Copy the formatted transcript to paste into ChatGPT or Claude for analysis
Ranking number one, page one on Google
will not serve you well in AI search.
[music] The waiting of frequency is
higher than the waiting of ranking. What
these companies are doing, they're
writing prompts in this way to force AI
overviews, but then the underlying data
is absolutely incorrect. You have
written about how 32.5%
of the AI citations are list.
>> Mystical work phenomenally well like
especially if you're focusing on bottom
of funnel uh search terms. Mystical are
like your 100% [music] best friend.
People run their sites behind
Cloudflare. By default, you're blocking
crawlers.
[music]
Welcome back to the Agency Insider Show.
I'm your host, Napit Koshell, founder of
Page Traffic. And today we have an
episode that could fundamentally change
how you think about search visibility in
2026 and beyond. Today's guest is James
Barry, CEO and founder of LLM Revs, an
AI search analytic platform focused on
helping brand understanding and track
their visibilities inside large language
models like charglexity and of course
various others. Now James has been
working at the intersections of SEO data
and AI not just talking about AI SEO
conceptually but actually building
tools. That's why we brought him on our
show just for you guys. In this episode
we're going to talk about AI search
reality changes for SEO. why traditional
ranking rank tracking tools doesn't cut
anymore and a lot more. James, welcome
to the Agency Insider Show.
>> Well, well, thank you so much for having
me. Really excited to be here and and
really excited to chat. AI SEO. That's
it. Or or that's I suppose that's that's
what we're calling it. Are we calling it
AI SEO? We calling it AEO GEO. Like what
what's your favorite term?
[clears throat]
>> My favorite term as of now is uh the AIA
uh SEO. I would prefer it because I I
think it makes sense but you know I
don't know AI SEO I I like LLM also but
I I think search everywhere optimization
should include everywhere but you know
everybody has their own terms so Jio and
AI SEO seems to be the favorite.
>> They do seem to be the favorite. I think
I saw that uh we did we did a bit of
digging through uh like Twitter,
Facebook, LinkedIn and uh Reddit as well
and just sort of uh accounted every
single mention of GEO, AO, AIO, LLMO as
well and a few different other
variations um in context talking about
the same topic that we were talking
about and it did come out that AI SEO
was the most commonly used term. I think
what I see from like being heavily
involved in Twitter is that AEO is on a
rise. GEO is like very heavily used, but
when you move out of like the world of
chatting to people who live their lives
on Twitter talking about what's the
latest update and all of that, AI SEO
seems to be the go-to term. So that's
what we're we're trying to um align
ourselves with.
>> Okay. uh now before we talk about uh the
SEO I want to start somewhere of course
most interview probably don't your
actual journey because you're seems your
background is fascinating you are not
just an SEO guy who stumbled upon the AI
I mean you are serious engineer with
open-source credentials correct me if
I'm wrong
>> yeah yeah that's that's exactly it so
I'm personally a software dev by by
background um built lots of different
projects over the years uh but have have
basically worked in growth, software
development and growth for the past
eight years or so. Um, and specifically
around organic growth, building free
tools to bring traffic and and you know,
maybe we can we can dig into it on a bit
further of my story if you're
interested. Um, but yeah, found a
massive opportunity uh when Chat GPT
first came out to um yeah, to capitalize
on on that traffic and I guess LLM refs
was born out the back of it. So, yeah.
So, So what was the first moment where
you realized that AI search wasn't just
a feature but a fundamental shift?
>> I guess it's a great great question. Um,
look, I think I remember chatting to my
mom and I know this is the classic,
everybody will say this, but she was
trying to book a holiday somewhere and
uh she she was on Perplexity and she'd
somehow got some hotel recommendations
of Perplexity and I think this was like
really early on like I knew of
Perplexity but maybe I hadn't even used
it at this point and I was like what is
this? like how how is my mom like she's
like I want sun I want like you know to
be this far away from the beach find me
a location find me a place to stay and
and she ended up booking straight off
this search I think she did a couple of
searches um so for me that was like a
real eye openening moment look there's
there's not only like traffic
opportunity here but like the conversion
rate on that will be significantly
higher than perusing hundreds of
different sites and then picking form,
>> right? So, so then uh so then what what
I mean what happened next? So, do you
decided to dive into it right away or
how does LLM Ref was born?
>> Sure. Yeah. So, so the story of of
myself. So, I worked with a couple of
friends of mine uh great founders Tim
and Sabah of a company called V.io. Um,
and they've been one of the fastest
growing companies in terms of pure play
organic SEO growth. Um, they're online
video editor moving into the space of
AI. It's a fantastic company, a
fantastic product. Um, and they did uh
nearly no paid advertising. It was all
through building free tools, building
landing pages that ask answered user
search intent. Um, and what we found is
that when chat GPT first started coming
out, we were able to look at the traffic
that was coming through and it was it
was fairly substantial. Um, outside of
that as well, we had a play with um, a
custom GPT that did really really well
when the GPT store came out, which I
know was a little bit of a flop overall.
Uh but we found a lot of success from
that and and brought in I'm not going to
share their figures but but many
millions of uh of users via that that
GPT just um getting getting traffic
through to to the V site. Yeah. [snorts]
>> Okay. And that this was in which year?
>> This was uh last year. Um this so uh and
I think it's God I can't even remember
the the years now. Every this this year
has moved so fast. So I think we
launched uh last year but I think when
um because to start with that GPT store
was only for paid users. So conversion
rates were really good but the numbers
were lower. Uh but when the GPT store
opened up for for every user then
suddenly there became a huge opportunity
for us. I think um I don't know if like
how much you're following the current
chat GPT new app store that I think got
released today uh as to today as we're
recording this as well. uh potentially
that's going to be a huge opportunity as
well for these businesses.
>> So when you built your custom GPT which
said brought you what exactly does GPT
was doing?
>> Yeah. So this was the difficulty with
GPTs, right? Because effectively they
were like just custom prompt rappers. Uh
you know if if you saw a lot of them
they were like copywriting tools. You'd
paste your your thing and it would align
it with your brand guidelines. uh but
they had this functionality to to add
actions which was effectively an open
API spec that you would add to your GPT
and it would be able to uh make
auxiliary calls out outside of uh open
AI outside of chat GPT. Uh and what we
had was a tool uh called V text and
video which was one of the first tools
that allowed you to go from idea or
prompt to full 30 second video that's
ready to share on social. And that's
what we built, right? It was you would
get jump into the chat GPT experience.
You would do all of your research. You
would write your script and you'd say,
"Yep, I want to make a video out of
this." It would get sent off to Vid and
and we would create that video for you.
So it was a really like wow wow moment.
um back back you know a year ago. Yeah.
>> Okay. Now uh of course you have worked
closely with technical SEO and data. Now
we uh do you think that the background
is essential for understanding AI SEO?
>> I think so. Let me ask you a question.
Are you digging into like to understand
it to you know sell to a client or to
understand it to found the company or or
or just
>> both? So I because most of my audience
is the agency owners or people who are
working in agencies that's what the
agency show is about. So I'm just
thinking when people think AI SEO is it
more technical which goes or is it
something different than what SEO is? I
mean of course I know the answer but
just for my audience at large what do
you think uh that background any sort of
technical SEO or data background is
essential to understand the uh AISU
[snorts]
>> yeah look I think there's with AI SEO
there's a whole load of new concepts
right but there's also a lot of existing
concepts that still lay true this whole
thing of like SEO is dead I don't buy
into this at all. Right. Um, so I think
that there's a there are a few different
concepts like the idea of tracking your
traffic now is like very very different.
You know, you're now tracking server
crawler logs as opposed to tracking what
is the referer that that somebody's like
GA calls when it when they land on my
web page. So I think there's a few like
concepts there that change. Um I also
think there's like a change in how you
track visibility as well is like one of
the most
I mean full of misinformation and uh you
know maybe over over complexities
where required. Um but I think you know
one of the things that we're trying to
do just to summarize all of that in one
of the things that we're trying to do at
LLM refs is make AI search really really
easy. Look, we've had [clears throat]
SEOs working with Google search get you
know typing in their keyword. You you do
your two to three word keyword and you
get a result and we've been working with
this for like 20 years. You click the
first blue link, if not you go to the
second one, etc. Right? Everybody knows
how to do this. Now in AI search, you
know, things have changed. Suddenly
search is conversational. It's way more
heavily personalized. But what we're
doing is we're trying to keep the same
concept of keywords instead of prompts
across our platform. So that's when you
come to LLM refs, it's our
differentiation with the rest of uh the
tools out there of which there are
hundreds now. Um you you work with
keywords. So that makes it really really
easy for for marketing teams to get
started
off the ground.
>> Okay. Now, now uh
so uh L&M refs has been intentionally
bootstrapped when competitors like
profound PKI have raised venture
capitals. Now I I am sure that's a
deliberate choice. So so what's the
strategic thinking behind being staying
bootstrapped and what does that mean for
how you are building this product?
Yeah, look, I I chat to lots of people
about this. Um, my my background is is
working in bootstrap companies. I think
that's the best way to build a
sustainable company. I think if you look
at, you know,
what what's what's the best, you know,
in my eyes, what's the best SEO company
out there? You know, in my eyes, it's
Hrefs. And Hrefs have done hund00
million in revenue a year. completely
bootstrapped as well. Uh they've created
this perfect model of of creating like a
lot of educational content, putting us
so much value out there that they've
just established themselves in my eyes
as as the leader in in the SEO space and
haven't needed to raise extra cash. The
other thing that I would say like the
the cash the cash can either help you
hire more, develop the product faster or
scale your you know your customer
acquisition,
>> right? Or I guess the other one is like
lower all of your costs. If you look at
LLM refs versus a lot of the other tools
out there, you know, the last time I
checked, you know, how we benchmark
ourselves in terms of our rank tracking
product, this is is on AI responses
analyzed. So as I said, we track
keywords, profound peak, whoever are
tracking prompts. But if you look at the
underlying thing that comes across all
of us is AI responses analyze. You know,
last time I looked, I even trying to
just find this little spreadsheet. I
think we came out as like 18 Yeah. 18
times cheaper than profound, 11 times
cheaper than than Pete, which tells me
that their venture capital fund is going
straight into Google ads, going straight
into, you know, hiring sales consultants
to go and close massive deals when
actually the target audience we're going
after are probably agencies more than
anything else, although we work with a
lot of in-house brands who are actually
[music]
looking for like great data quality
without having to commit to massive
spend.
Okay. So, so uh obviously the next
question, how do you compete against
wellfunded competitors while staying
lean? I mean marketing of course all of
this has I mean what is your uh customer
acquisition strategy?
>> Well, I guess it's a great question and
this is like an everchanging
[clears throat] strategy. I suppose
what's working really well for us at the
moment is just being out there
publishing educational content and being
trying to be at the forefront of like
understanding what AI search is. Um AI
search is changing all the time. Like
literally all the time the grounding
sources are changing. It's like it's the
most volatile search has ever been. and
we're trying to be at the front of that
and be one of the voices that can
explain it and sort of understand when
things change, why they change, and how
you can stay on top of it. Um, so I
think that's like our our core movement
is just being in touch with the people
who who know what they're talking about.
Um, and yeah, just just sort of getting
our name out there.
>> So So how many paid subscribers you have
if you wish some idea?
>> That's it's not something that we
publicly share. Um, I'm sure I could do
it off uh off camera, but it's not
something we we share publicly.
>> Okay. But but you are profitable.
>> We're profitable. We've got a team.
We've got three full-time. Uh, so we're
we're going well.
>> Okay. Now, most of the time you saw a
lot [clears throat] of studies yesterday
also came about that there is a
difference in responses when it comes to
v API and why while normal search. So
what exactly is LLM? Is it based on the
API calls making or how?
>> It is absolutely not based on API calls.
Um, so yeah, so I think most of most of
the good ones out there now are are all
based on real UI responses. That's
exactly the same with us across the
board. We track 10 different AI search
engines. I think that's like on the the
top level of what people are tracking at
this moment in time. And it's all based
on real browser automation of real
conversations. I think there was a thing
where there's like an 18% overlap
between what you're tracking in the API
and the UIs that specifically in chat
GPT. So it's the data quality is is
absolutely dire. The other thing that I
just want to be really clear on as well
is data quality is everything for us.
It's it's why our clients like us. There
are a few funny things I also see
constantly around Twitter where people
look into their um Google search console
and they see these weird queries which
is like give me an overview for and then
it looks like a prompt like a longtail
prompt. what these companies are doing
some of them and you know not naming
names uh but they're in order to like
save themselves costs they're writing
prompts in this way to force AI
overviews uh because you know when you
go to a Google search it doesn't always
give you an AI overview but you can sort
of force them in this way but then the
underlying data is absolutely incorrect
so we are crawling continuously to see
if we can find an AI overview that is
triggered you know and and fronting all
of the costs that it takes to do such a
thing. Not any of these crazy
hacks. [clears throat] You know, data
quality is key here.
>> Okay. Uh now, now how do you technically
approach tracking brand presence across
tools like Charglexity or Gemini?
>> Yeah, sure. So, one of the things that
we do differently again um or maybe not
differently, you know, there's hundreds
of different tools here, but I've tried
most of the tools, right? and and how
often have you logged into the tool,
you've set up your domain and it tells
you you have 80% brand visibility. And
then I go to a different tool and it
goes you've got 30%. And then I go to
another one, it goes you've got like
256. It just gives you like some magic
arbitrary
number. And I I don't know where it's
come from. I don't trust it because
they've all got different numbers. It
just doesn't make sense to me. So with
us, we will never give you a brand
visibility like magic number, right? How
we do our rankings is we do it purely on
competitor benchmarking, right? So what
we do is we look at the responses. We
analyze everybody who's been mentioned
and we measure them on two different
metrics. Now the first one of those is
your share of voice. So this is you can
be looked at as like a mention rate. So
when I look at a a prompt or a keyword
which when we talk maybe we should dig
into keywords in a second but when I
talk about keyword I'm talking about a
set like a a group of prompts um when I
look at that keyword I can say with
absolute confidence that for this search
engine for this keyword topic this brand
will be mentioned in 27% of responses or
whatever it might be. The second one we
look at is your position in in search
which is like the average order that
your brand appears in over time. Now
that is less important. So how we do is
we weight these two things together and
we create this this hidden score that's
then able to benchmark yourself against
your competitors.
>> Okay. [clears throat] And uh when you
are doing all this so so so so I
understand
I understand you said that how your tool
works is it takes keywords in prompts
because I've seen most of the tool and I
feel very odd and correct me if I'm
wrong
>> even before you get into the prompts
they give you the competitors and I I
find it very odd because I believe the
competitors are not what I think
competitors are what which are occurring
in prompts so I think so when when
somebody says the competition first
prompt later. I think that's not how it
should work. So, what is your take on
that?
>> I I completely agree. Like, how often
have you seen a brand do amazingly well
in AI search that isn't even ranking on
the first page of Google? It happens all
the time, right? Like I completely
agree. It should not be competitive
first. That's not how we we set up our
prompts. So, in LLM refs, as I said,
when you come to our tool, you set up
keywords, not prompts, right? And you
can think of this the same way you think
of your core keywords in SEO. So, you
know, a lot of our clients will come in,
set up their account, and they can
import them straight in from Google
Search Console, or you can bring it over
from your keyword lists in in whatever
other traditional you have.
>> Yeah. Cool. And and what we do there is
we then generate the prompts for you
based on real human conversations with
AI chat bots, right?
>> Okay.
>> Does that does that make sense?
>> Yes. Yes. Yes. Yes. And and and second
thing is when when people talk about
sentiments, I think it's an overblown
thing. I mean again I might be wrong but
I don't see negative sentiments as a
thing because how would you have a
negative sentiment when I'm asking about
what is the top I mean charge any LLM is
not going to give me answers and say
okay these are the best companies but
let me give you the worst as well. So,
so I doesn't even make sense to have
tracking as a negative sentiment. I I
see sentiments as something which is
quietly quite overrated. What is your
take on that? Again, I I feel like we're
very aligned here in plenty of things.
Uh agree that sentiment is not something
it's on our road map, but it's not
something that we're prioritizing. What
I think is more important is
misinformation and factchecking, right?
Like with sentiment in AI responses,
you're either going to be spoken about
overwhelmingly positively because
there's some sort of personalization
getting involved where it knows you work
at the brand already or something like
that. Or it's going to be very neutral.
Like it's very unlikely unless you've
specifically written a prompt in a way
that it comes through and it says the
worst podcast is the agency insider
show. It's it's just going to it's it's
just going to be very neutral about it.
On the flip side, there is a big issue
with misinformation, right? Like if I
was asking, find me a podcast that is,
you know, an hour long or something like
that, uh, about SEO for agencies, then
if it did not bring you up because it
believed that your brand was, you know,
only, you know, producing 10-minute
clips or something YouTube,
>> then that's an issue. So, this is what
we're prioritizing is tracking where are
things going wrong. And like a good
example is like a tire company I've
worked with recently. Uh that would, you
know, they were told their tires were
for motorcycles or something like that
when it when this new tire was made for
a truck. It it was like wildly wrong and
it was they've unpicked it now. But that
does that is a more damaging thing that
we should be focusing on over sentiment.
>> Okay. Now uh so from what you you are
seeing in real data so what does LLM
visibility
inside what does visibility inside LLM
actually look like? Is it just the
citations mentions recommendations or
something else entirely?
>> So I think you know it it goes into a
few different angles here. I like we
measure visibility in two ways. So, so
firstly is brands getting mentioned in
the c in the uh response and the second
one is having a citation which is
actually like a a clickable link right
what's less important is if you can get
your content in the sources but not
cited
>> like okay you really either want to have
a link or you want to have a an actual
mention in there. So they are the core
things that that we we also can tell you
whether you're getting sourced because
if you're if you're already getting
sourced then you've got a decent
opportunity to become a citation but I
wouldn't say that on its own is a is a
valuable um like thing you should be
measuring.
>> Okay. Now one thing I also find
interesting is that LLM visibility isn't
static. So so the thing is the same
prompt can return different branch over
different time. So how does so how does
LLM refle this variability or is I mean
or how do you go about fix fixing this
problem?
>> Yeah. So it's a really great question.
LLMs are probabilistic and
non-deterministic. You ask the same
question twice. They give you different
answers both times. Um and look, it's a
it's a major issue of a lot of the tools
that I try. Uh you know, you've got this
approach where you just rerun the same
prompt over and over and over. um which
is is part of our method but what we do
seeing as our method is based around
keywords if I can give you say 10
prompts underneath each keyword I'm able
to then analyze those across you know
whether it's chat GPT perplexity
whatever it might be and then only rerun
the ones that I need to remove any noise
so that we can find statistical
significance for your brand's AI
visibility
>> okay so so what does that mean I mean uh
so for For a brand what does exactly it
means if they have to track their
mentions I mean because they are
changing. So how does a brand should
should they be? So my question is you
you you track rankings you can track
rankings daily they shift but but here
what is happening is the the mentions
can change and they can change every few
minutes. So how do brand what does brand
take from this?
>> Yeah for sure. So, I think well, the the
key thing is I know this is a a horrible
plug for for myself and all of my
competitors, but you're going to need a
third party tool here to help you. You
know, you can't just go straight into
chat and run this hundreds of times.
There are tools that are specifically
built to to solve this problem for you,
which is running these things multiple
times to understand what are the
mentions, what are the citations, and
where can you find opportunities. Now, I
think the core thing here as well to
look at is you should be tracking a
broad range of prompts. Um,
specifically, you should be tracking
prompts that are basically saying
exactly the same thing in different ways
because they can have wildly different
answers as well. Another example of
somebody I worked with recently was is
in the the in building a link building
agency uh and they had come from one of
our our competitors, one of the big ones
that we've already mentioned today. Uh
and they said, "Oh, I have this amazing
search visibility and they came to our
tool and because we generate prompts
based on real user conversations, they
said something's not right. you know
these my visibility is far lower in your
tool than it is uh over in this other
tool and the reason is that they were
searching for the word I think it was
like backlink agency or something like
that and actually real user intent was
searching for link building agency and
their visibility I think they were in
second for backlink agency and then 16th
in link building agency. So getting a
broad range of prompts that are being
asked by real users in AI search would
be my recommendation. Uh but beyond if
you can't achieve that then just
quantity quantity quantity because the
more you're searching the better data
quality you're going to get out of that
as well. And and when people talk about
different stages tofu, mofu, bofu, which
one do you because I know the tofu will
not generate anything. So, so for for
our clients when they comes to
visibility, I say tofu is not going to
give you anything. You're just asking
question. They might not even they might
be grounded queries. So, what is your
take on this? Which type of which type
of a funnel generates the is good to
track and is is for good for brands to
follow?
>> I I think to be honest, I'm actually not
going to make the call on any brand's
behalf. I think it's when you start
making calls like that. Um,
look, I'm going to be wrong for for some
industries. [clears throat] It depends
on your industry. It depends on the
keyword or or you know sector that
you're going after. What I will say is
that my background is in revenue
revenuedriven SEO. So they are the
keywords. They are the they are the
search terms that I personally am going
after. And that's where I think you know
AI search excels as well. You know we
know that when people manage to find
that mention or get that click that
their conversion rates are going to be a
lot higher from at least what we've
seen. Um so that's bottom of funnel is
is where I'm investing. But again this
this lot strategies. Yeah.
>> Right.
>> Okay. Now let's shift gear to the
practical side. Our audience as I said
is agency owners and marketers who are
serving clients right now. They have
heard the high level stories about AI
search but they need to know how do I
actually do this tomorrow?
So, so, so if an SEO agency wants to add
AI visibility optimization as as a new
service line tomorrow, how would you
recommend they structure it and what
would the deliverables look like and how
should they price it relative to the
traditional SEO service?
>> Again, I I don't run an SEO agency, so I
don't think I'm the best person to
answer that. I think it's down to you.
Again, it's down to your clients. It's
down to, you know, what you think works
best works best for you. I have had a
conversation multiple times and maybe
I'm going to go on to talk about another
question that we might have in a second
about what would be like how would you
sell this thing, right? Let's say you go
to a client and you say I've now got
this how to sell. I say what should be
the deliverables look like?
>> Sure. In terms of deliverables, I think
you know the key thing is revenue,
right? How can you attribute revenue to
search? And I think there's there's a
really hard missing piece at the moment
with uh tracking where you can watch
server blogs and you can say right I got
a citation here or I got sourced here at
least which means you know then there
was a bit of traffic that came through
but there's no direct like UTM
parameters or referers that you can
follow or anything like that. What we
actually see is when brands start
investing heavily in their uh GEO that
they get this hidden brand dip uplift
from just Google organic search. You
know if your company if you're brand B
and you're you're just chilling and then
you start investing in like how can I
optimize myself for for specifically
chatting what you'll find is your brand
B direct search in Google will will
lift. So my answer would be figuring out
whatever that is and somehow mapping
back having a system for doing this.
>> Okay. And uh of course the reason why
I'm asking is that traditional SEO has
been all ranking position and optimizing
content building back links. Now a
search doesn't do most of these things.
So how should agencies rethink reporting
when they start using tools like LLM Rev
and what KPIs actually matters now when
it comes to the LLMs?
I think the the KP like the core KPIs
are share of voice, ranking, citations,
right? Like where are the citations
going to? Are they going to me? Are they
going to my high value pages? So, I
think they're what you should be
tracking. Um,
I think in in terms of the rest,
>> not being a an agency founder, I think
it would be unfair for me to to make
decisions on people. I don't want people
to make decisions based on my my
experience, which isn't in in running
agencies, you know, I I build analytics
software for for AI search engines.
Yeah.
>> Okay. And uh [clears throat]
now so let's let's talk about LLM's
understand brand. So what signals do you
believe matter most for shaping AI
perception?
>> So for shaping AI perception what
signals ma matter the most? Well I guess
it's a you know it's it's a great
question. So
well can you can you break that down for
me specifically what what we what we're
asking?
>> So so my question is that how how does
for LLM understands brand in terms of
how important is consistency entity
clarity how how does what exactly
matters when you have to shape a brand
perception in AI?
>> Sure. [snorts] Okay cool. I guess
>> yeah that's that's a great question.
Look, we know now that AI search is a
game of two halves, right? So your your
first half is can you get into the
actual underlying models training data
so that when the next version of Gemini
or or whatever it is is going to be
trained you can ask it completely
offline just via an API or something you
can say what is brand X and it will tell
you all about it and it will speak as
positively as it as it possibly can
about the brand including all of the
features you've got you know etc etc. So
that's the that's the first half, right?
The second half is now what we're
learning around search grounding. So
this is now when you're using chat GPT,
Gemini, whatever it is, complexity,
they're not only just making a search
into their own brain, they're actually
making these search, they're doing these
search grounding to get true
information, live information from the
web. And how that works is actually you
take the model itself and you say, can
you break down this big old seven
sentence prompt into one to five
different fanout queries is what we call
them. And these are like uh longtail
traditional SEO keywords, right? So
let's say it's got f I've got my seven
sentence prompt. It breaks it down into
five different sub queries. Each of
those queries is then going to be sent
off to some sort of search grounding
service. So, traditionally for Google,
obviously that was, you know, going to
be Google themselves. With Chat GPT,
it's it's a it seems to be a little bit
of a mixture depending on what account
and market you're based in between Bing
and Google. Perplexity seem to have
their own search index. And to be
honest, we've seen recently that Chat
GPT have got their own index as well.
So, what that does is it will return a
whole series of results. So that would
be, you know, say a hundred results from
Google for each search query. So now
we've got 500 results with a title, a
slug, and a meta description. And then
the model can choose right out of this
list, if I need more information, which
of these pages do I want to look into to
then understand more about about the
problem. So, this is where you're going
to find that actually there's a whole
load of pages that are getting cited
very heavily that aren't making it onto
page one of Google just because that
they're in that set and then a model is
is more heavily weighting them. And
again, that then leans back into how are
you getting into the original training
data set so that you are the trustworthy
source that that it picks up. Right? So,
these are these are the core things that
shape it. We've also seen recent
experiments from chatbt around wiki data
being a fantastic source for search
grounding as well. Especially I'm seeing
is if your brand name is like a a
non-generic
word and it's a it's the underlying
phrase of the service that you're
offering AI because it's this semantic
search it really struggles to pull out
whether that's a brand name or whether
that's like just repeating phrase. So it
uses wiki data the underlying wiki data
entry to be able to pull out what
exactly is a brand and and what is not
as well.
>> Okay.
>> Sorry that does that answer your
question there? Is there anything we
want to dig into?
>> I have a whole segment on this. So I'm
going to dig more into it now.
>> Okay. Cool.
>> So so my question next question is how
important are consistency entity clarity
and factual reinforcement across the
web?
>> Absolutely mega mega important. So how
important is clarity? Clarity is
everything. If we look at the most
recent uh I I believe it was a Prasana
Madhaven who's one of the principal PMs
uh at at Microsoft who obviously who run
Bing. He wrote a whole AI optimization
guidelines masterass and we've
summarized it down into a we've got a
blog post summarizing down sort of the
key points here. And one of the key
things here is absolute clarity um and
not overmbellishing the facts that you
have about your brand. It's
untrustworthy to AI for you to say I am
the leading X Y and Zed. What's more
trustworthy is to say I have had X years
experience in X Y and Zed and worked
with clients like whoever. And that
builds your, you know, it goes back to
the whole traditional Google EAT stuff
that builds your your experience, your
authority and your trust and uh it's
it's more likely to get cited in AI
search.
>> Let's talk about how we can further
influence these LLM models. So what role
do PR, citations, Reddit and community
content plays in training or influencing
these AI models?
>> Yeah, look, I think there's there's two
parts to that. The first one is is
obviously getting into training data. If
you're cited by like
publishers with with high authority and
high trust, it's going to, you know,
weight your brand more heavily compared
to your competitors. The second part is
is that actually we go back to those um
that search grounding process of now
we've got 500 different results because
of you roughly speaking now because the
model is actually looking at all of
these different results to then try and
find an answer. the more of those
results that your brand is appearing in,
the more trustworthy it is to the
answer. So this is where I say to people
that actually ranking in number one like
ranking number one, page one in Google
and not having any other one of the
results talking about your brand will
not serve you well in AI search in the
same way it would traditional search.
What would work better is having page
having number two, number three, number
10, number 15, 50 all talking about you,
which will then lift you to the top. The
waiting of frequency is higher than the
waiting of ranking in in uh search
grounding.
>> Yeah, I agree on that part. Now, do do
you think brand will eventually optimize
content specifically for LLM consumption
or are they doing it right now?
>> I can tell you they're doing it already.
They're doing it already. Yeah, I think
it's it's already started. I think that,
you know, the key thing is, you know,
you're still writing for humans. If
you're getting humans onto that page,
they're the only ones who are paying
you, right? Unless Cloudflare have their
way, AI is is not is not purchasing your
content. Um, so I think that's like the
key thing. But the second thing is, you
know,
we're shifting away from a world where a
user chooses what they they view. Like
if you think of um, you know, the Tik
Tok algorithm, right? I'm s I'm just
served content without even having any
choice in it at all. And it's just
learning my preference and personalizing
my feed over time. And I think
in a potentially sad way what we need to
do now as SEOs is optimize our content
for those algorithms as well which is
which is just effectively in our world
chat GBT perplexity and all of these
these other LLMs.
>> Okay. Now [clears throat]
so so so you have written about how
32.5%
of the AI citations are listicles.
>> Listicles work phenomenally well like
when especially so I guess the other
thing to look at here is like if if
you're focusing on bottom of funnel uh
search terms list mysticals are like
your 100% best friend. They're really
really great for AI search because the
first thing is they're super easy to
scan. Like if you know if somebody's
searching what is the best whatever it
goes you know does the search grounding
it finds this page will tell you what
exactly the best ones are. It opens the
page it goes H2 number one is this. H2
number two is this. It's it's really
easy for AI to read. And I wonder this
is sort of I'll throw this back at you
to see if you think the same if these
will be um what's the right word uh if
these will be like maybe not penalized
but the waiting of listicles will be
reduced inside AI search by the by open
AI by Google uh as time goes on because
these seem to be a real hack at the
moment in time.
>> I I agree with you. I will a personal
confession we have done it for our brand
quite a lot and it works but I believe
uh we are just running on a borrow time
on this maybe I think I will give it
maybe three or maximum six months when
this this thing will stop working but
>> I think
>> I think it's interesting that like
Google have so much control over search
uh and I think it's I I haven't seen
open AI have their moment Yeah, where
they've done their first helpful content
update. Uh, so I think they're they're
still figuring things out, but the
moment that they're immediately in a
pressure because of Gemini 3, you know
how much they are in a pressure. So I
don't think I think that will be the
last thing in their mind as of now
because it's just
>> true.
>> Yeah. Yeah, I think it is true. I I
wouldn't bet against open AI though. Um,
that's that's my I could be wrong. We'll
see in six months time. But but I I I
think you know especially with consumer
level
>> yeah like um users. I think it's really
hard to shake the default like it
happens right like Yahoo was once big
and then Google took over. Um but I do
think it's going to take a lot of
beating for for Open AI to to lose their
their monopoly at this point.
Now you specifically when you did that
uh uh 32.5% of AI citation my question
is that a follow-up question is is there
a content audit framework you use to
identify the highest opportunity pages
to optimize first if we have to or you
don't have anything like that
>> so we don't have automated tools so the
core thing in terms of LLM reps is is
what we do is
>> we give you the signals and signposts to
build your own effective AI search
strategy. So what I have seen in other
tools as well is when they give you like
recommendations do this and this will
happen because AI search is so new
things are constantly changing and
there's no real you know there are
things that work but there's no real
playbook yet. I could be right, but I'm
probably only going to be right
>> one out of a hundred times or or, you
know, one out of 30 40 times. At which
point, if if it doesn't if what I tell
you to do doesn't work, you're not going
to trust my platform and you're going to
go elsewhere. So, what I believe is
SEOs, we are smart people. We can figure
it out. Every industry is different.
Every keyword is different. Every
strategy is very different but if we can
give you all of the data to make the
correct decisions that that is where the
value ad comes from. Now in terms of
specifically to answer your question
around like content auditing I is to
look at what is already ranking in AI
search. So for your key phrase your
topic keyword prompt whatever however
you want to do it. um go and look at
what are what's what are the citations
that are there? What are my competitors
like? What content is my competitor
already doing really well with and go
and read that content and go and copy
that framework. Go and learn something
from them like you know maybe not
directly copy it verbatim one for one
for one but take inspiration from it.
Look at what works really well for them
and look at how you can build a piece of
content that is better than what they've
already produced.
>> Okay. Now of course UTRA 10 plus AI
search engine chat perplexity Google AI
overview claw Gemini etc each has
different biases for an agency with
limited resources which platform should
they prioritize first and how different
does the optimization strategy needs to
be for each?
>> It's a great question. I think you know
when I chat to everybody I think there
are really only two that that come up
time and time again like on on the grand
scheme of things. It is chat GPT because
that probably in terms of the pure play
LLMs it's it's over 80% of market share
and the second one is Google AI
overviews which is just you know that
probably dwarfs everything else because
they're everywhere now. So I think
they're the the two key ones you want to
be looking into. Um, that being said,
there are clients who are really hot on
on Perplexity as well. Depending on what
industry you're in, conversion rates
through Perplexity can be phenomenal,
especially for SAS plays. So, if you're
in that world, then take a look.
>> Okay, now let's get a little technical
for a moment. Can you walk us through
the must-h have technical setup? LLM.txt
controversial imple, but I agree there
is no harm downside in implementing it.
crawability consideration, schema
markup. So what's the priority order and
what mistakes do you see most often?
>> I think like we're effectively getting
straight into this is where AI SEO and
SEO are in bed together quite literally,
right? Um, so I think you know the key
thing is LLM's text I recommend setting
it up we you know it's not a priority
but it's it you know when you've done
everything else you've already audited
all of your content you're already doing
fantastically spend an hour or two and
go and get one of them set up and and
with us you can automate this as well.
Uh but just do not prioritize LLM's text
and think if I create one of these my
visibility is going to go through the
roof because you know maybe that will
happen in the future at this point it's
it's just not just not important. So if
I'm thinking of like technical SEO setup
I think yeah you you've sort of outlined
the the key things here. Number one, for
AI search, you need to make sure your
content is crawable, right? Like, right,
you there is a number of companies that
I've like chatted to who uh we've said,
"Oh, your your content isn't crawable."
And they go, "That can't be right
because a recent client said they found
this via chat GPT." Now,
>> that may be true, but what the issue is
is they found them via a third party
article or a third party source or
something like that. You can never get
your own content cited if you're
blocking AI AI crawlers. And the first
thing to do is if you're a Cloudflare
user, go and check out whether you've
got there's two tabs uh to allow robots
txt to not block crawlers and then just
to allow crawlers as well. We recommend
turning those off. Um you may if you're
a large publisher, you may make another
decision and and that's totally fine.
Um, so that's yeah, so crawability and
we've got free tools to check the if
your site is crawable as well and what
AI can understand. I think the second
one is serverside rendering. You know,
with Google they've done, you know, part
of their moat really is the fact that
they're able to crawl a page, run the
JavaScript, get all of the content out
of it, and and add it into their search
index. In AI search, that does not
happen at all. Even if we're using the
search grounding, by the time that the
page the the AI crawler comes back to
crawl that page in real time and pull
data out of the out of the page content,
they're just not going to see anything.
So, you need to be doing server side
rendering. If you're not, I mean, all
the WordPress sites and all of that will
already be set up there, but just one to
keep in mind. Um, and then I guess your
final one as well is getting your schema
markups lined up. It's a traditional SEO
strategy. You should be doing it anyway.
Uh, but if you've not, that's super
super highly important. Especially
getting FAQ schemas set up is what I
find work really well because when
people are typing verbatim questions and
those queries get split up, the more FAQ
schemas you've got in there, the more
chance you are to get verbatim a chunk
of text lifted straight into an AI
overview or something like that.
>> Okay. Now, you also have built an AI
crawability checker tool with LLM prep.
So what are the most common issues it
surfaces that the site site owners don't
don't realize are blocking AI
visibility?
>> I 100% the most common one and I I get
people who email me about this every day
going um oh something something's wrong
uh it says it can't crawl it. It's it's
the Cloudflare block. People run their
sites behind Cloudflare and there was
Cloudflare a few months ago flicked a
switch that basically said by default
you're blocking crawlers and it's it's
just one to check out. It's one like we
have a blog post about why you may or
may not want to allow them in but that's
the most common thing that we see coming
up. The second side of that is stuff
that's blocked by by JavaScript. So for
example, I'm thinking pricing grids. you
know, if you're selling uh say credits
or like a you know, a a tiered system
and you have like a nice pricing slider,
AI obviously can't move the pricing
slider. So, it only knows the the
initial one that it's set on and doesn't
know the rest of them. And what you'll
find is when somebody's doing research
around like maybe evaluating two
problems, uh, two services and asking
about pricing, which obviously a really
core thing that someone's going to be
researching before they make a purchase,
it'll end up pulling the pricing
information off some other source that
is unowned media by you. And in a lot of
cases, we see that to be like competitor
pages. So if you're going, "What's
what's the best between this service and
this service?" and it can't pull any
data from your page. It's going to go
for their comparison page where they've
said horrible things about you I'm sure
um that are not not quite true.
>> So what you are saying is each of the
brands should just not do AI crawability
and LLM text and schema but should also
put one or the other comparison page
with the competitors. I mean adding to
listical comparison pages do
phenomenally well like bottom of funnel
ready like to buy users they they come
up all the time and I think you know
we've even personally had success with
this where somebody's about to purchase
another tool and they go what are the
alternatives to this tool and you know
you can get lifted and then suddenly
it's like you know they're aware of
tools that didn't exist before and and
suddenly you're in the conversation. So
yeah, 100% I'd prioritize um comparison
pages, especially if you've got large
like monopoly
competitors.
>> Okay. And what's one misconception about
optimizing for AI that you strongly
disagree with?
>> Oh, it's a good question. There's there
is actually a few of them out there. Um,
I think the whole idea that that AI
search and SEO are
different, I disagree with, but I also
think the idea that AI search and SEO
are the same, I disagree with. I think
they are two cousins
that are packaged quite differently and
I think there it will be interesting to
see what happens with like traditional
Google search over time like they might
actually phase out having like a results
page you many years into the future. Uh
but I think at this time it's something
that you need to be investing in on both
sides especially as the growth in AI
search is is on the way up.
Does that answer your question?
>> Uh yes it does on a yes. Now now let's
this is we come to the last segments of
our show. The future of SEO agencies in
an AI world. So for agencies client
relationship reporting is everywhere
everything. Now, how should agency be
structuring AI visibility report and
what metrics matter most and how do you
connect AI citations to actual business
outcomes like traffic and revenue?
>> Yeah, good question. I think you know
ranking is the key thing. Your your
clients want to know whether they're
ranking higher or lower than some
competitor. Share of voice is also a
super important one. They want to know
and how many of those like responses
that talk around prompts that are
important to them they're getting
mentioned in. I think the second part
here is all of the citations like how
often are they getting links that are
clickable that take them out to the
site. You can track that in two ways.
The first one is actually tracking when
somebody clicks a link from say chat GPT
AI mode whatever it is through it will
come with a normal referral link that
you can find in your GA right the
problem is you've then got links that
are unttrackable so for example how many
times have you gone into AI search
you've asked a question and it gives you
a brand name without a [clears throat]
link next to it the next thing you're
going to do is you're then going to type
into Google brand name and click click
the first one so what I would do is
probably run like a threemonth benchmark
across their traditional non-organic
search uh just to understand where it's
going, what the current growth rate is
and then for your next period, how has
that changed over time.
>> Okay. And uh all right, so with that we
come to the end of our show and James
this has been a fantastic conversation.
I I know a lot of people have
misconceptions about LLM tools and how
they run it and I think some of those
get cleared by the way LLM refs uh
works. I might have taken a trial and I
love the free tools which you offer and
I would encourage my audience to go out
and try free. It doesn't cost a dime. So
I would definitely recommend that and um
I really appreciate uh how practical
approach you are doing giving for the
AISU. So it's not something fluff you
have. So that's especially in a space
where it's so you know the AI uh tools
space is actually full of more
speculation than the real thing. So I I
thank you for that and for uh and thank
you for giving your time today.
>> No, my absolute pleasure. I loved having
a conversation and to be honest with
you, this is what what you know myself
and the team I I dream about this stuff.
So I love chatting AISO. I'd love to
chat with anyone who uh who fancies it
at any time. Um and the same with the
rest of the team. So reach out and come
say hi.
>> Right. Right. Right. And for everyone
listening, we will link to the James
work LLM uh his tools and everything and
talk and resources in the show notes. If
you enjoyed this episode, subscribe,
share it with your team and leave us a
review. Until next time, keep testing,
keep measuring, and keep adapting to
where search is really going. Thank you.
Nikhil Kamath
