noise, could tape it back to where Brad was talking about. Schema is going to be really important, right. Making sure that's done correctly. Precisely. Make sure that you're not blocking your bots.

00:49:50:01 - 00:50:10:14
Speaker 1
That would be bad. And I would say probably leveraging if any, at least if you're a health care provider, leveraging the expertise of the people you have on your team, usually presumably have doctors that have some credentials. So there's there's a lot here. We don't expect you to follow everything we talked about today. And there's much, much more out there.

00:50:10:16 - 00:50:17:23
Speaker 1
If you're I'm going to sort of invite people to call us in a second. Brandon. But any other final comments?

00:50:18:00 - 00:50:43:22
Speaker 2
Yeah. Obviously. Yeah. Because that's, that's that's still somewhat for, for whatever reason, a controversy. That, that that's being you. So we know that people use it like Google uses it. It's part of the knowledge graph. Well, it may not show up always as a nice surface. And in the search results like it does say for like products, for instance, it does have its use.

00:50:43:22 - 00:51:01:16
Speaker 2
It's an I actually have seen it. Excuse me, I have seen it impact even in cases where it's this like lesser or lesser bit of schema, like somewhere in like the travel industry or whatever. Like there's all these experiments you could be running. I think the, the question that a lot of people have is like, where is schema used?

00:51:01:16 - 00:51:22:06
Speaker 2
And it's used in a lot of subsystems. It's also used as part of how the final answer gets spit out. So again, this is garbage in garbage out situation. If you if you're putting good stuff in in a clear way, you're going to get good stuff out on the other side. You're just hoping the album removes the friction from that path, right?

00:51:22:08 - 00:51:55:11
Speaker 1
So I'll just, invite listeners, readers, viewers, whichever way you're consuming this content to if this is an issue for your organization or whether you're a SaaS company, telehealth provider, traditional provider, new kind of provider, multi-specialty hospital, health system, pharma device, whatever health plan, we really do have this expertise here. Brandon. And our team has created an amazing a variety of people at all different levels for the paid side, the organic side, the content, the technical, the development.

00:51:55:13 - 00:52:16:21
Speaker 1
So if you're interested in all of these kinds of things and you need help, I will put our team against just about anybody there. Brandon, I think we've got a very, very, very strong team. So we're happy to talk to you in detail if you're looking for help. Thank you for attending, everybody. Brandon. Awesome. This is fun. Redoing a webinar with you again.

00:52:16:23 - 00:52:22:15
Speaker 1
This is part of our DNA. We love educating people out there and, trying to help. So thanks for timing.

00:52:22:17 - 00:52:25:09
Speaker 2
All right. Thanks everyone.