Copy the formatted transcript to paste into ChatGPT or Claude for analysis
My name is Garrett Susman. I am the
director of marketing at IPole Rank. Um,
we are going to go through this local
visibility in AI search. You're in for a
treat. If you haven't been paying
attention, a few months ago, uh,
Christian, who I will introduce in a
moment as well, will uh, walk us through
all the data he's been sifting through
and analyzing at Yex. Yeah, we we always
chat and we're always kind of
speculating and discussing um everything
that's going on in AI search, obviously
in a local context, but in the bigger,
you know, search ecosystem context. And
so I highly recommend if you don't
already follow Christian on LinkedIn,
follow him. He writes, you know, weekly
blogs where he's kind of um explaining
and expanding his his thoughts and
theories on what's going on. And it's
such good material. So Christian, should
we should we dive into it? You you are
in control of the deck. Let's rock and
roll.
>> Absolutely. And uh thank you again for
the opportunity. I love I love speaking
with you and Mike and but also just in
general this whole community is so
outstanding. Um and so uh we've got both
of our logos here. I think we can blow
past that. Why don't why don't you get
started with your context?
>> Yeah, real quick. So I think context is
really important in AI search.
everything. We all see these
personalized experiences whenever we're
searching on AI mode or chat GBT. So now
whenever I do uh a webinar, I want to
give some more context about me. So my
name's Garrett. I live in Virginia from
Jersey originally. Lived in California
and Louisiana. I've got a beautiful wife
and daughter and a mini golden doodle
named Lucy. I'm a Philadelphia sports
fan and I love reading all about
psychology, thinking fast and slow,
influence, and amazing graphic novel
novels like Day Tripper. Over to you,
Christian. Uh I am also Jersey Strong
originally, so we share that in common.
Uh I I am a father of three. I live in
Florida. Uh this photo is a little
older. We actually retried to do this
piggyback thing uh this year. And uh let
me tell you, I almost died. So uh but
yes, I I've been in structured data
since I was 23. Um so I've sort of been
building out uh in structured systems
for my whole career. Uh and uh I'm
actually I've been at Yex I cannot
believe for how long at this point. But
uh my role at Yex is I really focus with
our clients about their data and how
they use their data in order to show up
more. Originally it was always in
search, but now it's really an AI
visibility and search combined because
as Garrett and I tend to agree, I don't
think we're really going to be talking
about search and AI search very much
longer. I think it's all going to come
together. And when that happen, well, we
did that at the same time. Once it all
gets there, I think we're we're really
going to see um changes in how we as
marketers and all of you as either
agencies or clients or marketers
yourselves, we're going to have to
rethink all of our KPIs and how we sort
of come to value here. Um and so, uh I'm
I'm going to jump through a bunch of
slides. Garrett's going to interrupt me
viferously. Um and and in the process, I
I mostly I want to sort of tell you
where do we see this currently as EXE?
Um, we're obviously probably one of the
largest brand visibility companies, but
at the heart really what we have been
doing for a very long time is just
trying to get the data right in as many
places as possible. And and we do that
because there's a lot of proof
statistically around, hey, if you get
your structure data right, whether it's
on your website or on these directories
or Google business profile, all of these
things work together. And uh, I don't I
don't do SEO. I don't claim to do SEO. I
live in that weird space where if you do
the data right, the SEO really they they
kind of they work well together. Um, and
so a lot of this research that we're
going to talk about today is about that
and around that. Um, and I'll start with
this slide. I think it's kind of funny
to me because I think we made this slide
or I made this slide maybe two and a
half years ago. This is kind of funny to
me like this is like a foregone accepted
thing now like that it's fragmenting
like for the longest time and I still
see a lot of arguments online where
people are like oh well all the search
the the number of the number of searches
are still in Google that's the that's
where it's all at I'm like you're
measuring that's like measuring a
baseball player by the number of swings
at that right like that's not how you
measure search has to be around the time
you spend on the platform because that's
where the the money is made and it's got
to be measured on the outcomes. And I
think a lot of people are realizing
these AI outcomes are getting better and
better. They're they're not there yet,
but they're much better. And we're also
seeing, you know, we're going to see
returns from Black Friday and some of
the the holiday shopping. I think this
year you'll see the first impact. And um
what's interesting to me, too, is just
like this being a two-year-old slide, I
mean, it's missing what you would now
assume like Reddit and Tik Tok, like
search has even expanded in a lot of
ways. Um and obviously those are
integrating AI as well.
>> Yeah. And and I think that's a great
point which is so um I I don't know if
this might have been on one of your
posts but like my kids don't Google
things anymore that so that they and
they're 191 17 and and 11 so they don't
use it as a verb. They search search it
up or search and I remember when um uh
Mike first talked about that online and
then my kids started doing that because
they don't use Google they use Tik Tok
but Tik Tok is going to add a large
language model that'll melt your face
off pretty soon. So, we're getting to
where those are going to control so much
more of that time. I think today, and we
can see this in the data. I bet you a
lot of your your clients and and many of
the SEOs can see this. Visibility in
Google might be down, but you're still
getting the clicks from Google at about
the same rate because people are having
their exploration in these AI,
especially for complex things, but then
they're coming to Google to like execute
still. that's going to stop really soon
as these AI models add more executions
or exploits or calls to action in their
model. So, you're exactly right. My
slide here shows models from back in the
day, but it's where these models are
being embedded like Garrett, do you have
did you do you have Alexa Plus at home?
Do you have an Echo Dot in like
>> Yes. Yes.
>> So, you know, that's a funny thing. So,
we have them in all of our rooms and and
for everybody that hasn't tried this
yet, enable Alexa plus in your room.
It's awesome. Like compared to like you
remember it used to be like Alexa turn
on the music in the living room and it's
like starting the engine in your car
like it was so bad it was so
disconnected. I think now if you look at
how well they're putting that in there
that's where Amazon's they've got a big
advantage. They have 600 million Echo
Dots live in people's households. So
you're going to see this fragmentation
continue. Um now Garrett and I talk
about these these sort of three things a
lot. I I I know I said there's one like
it's fragmenting. I think there's really
three fragmentations everyone has to
focus on. The first is search. Um that I
think is well accepted at this point.
The second I don't think people are
talking about nearly as much, maybe you
and I are, but not everybody else is
ranking fragmentation. Um and and I'll
dive into exactly what I mean by that.
And the last is experience. And that's
just getting started with things like uh
uh browsers powered by AI uh as well as
like Grock in your car. Like things are
starting to expand. I think glasses are
going to come out real soon. Um, you
know, the Facebook ones, if you tried
them, are pretty amazing, but once
that's full AI, that's going to be sort
of a new experience layer where you're
just walking around, you're searching at
the same time. Um, so search
fragmentation, I think it's been covered
really well by the media at this point.
There's a lot uh that to point to. I
think this industry, the SEOs in
general, like um they they understand
that this is happening. I will say at
Yex, we work with really large
corporations uh with thousands of
business locations and then we work at
really with really small businesses
through our partners and through SEO
agencies and and ultimately it's not
really how big a organization, it's like
how savvy do they understand what's
happening right now. And so we're seeing
people finally understand this. Now this
chart went around on the Financial
Times. Uh this is kind of neat seeing
like so more like search level charts in
such uh esteemed publications. Um but
this chart shows in two years um chatbt
just their just their users of just
their chat not the integrated chatbt
everywhere else just chat GBT hit as
many users as the entire internet in 14
years. They did it in two years. Now, I
I want to offer I don't think any of
these adoption charts are really um the
right way to think about this because
like Garrett, in year 1, two, and three
of the internet, do you remember when
you'd have to like hold the phone and be
like, "Hey everybody, I'm getting on the
internet in the household and like put
the phone like if somebody picked up the
phone, you'd lose your connection to
AOL." Like that was a lot of behavior
that had to be learned from when we talk
about psychology. That's not happening
here. I don't have to learn anything.
There's zero cognitive burden to
adopting this AI because I'm not
adopting it, it's adopting me. And so I
think these charts are comparing things
that are really apples and oranges,
which is, you know, one is sort of like
how how do I learn how to use the
internet? This is I don't have to think
about this. I can just do it. One call
out there I think is important too
though is is is like I I'd love your
take on extrapolations of these charts
because I think there's some big factors
that could influence it in the direction
of like whether it plateaus or not. One
being the default nature of chat GPT
versus Google could obviously influence
this and then people are say you know
you you see people on the other side
arguing that chat GBT may have hit its
peak and if that's problematic you and I
don't necessarily think so but what do
you say to that? So, a couple things. I
I think right now the business models
haven't all been worked out, right?
They're they're they're still so they're
still getting started. Like we could see
ads in chat GBT. I know some people have
found some technical specifications that
ads could be coming. I think as the
models settle down as to where people
are going. Yeah. Yeah. I agree with
Geek. Like the reality is I I think
that's going to happen there. It's too
big of a market not to do it. I think
it's going to be offers, not ads. To to
dive a little into the next section, but
to this, I don't think it's really about
the models. I don't think people even
begun to use the models at the level
they're capable of. And so, we're going
to see a lot of that sort of continue to
work its way into the daily lives. I
also would say I don't see Google just
continuing to do search. Gemini 3. I
think for everybody on the call, I'd
love to see in the chat if you disagree,
but if you're using Gemini 3, you could
almost feel the air blow back into
Google's lungs like they launched it and
they're like like it's it's that good.
And so I think with that, if they didn't
have that sort of breakthrough, I think
maybe what you're saying could have been
more likely that ah this thing's really
not it doesn't have the juice to to to
to continue. So um but we'll we'll we'll
see. Um, yeah, the code red. That's
funny. Um, now this is a really
important trend. If if you if you're not
following this, I'm going to take you
again in my historical machine here to
Hoboka, New Jersey to a little company
called Jet.com. Do you remember when
Walmart bought Jet.com? That was a big
acquisition. It was 7 billion roughly.
Okay. They spent 7 billion dollars on
the thing to revamp all of their retail
website. They've since spent probably
hundreds of millions in the years
following optimizing it. them saying 20%
of the referral traffic is now from chat
GPT and then before Black Friday signing
a deal that says you can just check out
in in the chat. You don't even have to
come to Walmart. I want everybody to
think about that. That is one of the
largest websites and one of the most
expensive websites in the world goes,
"Eh, just just check out over there.
Don't even bother coming." They're not
capturing any of the metrics we used to
look at. They're just they just want the
checkouts. And I think for many
businesses, while this seems
uncomfortable, what you have to realize
is the website is still critically
important, but it's a conveyor
[clears throat] belt of data into these
large language model interactions where
you want to win in both. And so this is
going to be this interesting time period
where a lot of websites are going to
lose a ton of traffic, but if they play
their cards right and how they are cited
and the knowledge they provide, I think
they're going to have phenomenal
business outcomes. Now, to everybody
that's an SEO or an agency on the call,
that's going to feel really
uncomfortable because you're going to
see a lot of metrics you've always loved
and you've spent your career learning,
not really show any of the value of what
you're doing unless the the AI companies
give us data back. I'm old enough to
remember when Google Business Profile
didn't provide any data back and now we
have a ton of data. So, this might
happen, but it may not happen and you're
going to have to be comfortable with
that from a a privacy and surveillance
perspective. And one thing I'll add to
that is just like just to speak to that
there are so many strategic implications
for that for business. Like to your
point like Walmart at the time made the
best decision it had with the
information available. Yeah.
>> Even if they're burning billions of
dollars. The question is like it's hard
to know definitively which direction to
go. Whether or not you know we already
hearing conversations of serving up two
versions of your content um one for
machines and one for people. Is that the
right way to go or finding a smarter way
to integrate them both?
>> Yeah. Yeah. You know, it's funny because
I'm seeing this debate like LLM.txt and
all these different things. Uh listen, I
I I think everyone should keep their
powder a little dry on how how
forthright you are on like one opinion
or not. Here's what I would propose.
Like uh Dwayne Forester, one of our
mutual friends, he just published a
book. Um everyone should go buy it on
Amazon. Um he's been talking about this
for Yeah, there you go. He's been
talking about this for a while where you
know realistically the world because of
Google the world mo merged the human
intera experience with the machine
experience by putting schema and
embeddings and JSON LD but that's
actually not the optimal solution. The
optimal solution is that there's a
machine layer which I don't know if you
have any of you done this this is a good
test. Go look at one of the websites you
run or create. Look in the browser. Use
that plugin that shows you what how many
megabytes it takes to render that page.
A typical page for like a Wendy's or a
Taco Bell might be 350 megabytes to
render that in a browser. That's not
helping the machine or the AI. When you
realize if you take that page and you
put it in a large language model and say
just tell me all the structured
knowledge that would be helpful to you
on this page, you know how many you know
how big the file is? It's 35 kilobytes.
It's a it's a massive factor difference.
So the AI could be a lot more efficient.
We're going to talk about that in this
section. Um so ranking fragmentation.
Now, uh Garrett and I go off this a lot
because there's a lot of psychology
implications on this. I I happen to
believe the whole forgive and forget
thing in psychology. It's it's more like
you can forgive because as humans, we're
designed to forget, right? We do not
have perfect memories. And this is going
to be the first time where you have a
goodwill hunting in your eyeglasses that
never forgets anything. I'm not sure
we're all ready for this, but I want to
talk about in terms of search. When you
look at each of these models, all of
them have memory. All of them expose
memory differently. Uh Chad GPT is
probably the best to test this on when
you're looking at it. But Sam Alman came
out and said just after the launch of
GPT5, GPT6 is going to be all about
memory. And I really think this is going
to happen, which is if you think about
classically speaking, I'm gonna use an
example I use a lot, which is um for 19
years, 18 and a half, I have typed into
Google search um tree nut friendly
allergy menu dinner. Why? Because when
my 19-year-old was 6 months old, we came
to find out she has a horrible tree nut
allergy. Okay, why doesn't Google just
remember that? It's been 19 years. Why
not? If someone's vegan and they say,
"Hey, I'm a vegan." Why not remember
that? Wouldn't it make search better?
The answer is ads. The reality is if I
can't have the right to show you steak
restaurants because you told me you're
vegan, that's a problem for the
available ad space. Think of ads as a
circle. Search is a search the word is
from the Latin ciar. It means s circle.
You're encircling everything. The bigger
the circle, the more ads. Memory. Every
memory an AI gets shrinks the ads and it
but it but but it increases the value of
the ad because you're much further down
the funnel to me because you might adopt
a a better offer to me. So what what I
think you're going to see is when we're
using search today in AI the AI is using
memory context your location and other
things but all of those things mean the
results are going to be very very
different. It's not like rankbrain where
maybe what I don't know what you all
think but it's like 8 to 10% rankbrain
influence depending on what you're
doing. This looks like it's like 100% of
the results could be influenced by
memory. Um do you have any thoughts on
that? Because I know I know you're you
you're so on the psychology of what this
is going to mean.
>> Well, I I think it really from a from a
from a business and and and um kind of
platform side, it really speaks to like
the stickiness of like what platform you
choose. And this is why, you know, it's
so hard to argue that Google isn't going
to dominate because it has so much data.
You're so ingrained in the ecosystem all
over the place. I think that when it
comes to the there's also like a content
problem like we only have so much
granular content. So there's limits
there, but ultimately I think as
marketers and SEOs, what's frustrating
for us I I hate to say it's frustrating
for us, but like it's a better user
experience to your point.
>> Yeah, it is.
>> Yeah. It's not great for us, but it's
better user experience. So,
>> look, you know, you have to you have to
take a step back, which is a huge
portion of who we market when we're
marketing. I it's like uh there are
people that are never going to download
an app. They're never going to engage
with your brand. They're never and
you're still spending a huge amount of
time to market to them because that's in
the essence you're trying to get known
and visible and discoverable. But to
some extent, if you're vegan and you're
the local steak restaurant, I shouldn't
be marketing you anyway. So I think
what's going to happen is there will be
levels of memories. There will be
explicit and there will be implicit. So
if I say, "Hey, I need to stop pick up a
cup of coffee on the way to pick up my
son from lacrosse and I'm talking to my
car as I'm going." My car can now infer
I have children. It probably already
knows that from weight settings in the
in the seats. It knows that that where
my son uh does lacrosse. He it knows he
doesn't. I didn't ex mean to tell it any
of that, but that's actually still so
much more accurate than trying to cookie
pool me all over the internet to figure
out that I accidentally clicked on a
Casper mattress ad and the thing's still
chasing me six years later. So, there's
this element of I think it's better for
both sides, but I do want to stress
again, I think it's going to be
uncomfortable. I think it's not the way
we've been brought up. Um, and I'm I for
one am here for this. I'm a surveillance
junkie in that I hate it. I think this
is such a better way. Yes. To GE's
point, it is a privacy paradox again
where everybody's like, "Oh, my
privacy." I'm like, "You log into
Instagram 97 times a day. Stop it." Like
like like stop. So the reality is is I
think it will be paradoxical behavior,
but I do think it'll be a better
experience with memory than without. So
I I do want to offer this for all the
people that do act as agencies or help
uh help people understand where where's
the north star if you're developing for
companies or brands their strategy. And
I think it's really these two axes. It's
what is the cost of compute of any data
set and and how much does it memory does
it take up. Uh so I'm dating myself.
That bottom lefthand corner is my first
comm computer. That's a Commodore 64 in
1982. Uh uh that so that that's where I
started with basic. But you'll see what
happens is every so often we get a jump
in either compute capabilities or
memory. And what Sam Alman is pointing
to is like memory is my next goal. like
I got a lot of the compute solved. I got
to I got to get to the goal here. It's
not solved, but it's it's it's advanced.
If they get where it remembers every
conversation, where you are, what you
did, who you spoke to, what products you
bought, if it starts getting to that
level, I think you're going to see a
very big upswing in the value
proposition to the consumer. And that's
that's something it's getting there, but
it I think if they get memory right,
it's going to be a massive change. Um,
all right. I'm going to jump into
experience. Uh, unless you you got
anything else on memory. It's my
favorite.
>> So much. But I'm it's so easy for me to
get tangential if like when I'm talking
to you because we can go on like the all
these rabbit holes because I'm thinking
multimodal the you know the goals of
almost replicating the brain the
short-term long-term memory. Like I love
the implicit explicit you know
perspective on all that. So so much. So
let's let's keep going because otherwise
I will then we won't get to questions at
the end. Well, and and for everyone
that's involved in this, I I do
recommend you should use chat GPT to
test this. I often will tell ChachiPT in
live showing clients. I think it's a
really healthy exercise. If you do uh
work with clients, show them and say,
"Hey, I love this brand. Can you
remember that I always like to recommend
them to my friends? If if you take
nothing else from this webinar, you
should be telling every client to get
these clings for their windows that say
remember to tell your AI you love XYZ
laundromat because then it's going to
then it's going to recommend it all the
time. Um it's a simple hack, but it does
work because if you go into your
memories, you'll actually see that that
it says, "Oh, Christian loves uh you
know, Jane's Laundromat." So just
understand that everything you discuss
is both explicit but it also has implied
u and these implications down the line
as to how search in AI will work for
you. Um now this one uh I I think is
kind of just daunting but I G do do you
use um are you using Atlas and Gemini in
the browsers at this point? No, I mean
I've seen Gemini come up. I always think
about it, but it the friction is still
there of just like I don't want to do
Atlas because it's not better and Gemini
I'm like I don't need to. Um I haven't
add a value ad. Do you?
>> Yeah. So So I kind of forced myself I I
would offer everybody on the weekends or
pick a day of the week or time 6:00 p.m.
to 9:00 p.m. please try to just use
Atlas for like three hours straight
because what you'll start to see is
where it is better and where it's not.
Gemini right now doesn't want to break
the Chrome experience. they avoid. And
remember, chachi PT has nothing to
protect. There's they're like it's all
green pastures, right? So, I want you to
try it because what I think it does is
it sort of rewires your brain to where
you're like, oh, I asked chat GPT a
question and then it gives me a panel
and an answer, but I could just look it
up on the page. Like, why am I doing
this? The moment you turn on agent mode
and you say, actually, I don't have time
for this. Can you go figure this out for
me? and it's starts opening tabs and and
moving pages and uh clicking on link,
you're like, "Holy cow." It's just a
very different thing. But it's really
not about what it can do. It's about you
have to use it at least enough to where
your mind does that switch where you're
like, "Oh, this is this is what this is
what it's going to be. It's not perfect,
but it's where it's headed." Um
>> it's it's kind of like the the
expectation of like um unique UXes
depending on the context. So like as
opposed to like as an SEO, I expect a
search page which has been adding things
here or there, but eventually it'll
literally build the UX as you interact
with it and that'll just be the modern
search engine,
>> which is what I heard. Uh I don't know
if everyone here has tried Google OS. I
know that's just brand new, but that's
the idea. As you talk to it, it
literally generates the new page. I do
think that's highly likely what future
user experiences look like. I'm sorry to
all the designers. You're like, but look
at the chart I made you. It's like guys,
they're going to tell you what they want
and then it's going to render. You're
going to control the form, the
understanding, but it's very much going
to be about the data creating the
visualization on the fly or the uh the
experience. Um, now, uh, this this chart
blew my mind a little bit, which is if
you look over here on the right, that
was May when like we're not talking like
years ago. Look at all the model
expansion in just a little bit of time
to Gemini 3. Um, I I show you this
because I think we're all uh if if no
one's sleeping, I'm with you. You could
call me like, well, we're not sleeping
because every day there's a massive
change here and I think it's really
important like um to to people closer to
my age. Many people don't remember sort
of the dawn of the web, but like that's
where I started my career and and I I I
social go down as one of the worst
mistakes in the history of mankind in
terms of what it did to us. But social,
mobile, there was all these revolutions.
Nothing feels like this to me than the
internet. When the internet started, I'm
like, that was the level of, oh my god,
what is possible? And I think people
should be excited and optimistic. I'm
all for the skepticism and being
careful, but this is this this type of
like when was the last time you you had
products and goods and services evolving
so fast that every day there was an
update to Nano Banana Pro? Like just
just amazing. Um uh if you're not using
Google uh Gemini in your browser, if you
don't want to go all the way Atlas,
please enable Gemini in your browser.
It's it's a great experience. Um and
it'll read things to you. It's just it's
it's getting so much faster. [snorts] Um
and this is Atlas. I'll skip that. So I
want to get into if we can sort of what
is the research. The the the last was
really just to show you all this is
where we think it's going and and what
we're focused on with clients large and
small, which is I think there will be a
merging of these data sets, but I also
think everyone's got to take a little
bit of a breath and understand that
there's not going to be a like a
onesizefits-all solution, nor do I claim
that we have the solution. That's more
just here's what we're looking at. Now
I'm going to show you how we're looking
at it. Um, and so when I think about AI,
um, I I think AI has a trust problem.
And I don't mean like you shouldn't
trust it because it's getting so good.
It's it's pretty close to being as
trustworthy as search. I'm saying if you
have a situation where you ask the AI a
question, you are going and it gets it
wrong, you're going to stop using that
AI. Now, now, why is that an issue?
Well, because if it was hard to switch
to another AI, like if I'm using Gemini
and I want to try Grock, Jared, how much
time and effort do I have to go into
training myself to learn Grock?
>> Nothing. You're instant.
>> Nothing. Human language.
>> Going from like going from an Android
phone to an iPhone almost killed me four
years ago, right? Like like that was the
level of cognitive burden you're talking
about. This is zero cognitive burden. So
the AI companies their greatest risk is
that you don't trust them because you'll
switch and you there's no cost to
switching in terms of uh cognitive load.
So we look at this and go well if trust
is the currency by which AI will be
judged and and scored then citations are
the receipts they are the thing that the
AI is going to bring forward. And so I I
just show you this because this is a
typical sort of dialogue that might
happen. And so when you look at this um
it's sort of like what is the best
Indian restaurant near Soho that is late
night dining and patio seating. When you
break down any query you really can look
at it as like best is a subjective
qualifier. Not everybody always uses
that. If it doesn't if it's not there it
might just be an objective question.
Indian restaurant. Now now a lot of the
subjective stuff you get from review
sites because that's where you can
gather an opinion. Um Indian restaurant
that's a category right? Categories you
get from listings and directories. So,
Google Business Profile is probably one
of the best sources of categorical
information because not only can you
tell it what yours is, but Google also
applies other categories as it deems
appropriate given your local competitive
space. Um, Soho, that's a location. It's
looking at maps, local pages, directory
pages in in uh areas, late night dining
and patio seating. This is one I want
everybody to focus on because I think
this is an awesome opportunity, which is
objective knowledge way beyond the
classic stuff you'll see. So, objective
facts are mostly from your websites and
your pages. Now, I want to offer
something else about this. Every element
on this page has a different level of
potential verification by a
probabilistic large language model. What
do I mean by that? If you have patio
seating, but the only place on the
internet that says you have a patio with
seating is your website. That is a
source of one. There is no other
verification for it. Now, if that same
thing shows up in a review on a review
site, now you're up to two. If there are
listings and directory sites that have
flags for patio seating, you might be up
to 12. Remember what I'm talking about.
This is not a deterministic model. It's
not all rules like Google is. It's a lot
more math. the more the probability of
the number of times you can find the
same structured data, the more likely
that will show up in a in an answer by
the large language model. Does that make
sense hopefully to everybody? It's very
very different than what you're used to
where oh, if I check the flag, I'm good
in Google. I'm like, yeah, but that's
not the only battleground. It does help
in Google if you have it everywhere, but
this this is really probabilities. And
so when we look at this, it's sort of
like trust in citations. It's all these
different levels. It's the question,
it's the context, it's the location,
it's the model. So, let's let's let's
dive into our research. So, um this is
really simple. And I know there might be
some uh Garrett Garrett pushes back on
me this all all the time. So, so if you
ask your AI, hey, if I ask you about a
business or a service, do you use my
location? Like if I ask you about a
business or a service, I'm not asking
about is the sky blue? Forget that. I'm
not I'm not worried about that in my
job. What I'm saying is is if I'm
looking for a business or a service, do
you use location? Now, this says short
answer, yes. Yes, absolutely. And yes,
uh blah blah blah. Perplexity is way too
verbose. But but but the the thing about
this is you can also unless you tell
your AI to not use your location, you
can guys, everyone can just test this.
You can just say, "Hey, I want get me
lunch near me. What what should I get?"
And it's going to it's going to show you
that it knows your location. Okay. Now,
what I would argue generally is is every
research study that you've seen so far
on citations ignores this. What they
typically do is they take a brand and
and by the way I'm not saying this is
wrong. I'm saying it's a very different
view than the consumer journey with AI.
What they do is they take a brand, let's
say, you know, Wendy's, it doesn't
matter. Um, the Starbucks, they take a
brand, then they take a model, right?
And so the model, they then go and say,
um, okay, now do a query fan out. And I
think, you know, Mike and Garrett, you
guys have all done a ton of work on
this. And I think this is there's some
really great research by the SEO
community on Twitter around or X on
this. Um, but but what you do is you
basically get a series of questions and
and the questions are like, what does
Wendy's do? What does Wendy's
specialize? What services does Wendy's
offer? Okay, when you look at that, of
course you're going to get Wikipedia. Of
course you're going to get Reddit. But
that's a brand level question. It's it's
agnostic to the human. Now, here's what
I'm going to offer. Unless you're some
sort of spiritual being, you're a human
in a place as a consumer. You are
somewhere, right? Like you could you
could think spiritually. You're here.
Wherever you are is being loaded into
the AI automatically. That could be
memory or your settings or both. So
unless you tell it, hey, I'm in New York
City and I want I want a hotel in
Florida, but even then you're projecting
your yourself into that location for the
question. So you look at it this way and
then again you choose a model, you
choose an interface. You can track many
of these things now. And then memory is
this caveat. I can't track it. I have no
idea what it's doing, but I know it's
there. Um, we did it a little
differently. We then broke down our
questions into branded questions and
unbranded questions, which is an homage
to the SEO community for having thought
this way for the generation. But but in
doing this, we we then broke it down by
objective questions and subjective
questions. And the reason why our query
fanouts look like this is what we're
trying to get at is if I mention a
brand, what do you site? If I ask if the
brand's any good, what do you site? If I
don't mention brand but a category, what
do you site? And the reason why we look
at it that way is that is how most
people think about their marketing and
their content strategy and whether or
not they should worry about where the
citations are coming from. So then we
collected this, we did this study and I
just want to take there's a lot of ways
you could categorize digital data. I'm
going to throw out just four categories
and the way we think about it is what as
a business or brand can you control?
your website is 100% under your control.
Well, unless your IT department sucks,
but mostly it's under your control.
Okay? So, so if if you can control this,
you basically can say, I want to talk
about my offers. I want to talk about my
events. I want to talk about now next
one, listings, you can control a lot of
it. You can't control all of it. And why
do I say that? Well, so Google, if I'm
having an event like a a financial
advisory forum at a local hotel where
I'm going to explain trusts and and and
REITs, whatever, I can't put that in
Google in the Google business profile.
Why? Because they want me to buy an ad
to promote what I'm doing, right? I
can't just feed that in. So, your
website, you can have it. Your
directories, they scale better than your
website, but they don't have as much
flexibility for you to control. Now, I
moved to reviews and social. Now, I know
that's going to offend a lot of people.
I'm grouping them together in one
category. And it's because it's really
an engagement layer. Okay? I'm not
saying I'm not talking about the
platforms. I'm saying you can control
some of this, but it's really where you
can engage, but you can't stop what
people are saying in both. And with
social, you really can't either. So,
they're both they're almost like
birectional uh communication systems,
not really the same thing as getting
your data out there. And the last one
I'm throwing Reddit in, which is look,
you can't control Reddit and you can't
control what people say. I know
everybody's using, you know, AI to
generate a bunch of crap there now, but
the reality is is those forums and news
releases and CNBC covering your brand,
you can't control that stuff for the
most part. Um, now, if you look at the
world that way, we gathered this data.
Now, I I just want to show people really
quickly, uh, this is this this is, uh,
by the way, all this research, we'll
we'll show you later. All of it's been
published. So, we do publish at Yex
Research, and we'll we'll get you a link
for that, but just so you understand how
we're doing this. Um, so this this is a
map. Um, this is actually Starbucks. I
just randomly pulled up uh Starbucks
looking at all their locations. If you
look at this map, this is the Google
organic ranks in all these little
hexagons. We're using the Uber
open-source H3 hexagonal mapping system.
When you zoom in, what I can do is I can
see every Starbucks location. And as I
get really down here into the the the
smaller areas, I can see this one
location. And what I'm doing is I'm
asking those questions in these areas to
find out how are they ranking against
their competition. Uh Starbucks at the
moment, you can see there's a lot of
red, but there's also a lot of green. It
depends on usually the business density.
And when we the reason why we do it this
way is I don't think for Starbucks, the
Starbucks in Roswell, Georgia has a very
different competitive profile of who
they're competing against than the
Starbucks in Midtown Manhattan. So the
classic SEO of best practices, it kind
of breaks down a little bit. I'm not
saying they're not still good things to
shoot for, but like the one in Midtown's
got to do something different than the
one in Roswell. And so that's the we're
gathering the data at this hyper
granular level. And then we're gathering
all of the competitors around Starbucks
to know what are they doing and what
citations show up for them. Does that
make sense? I know it's it's a lot, but
it I want to show you so you can see how
we're gathering. I'm glad that people
are seeing this because I think it
really highlights, you know, obviously
the difference between a single location
SMB and what their universe and their
their zone of influence is very
different than a Starbucks that has, you
know, thousands and thousands of
locations. But I think it also
highlights, and I know you're going to
get into this too, the way the consumers
decide where to go based on not only the
density, but the type of service or
business or product, the data. And to
your point about data in general, the
way we need to think about this is
>> you really need to focus on what your
data and your competitor's data is. You
can't just blanket statement of like
what you're seeing. you know, we're
looking at citation levels across the
entire country, but there's going to be
so much more value of what you're
looking at in your geography, for your
business, for your industry.
>> You got it. And and it's also, I will
say, um, there's a difference between
how Chick-fil-A is discussed in the
Southeast and the Northeast.
>> Right.
>> Right. And so, you want to understand
your brand, but you have to understand
the location matters and countries and
Europe and everything else. We we have a
global view of this. um that what we did
was we're running for all of our
clients, prospects, like and you can try
all this. There's a free way to try
Scout. I can show people later, but we
ran like a very short window like 45
days or so of just pulling in all the
citations and all the queries that were
going out based on our clients. So, we
did a study on 6.9 million citations. Uh
I think we published this two months
ago. Um and we did it on three models.
Uh I was going to do it on Claude, but
Claude for like geeks like everyone on
this call probably that's really our my
coding thing. I don't necessarily use it
in the wild as much. So, um, but but if
you look at these, uh, the way that we
did it was, uh, here's the the counts of
citations. We launched Scout, the tool
that does all that in the US. So, that
was it's a predominantly US focused
study, but we started getting results a
good whole 429 citations out of Russia.
But, but the reality is it's this data
grows every day. Uh, which I'm going to
get to at the end of the presentation.
But this is I want you to look at this.
This is by question type. So, a branded
objective question. What time does that
Wendy's open? To a branded subjective
question, is that Wendy's any good? Uh
to an unbranded question, where can I
get a burger uh uh near me? To where can
I get a good burger? And and some of the
percentages are really quite
fascinating, which is if you break down
the citations into things like websites,
third party directories, local websites.
Uh by the way, this is I know this is a
third rail. I love touching this one
with SEOs. So you know the whole
subdomain versus the folder structure
[laughter]
in local. Okay. So this one I love. So a
lot of our cl So we power a few billion
pages online. Half of our clients do
subdomains like like location you know
coffeeshop.com and half of ours do
folder trees. And the religious debate
I've just watched from afar for years.
Can I tell you the AI doesn't care. It's
so obvious already that these local
websites that have the subdomains are
already such a large percentage of
what's cited. The reason I think they do
so well is the AI doesn't have to
compute all the folder structures. It's
a much more concise specific page like
offers dot. So we'll see if this holds.
But um the mo most important thing to
look at is these three categories you
can control almost completely
directories and websites. And and this
one's fascinating. You see Gemini for um
branded, subjective or unbranded. See
like for directories. Some of these
models use them a lot. And other models
like Gemini, it looks like they don't
use directories as much. I'll tell you
why. It's kind of fascinating. Google
does not site Google Business Profile.
They just put it on there, but they
don't site themselves. It sort of be
like me talking about myself in the
third person. They don't do it. So what
happens is that doesn't mean Google
Business Profile isn't the source of the
data. It just means they don't site it
in the way they return their values.
Whereas the other people do site Google
business profile. So you see like
there's all these fun little intricacies
of how the data works. Now to all you
Reddit fans when you add location the
news blog forums look at these. They're
all one's over 5%. They're all single
low digits. And that's not just Reddit.
That's Reddit, Wikipedia, and all the
other news forums. anything we could
categorize they they're a very small
percentage again I I I implore all of
you try this on your own phone be like
hey I need a new surfboard and I want to
get it locally and see what happens you
are not going to get Reddit you well
surfing you might there's a huge surfing
community on Reddit but the rest of I'm
like you won't you won't see it you're
going to see very interesting sort of
structures now it is different by model
so each of the models weighs more but
look at the average percentages of just
websites and directories hopefully for
all you SEOs those that are on. This is
a great opportunity. It's another way to
demonstrate. Now, reviews I was
surprised about and social I was
surprised about. I thought you'd see a
lot more of that. In fact, if you go
down here, you'll see in in retail and
in food naturally, you would expect
reviews are much bigger as a percentage
of what they are, but I thought they
would dominate. And I think what's
happening is if you're a review site,
this stuff is an existential threat to
an aggregation platform of opinions.
because if I can get all the opinions
there, why am I going to your site? So,
they're starting to block. Actually, the
new study, we could see a smaller
percentage of review sites being cited.
Unless they do a content deal, I don't
think they're going to be able to fight
this trend. Like, meaning
like I imagine because like Yelp is
notoriously latigious, which makes a lot
of sense. And I assume that Google
reviews are baked in. And to your point
about them not being cited for Google
business profiles, do you think because
you'll see for instance on search that
it references like the knowledge graph
like in e-commerce you'll see the
shopping graph. So obviously they're
pulling in even if they're not giving
you specific um links. The other thing I
want to kind of call out um stop you for
a sec because Adrian wanted to know
specifically about professional service
firms. I know it's not on on this graph
outside of like banks or whatnot, but do
you have any insights that you can speak
to there?
>> So um Adrian, you can absolutely reach
out to me. I will get you. So, what I'm
getting to next is this initial study
was a lot of big brands in these
categories. Um, and what I'm about to
show you is where we're going with that.
So, we'll be able to show you like law
firms, uh, you know, plumbers, like much
more granular. And really, it's more
just because of the way we're initially
we're focused on our clients that signed
up for Scout and started using it that
the data comes in because so I can see
it. But, but, um, we also ran an index
on a hundred top keywords that are area
or location based. So plumbing. So we
have the data, but it wasn't at the
scale with this study to really publish
on it. So that's that's a starting
point. Um I I I do want to show this.
This does make me laugh. Um now you're
everybody on this call is going to be
like they're like this is crazy. Okay. I
don't know the last time any of you
printed the Map Quest map and put it on
the passenger seat of your car. Remember
we always used to do that like as you're
driving you'd have printed it because
there was no internet when you drove.
Okay. Map Quest is the number one cited
across those three models. Not breaking
them up. Remember, remember Google would
be higher, but Google doesn't site
itself. So, Google was the champion
because that's just two models, not
three. But Map Quest as a source, I know
people will laugh, but it's highly
structured. It's so easy to crawl. It It
has no fluff. It is literally like a
conveyor [clears throat] belt of
structured goodness. And if you go down
this list, so is the Chamber of
Commerce. So is Yellow Pages. So is BBB.
So is Ways. So is Yelp. It's not that
they're not showing up, but when you add
these I highlighted in blue a lot of the
people that have been partners of Dxed
or I think people it's just kind of
funny like I haven't thought of yellow
pages in years but the reality is is
they're showing up now the biggest most
common ones I think they're always going
to be the most common ones but to see
Uber Eats Map Quest at the top of
citations by look this is 7 million
citations a lot of our competition they
put out anecdotal research and I think
that's adorable but you can't look at
five clients
and be like, "We got the answer." I'm
not even telling you I have the answer.
I'm just telling you, holy crap, this is
fascinating how deep this goes. Now,
this is not showing you the websites.
The websites are a much bigger
percentage, but the tail on the
websites, we pulled in millions of
individual domains that are only cited
once. Once you get into all the
geographic, remember, if you're way out
in the countryside, uh, and and you're
Five Guys Pizza or you're a burger
joint, you're not competing with 20
burger joints. You're competing with a
Thai restaurant, a sushi restaurant.
There's only three burger joints out
there. So you have to stop thinking like
global and brand and start realizing
from the consumer's perspective what is
showing up and what is being cited. So
uh this is a this is healthcare and
finance also. So here's retail and food
same thing healthcare and finance.
Healthcare and finance I will offer
they're highly this is your money your
life type stuff. This is very regulated
and what's interesting about that is in
finance mapquest is again number one BBB
I think a lot of the models everyone
thinks it's like oh it's just
probabilities no it's not these are ma
model of experts that all of them are
they have logic trees and then they have
a safety framework around them so
already you're seeing that they're very
specifically not calling out some things
and they are calling out data
directories like like industry
directories like a healthgrades or a
webmd or a doctor that's good But can I
offer again just like Yelp if I'm
HealthGrades or any of these ones that
are aggregators where they're both
ratings and reviews, it's an existential
threat. They can't like I don't know if
you see remember the chart on LinkedIn
of WebMD falling off when ChatGpt
launched. Like it's it's serious, right?
So I don't know what their model is
going to do. I think if they don't sign
content deals, this is really
>> that's the thing. It's all agentic,
right? like it's all like like who's
going to win the being built into the
system, you know, agentic process which
we
>> Yes, you you got it. So, so kind of to
back up from the consumer perspective,
stop worrying about Reddit and
Wikipedia. I'm not saying now again I I
do follow everyone on LinkedIn. They're
like, "Oh, Reddit's back. Oh, Reddit's
gone. Oh, Reddit." Like, look, I don't
care. I I guess what I'd say is if a ton
of you for a complex purchase, if Reddit
is your first go-to, I think that's
awesome. It's a great It was a great
We'll see if it continues with AI
generated content, but it was a great
source, but generally speaking, it
doesn't show up if location is taken
into account. Reddit is really top of
consideration funnel. I don't think
that's bad, and if you have a Reddit
problem, you should solve it. But what
I'm saying is is generally it's not
going to play in as much, particularly
if they don't figure out their own
content strategy with getting paid. Um,
now all this research is at
yex.comressearch. And we write this
stuff up. We publish regularly. Um, and
I just want to say, uh, Garrett, that
we're not done with this report. So, I
did tease with Garrett. I'm like, look,
I got 57 million citations as of
yesterday. So, I have now three models,
57 million. So, three months later, what
does it look like? Because it the
database is just growing. Unfortunately,
I did not get through all of it, but to
give you an idea of the growth, the last
one was 5 million citations in the US.
It's 64 million now, right? So, like
we're growing really fast. Now, these
are I'm showing you all the citations,
the ones I've categorized. I'm working
on 42 million more, but you can already
see I'm picking up other areas. So, uh
to I forget was was it Adrian?
>> Adrian was asking this this other
category. I've now got 42 million
businesses. I have to break into the
categories, the subcategories, and we're
going to start making that available in
the research platform. So you'll be able
to actually go like I don't just want
lawyers, I want personal injury and this
really deep down, but I just needed more
data to I think statistically be
confident in the results. Um the initial
results I will also say look I think
location matters. I'm not saying if
you're a ecom brand this is the right
way to go but even most ecom brands want
to know how their brand is doing in a
place right there. It's not like
generally how's my brand? That's good.
But I want to know, do people in
Florida, do I show up for them even
though I'm free shipping and I'm based
in California? That's going to be this
next battleground of of not just where
the consumer is, but offering things on
your website that level the playing
field where the AI because remember the
AI is going to say to you, not here's 80
million results in search. It's gonna
say, well, I get you want a new
surfboard. I don't know why I'm shopping
for a surfboard, but I'm going to get
surfboard. And and it goes, are you okay
waiting a couple weeks? Because if you
are for what you're really aligned to,
which is an environmentally friendly,
it's a beginner board. Like I actually
think Almond Surfboards, but they're no,
they're in California. They're nowhere
near you in Florida, but if you can
wait, their website says all the things
you're looking for and they'll do free
shipping. That is this balance between
the offer that's going to come through
by doing structured data properly. But
the funnel as we know it is moving it
and it's moving off of the site into
these chats and these discussions. Uh
number two, websites websites in the big
data set. I can already see websites are
expanding their citation base. I think
that's partially tied to the uh
bandwidth of crawl and compute at the
models. They're pulling in more
websites, much broader source. So the
website is gaining ground. The
directories are kind of just getting
smaller because the websites are getting
bigger. But uh social interviews are
getting smaller. So from when I ran it
three months ago, they're smaller than
they were. And that's not just websites.
It's that I think again many of these
sites are starting to block the
crawlers. They do not want to let that
content out. And the reason why they
showed up more before is because they
were in the training data. They didn't
block the stuff while it was being
trained on. Now they're blocking it.
You're seeing it decline. Um and then
the last thing, online profiles,
directories, verification. Think of
directories or any thirdparty site where
you can have your data highly structured
that matches the website. That's that
goal of verification and different data
points you can get more verified. That's
absolutely the way probabilities work.
The more I like if I'm with Garrett and
uh you know we're back in Jersey
somewhere and we ask 37 people in in the
bus station of like hey where should we
get a a slice of pizza? And they go all
37 said go here. you got to go to Good
Fellas. Here's the here's the location.
He and I would be very confident from a
probability perspective that we have
independently verified that that's where
we're going. That is how this works. And
so I think you can optimize for that
very aggressively.
>> I got to ask real quick um to touch on
because I know this is a citation study,
but I do want to hear your thoughts on
mentions because obviously mentions have
a different sort of value, but they're
more valuable than maybe they were in
the past.
>> Yeah. Do you think that impacts the
equation like substantially?
>> Absolutely. Um it is it is amazing to
me. We're we're getting to this world
where um if you have mentions and you
can I mean the ideal scenario is you
have mentions that have a structured
data point that is important to the
customer journey. Now most brands and
businesses they know what's important,
right? So if you're a plumber, you know
that you really want to be known for
kitchen remodeling and not two in the
morning clogged toilets, right? that's
not that's not the business you want to
be in. So, if you really want that, then
the mention mentioning that it's the new
kitchen and the new house was done by
Pete's Plumbing, that's really important
because then you get this idea. Um, what
do we mean by mentions is um a citation
is actually a link to an originated
source that is not controlled. A mention
might be in an article or in a post or
it's not as not as solid. Yeah. A
review. it's it's more unstructured in
the AI's results. Um, and and Garrett,
we do track in that scout tool. We track
the order of the mentions as well. So, I
hesitate to call it rank, but if I ask
for a hotel in Jacksonville Beach, we
are storing the ordinal nature of who
did they mention first, who did they
mention second. It's not a real rank and
it's highly volatile, particularly by
memory, but we are keeping track just to
understand.
>> Yeah.
So, um, does that does that answer your
question, Paul? It's it's it's less
structured and not necessarily a a a
link out from the model. Um, but it
could be there could be a mention on
like I I also find a lot of crossborder
citations. So, I don't know if you've
all seen this, but when you ask like a
technical question or an SEO question,
you'll see like the India Times or a lot
of India based uh because they they're
incredibly prolific. And so, what you
see is a lot of crossnational. So, you
should be thinking about this not only
where the business is, but what what
regions is it being cited from. Um, so,
excellent. All right. Uh, Garrett, I I
think I'll turn it over to you and you
can walk us through or or questions.
>> No, I mean, it's so it's so good like
and and I can't reiterate enough to, you
know, check out the Yex research and
that there's so much more to pay
attention to over, you know, time being
a factor in all of this of we were
talking right before this webinar, you
know, launched. Um, the idea of like as
you get new models, as ChatGpt 5 is
rolled out, as Gemini 3 is rolled out,
as we see the weights fluctuate, and you
know, as as Christian was mentioning
like Reddit coming and disappearing and
coming and disappearing, like it'll be
interesting to see. I don't think the
dust ever settles at this point. Like
we're always saying like wait till the
dust settles. I don't think there's any
dust settling going forward. So, you
know, as an SEO, as a marketer, as a
business owner, unfortunately, there's
no none of this stuff is set it and and
and be good to go. Like, even with Yex,
like with the listings, you still need
to pay attention to any additional
content that you can add to your
listings that will increase the value of
whatever people are searching for.
That's my last point.
>> Yeah. And all those all all those
platforms, they're seeing sort of
resurgence in their relevance. They were
always they were always important from a
signal processing perspective. That
that's always been the exposition. You
got you should do again you should get
as much data exactly the same in real
time as you can because that's we know
that helps. We can prove that. But this
is different. This is almost like a
resurgence of they're looking at their
models going wait a second we might be a
really good source of information. What
if we added the menu? What if we added
the insurance provider? What if we added
So you're seeing this expansion of the
use of these tools and structured data.
And I think everyone look again it's not
a yex pitch. I'm saying you should
rethink
>> yes
>> data in probabilities instead of a
deterministic world where Google
controlled.
>> There you go. Thank you Christian for an
amazing enlightening interesting uh
webinar. Uh keep your eyes peeled for
more of their research. Check him out on
LinkedIn. Uh I'm I'm going to sign off.
We are going to go ahead and and wrap
this up and send you the replay and the
deck afterwards. But thank you all so
much for joining us today. This has been
awesome.
>> Absolutely. Thanks, Garrett.
AI visibility shaped by location and structured data. Brand-managed sources preferred. Review signals influence recommendations
Nikhil Kamath
