Copy the formatted transcript to paste into ChatGPT or Claude for analysis
Seammerris just dropped a massive 10
million keyword AI overview study. Just
dropped it. Dropped it yesterday and I
skimmed it. It has a lot of interesting
facts. It's a big read and I started
reading it and I'm like, you know what?
Why am I reading this alone when I have
an audience of SEOs who are just as
interested in this as I am? So, I'm
going to read this. I'm going read this
with all of you. And if you're watching
on YouTube, I'm going to share all the
graphs and everything that this mega
article shares. I'm going to share it
all as well. Thank you to Semrush for
making this. All right, let's get
started. Semrush AI overview study. What
2025 SEO data tells us about Google's
search shift. Google's rollout of AI
overviews is arguably the most
disruptive change to the search
landscape since the introduction of
featured snippets. It's true. our deep
dive analysis based on tens of thousands
of keywords SER feature overlaps and
indust if you don't know what SER means
it it means search engine results page
SER feature overlaps and industry which
is like the page of Google and industry
specific visibility shifts reveals how
and where Google is ch is changing the
rules and what SEOs publishers and
marketers need to do about it we analyze
10 million plus keywords including the
clickstream data data from Datos, a
Seamrush company that trigger AI
overviews in the US on desktop to find
the answers to these questions. So these
are the questions that will be answered.
How many queries trigger AI overviews
overall? What kind of queries are
triggering AI overviews? Which
industries have been impacted the most
by AI overviews so far? Are AI overviews
driving more zeroclick searches? This is
what was found. And so here's the key
takeaways just right at the top. So,
these are the key takeaways. AI
overviews are on the rise. We all know
that
13.14% of all queries triggered AI
overviews in March 2025. That's up from
6.49% in January 2025. Informational
content is most likely to trigger AI
overviews.
88.1% of queries that trigger an AI
overview are informational. That's huge.
Hrefs just just did a similar study and
they found that it was
99.2% of AI overviews were triggered
byformational queries. Next from
SEMrush, the number of navigational
queries that trigger AI overviews has
doubled since January from 74% to
1.43%. So that's people who are googling
looking for like a brand login page for
example. They're trying to navigate to
something that they know of. Another
one, science, health, and society are in
the crosshairs of AI overviews. Science
is up
22%, health is up 20%, people in society
up 18%, and law and government up 15%.
So, these industries are experiencing
the largest share of growth of AI
overviews, indicating a rapid shift in
how users get answers in these high
trust, information dense categories.
That's that's crazy. So, Google is
summarizing is using its AI to summarize
answers in really high trust information
dense categories, but I bet you that
people will want to click because when
you have something that's so important
in your life, you're not going to risk
an AI hallucinating and you're probably
going to go in and double check
yourself. That's what my intuition tells
me. Zero clicks. On average, keywords
that trigger AI overviews show higher
zeroclick behavior likely due to
theirformational nature. But our same
keyword analysis shows zeroclick rates
actually declined slightly after AI
overviews were introduced. Very
interesting. Here's a quote. The
continuing rise of AI overviews across
all types of keywords suggests a shift
towards more direct interaction within
search results. The implications for
digital marketers are increasingly
prevalent. Ensuring visibility now means
optimizing for AI overviews, a challenge
that Semrush is uniquely equipped to
address. Thank you, Semrush. All right,
so why should you care about AI
overviews? AI overviews are AI generated
summaries that appear directly within
Google search results. Built on Google's
generative AI models, they aim to
deliver an instant synthesis of relevant
content pulled from the web. Unlike
featured snippets which extract a
sentence or two from a single source, AI
overviews attempt to consolidate
knowledge from multiple sources to
present users with a unified answer. In
short, they're turning Google into both
a search engine and an answer engine.
Three things. These summaries appear
above organic results. They reduce the
need for users to click through to
websites. They may cannibalize traffic
from publishers, e-commerce platforms,
affiliates, and content marketers.
That's if you're going afterformational
queries. Don't go after them. Method of
the of the study. Seamrush analyzed 10
million plus keywords to define the
share of queries triggering AI
overviews. Monthly changes in AI
overview share by intent type. Oh, cool.
So, what's your what's the intent be
behind the search? Keywords that trigger
AI overviews. This this is everything
that Seamrush analyzed. They analyzed
10,000 keywords to define keyword
length, keyword difficulty levels, and
search volume levels. SER feature
co-occurrence with AI overviews. So what
other SER features are occurring? Domain
visibility by industry. 200,000 plus
keywords to define zeroclick rate trends
from January 2024 to March 2025. This
allowed us to observe what kinds of
content AI overviews target, how that's
changing over time, and what SEO
strategies are still effective. Oh,
zeroclick behavior, the biggest SEO KPI
shift. Our AI overview stealing clicks
from the organic search results. To find
out, Semer partnered with Datos, a Semer
company, to analyze over 200,000
keywords, each with a search volume over
100. So, all the keywords that were
studied for this for these 200,000
keywords, search volume was over 100
that either triggered or did not trigger
an AI overview between January and March
2025. To understand causality, we also
tracked a set of keywords that then that
did not trigger AI overviews in January,
but did in March. This allowed us to
directly compare zeroclick behavior
before and after an AI overview was
introduced for the same keyword.
Keywords that trigger AI overviews do
tend to have higher zeroclick rates on
average, but this may reflect
theirformational intent, not causation.
Although keywords with AI overviews have
higher than average zeroclick rates, the
trend is not consistently rising.
Between January and March 2025,
zeroclick behavior among these queries
actually declined slightly, challenging
the assumption that AI overviews always
reduce clicks. This points to a more
complex interaction between query type,
user intent, and how AI answers are
delivered. When we analyze the same
keywords before and after they started
showing AI overviews, we found zeroclick
rate slightly decreased from
38.1% to
36.2%. That suggests AI overviews do not
automatically increase zeroclick
behavior. And that makes sense. And that
makes a lot of sense. Language is
nuanced. search is nuanced, information
is nuanced. If there's something that's
really important, you might want to dig
deeper and make sure, again, like I
said, the AI is not hallucinating. Or
you might be looking for a unique voice
on on a topic, in which case, you're
also going to click deeper so you can
actually hear from the author rather
than the AI synthesizing the author who
might emphasize the wrong points. You
see that all the time, the AI will
emphasize the wrong things. It might
summarize things a bit accurately, but
it's it's not emphasizing what's most
important. And you see that, too. Here's
another quote. AI overviews are
reshaping the search landscape in real
time, and marketers can no longer rely
on rankings alone. Visibility now means
earning a spot in the answer itself. To
succeed, marketers will need to create
content that works within the evolving
complexities of AI search, and that
gives users real value. No clicks
required. I personally think that's a
BS. I think that's BS. I think the AI
overview is it's it will cite it
sources, but just seeing the brand in
the AI overview is really not going to
do that much for for boosting the bottom
line and for getting you sales and for
getting you customers. I think the best
thing to do is to not go after the
keywords that trigger AI overviews and
go after keywords where even if go
basically transactional keywords and
even if an AI overview is triggered for
a transactional keyword, people are
still going to click through because
that's the intent behind the keyword.
People want to take action. That's what
a transactional keyword is. What
percentage of queries are triggering AI
overviews? We looked at 10 million plus
keywords and defined the percentage of
all queries that are currently
triggering in AI overview. And the data
shows they're rapidly increasing in
2025. AI overviews were triggered for
6.49% of queries in January. That
climbed to
7.64% in February, an 18% increase, and
then up to
13.14% by March, 72% growth from the
previous month. Woo! Oh, wow. That's
That's a big That's some big increases.
This suggests AI overviews are more than
a passing feature. They're changing how
information is found online, most
notablyformational content, the kind
that provides answers and insights, has
been most affected by AI overviews. By
the way, compact keywords, my 13 plus
hour SEO course, is about going after
transactional keywords, notformational
keywords. Though the lessons that I that
I teach also work for ranking for for
informational keywords. I used them one
time to rank for anformational keyword
above the New York Times five minutes
after I posted the article. I posted the
article five it was an it was targeting
anformational keyword. It was during
COVID. Five minutes later I was ranking
number one and previously the New York
Times was ranking number one. So, these
lessons work, but I just totally advise
go after transactional keywords. You can
learn more at compactkeywords.com. And
if you've checked out the compact
keywords landing page in the past, I
just put up a new explainer video for
it. It's a 10-minute explainer video
showing what the compact keywords method
is, saying why it's better than blog
SEO, which is not what not what compact
keywords is. compact keywords goes after
transactional keywords with scenario
conversionbased pages instead of blog
SEO instead of articles. Anyway, there's
a ton in this video. I spent a long time
making it. If you like this podcast,
you'll like the video. compact
keywords.com. Here's another quote from
the Seamrush study. Google's AI
overviews now appear in over 13% of all
searches, and that number continues to
climb. We know that inclusion isn't
random. It's a trust signal. Brands that
create consistent and expert level
content power AI answers. In this new
SEO landscape, authority doesn't just
rank, you become the answer. That's
another reason why going
afterformational searches is difficult.
It's if it's more competitive and you
need even more domain authority for
yourformational page to rank, you're
better off going after transactional
keywords where there's just huge gaps in
the market because people don't target
transactional keywords because they have
lower search volume even though they
have more desire behind them. So, you
don't need as much authority to rank
really well for transactional searches.
That's crazy. These are these are the
searches where people are searching to
literally take action. Maybe that's a
purchase, maybe that's use a product,
maybe that's do a discovery call, but
these people are searching to take
action. And because there's just
naturally less of those people in any
marketing funnel, these people are
undertargeted and that means you don't
need as much authority. And I've also
just seen that for years and years and
years and years from my brands and other
people's brands and a million brands. In
fact, yesterday I saw it from a from a
search result I was just googling
randomly. I saw from a search result
that had a domain authority of three out
of 100 just because it was targeting the
right transactional keywords. All right,
so here's some hard stats. What kind of
queries are triggering AI overviews? We
analyzed 10,000 terms that trigger AI
overviews. Here's what we found. Typical
features of keywords that trigger an AI
overview. Top level keyword metrics.
Average search volume 338. Median search
volume 70. Average CPC, average cost per
click is 53. Max cost per click is $26.
Wow. Average keyword difficulty is.38.
These values show that AI overviews
currently tend to appear for longtail
low difficulty informational queries,
not high volume transactional keywords.
This is consistent with Google's
previous rollout strategies for features
like featured snippets and people also
ask those began with fact-based queries
before expanding into ambiguous or
controversial topics. For example,
theformational query, what is BMR
currently triggers an AI overview. BMR
or basil metabolic rate is the number of
calories your body burns at rest to
maintain basic life sustaining
functions. Oh, while the commercial
query best protein powder does not
trigger an AI overview. So, the question
is, is Google just testing AI overviews
in low-risk areas before expanding them
to more queries? We analyze 10 million
keywords to define how the search intent
of queries that trigger an AI overview
has changed over time for clues. The
search intent of queries that trigger an
AI overview. This breakdown confirms a
clear insight. AI overviews are not
targeting conversionheavy content yet.
That's what it says. And even if it
does, I just want to emphasize, I want
to I want to repeat, even if it does,
people are are searching to convert. So
the AI is just going to recommend the re
recommend brands for the searchers to
convert and you are putting out content
that trains the AI to recommend your
brand. Seamrush continues, "But the
percentage of keywords triggering an AI
overview with commercial, transactional,
and navigational intent has slightly
grown since January. Commercial queries
are up from
6.28% in January to 8.69% in March.
Transactional queries up from
1.69% to
1.76%. Navigational queries up from 74%
to
1.43%. Here's what this means. Top
offunnel informational content is the
most at risk from AI overviews. Only a
small share of keywords that trigger an
AI overview are commercial,
transactional or navigational, but there
has been an increase in the number of
middle of the funnel and bottom of the
funnel terms that trigger AI overviews
since January. Slight increase.
Navigational AI overviews are doubling.
A major signal that even branded traffic
is at risk. And but it's a it's a
navigational I still I wouldn't be
worried about navigational AI overviews
because people are trying to go
somewhere. Oh, my coffee is ready.
Excuse me for a second. I'm back. I'm
back with my coffee. If you're watching
on YouTube, I have this thermos. It's a
Brooklyn little league soccer thermos
that we probably got decades ago in
Brooklyn. And I take it with me when I'm
traveling. And I'm I'm traveling through
Europe as I record this. And I have this
uh this thermos from Brooklyn with me
just as I as I travel. I don't need to
bring it with me. It takes up extra room
on my luggage, but I like having this
Brooklyn thermos. That's where I'm from.
Just with me wherever I go. High
opportunity keywords for now. It's from
Seish. High opportunity keywords. The
vast majority of AI overviews are
focused on low CPCformational queries.
That means commercially valuable
keywords, especially those with CPC
above $2 and keyword difficulty below
30% are still largely untouched. We
filter these using Semer's CPC and and
KD data to identify queries that are
high value, not triggering AI overviews,
still relatively easy to rank for. These
are your low competition, highreward
keywords, and ideal for targeting while
AI disruption is still limited to the
top of the funnel. Examples include, oh,
and there's a ton of keywords here.
Start your own insurance business,
removable prostadontics, how much to
install sliding glass door, employee
performance review questions, how often
should you do cryotherapy, best online
interior decorating schools, why is my
gas heater making noise, how do I become
a music teacher, cost to replace roof on
mobile home. There's just this long
list. that I'm not going to read it all
to you, but these keywords are
commercially valuable and not yet too
competitive and they don't trigger AI
overviews for now at least. You know,
I'll say this for a study that says AI
overviews don't always decrease
click-through rate. So far, a lot of
this study is about how not to trigger
AI overviews. So clearly the authors of
this study feel threatened by AI
overviews enough to say let's try not to
trigger the AI overviews and it makes a
lot of sense. I mean a lot of people
will be satisfied with an AI overview
and will not want to will not want to
click deeper. Some will again for
complex terms or terms that are high in
trust or terms where people want a
unique voice but not always. So I get
it. Keyword shape. What do AI overview
queries look like? We analyze the
structure and length of AI overview
triggering keywords and those that do
not trigger an AI overview. Keywords
that trigger AI overviews tend to be
slightly longer in word count and more
likely to include clarifications,
comparisons, or definitions.
Classicformational query shapes. These
queries are longer than navigational
keywords, but shorter than
conversational searches. They're often
structured as definitions. What is
intermittent fasting? Comparisons.
Difference between debit and credit
clarifications. Can dogs eat grapes?
These kinds of queries have been the
cornerstone of most SEO content
strategies for years. Now brands will
have to compete against AI overviews to
earn attention and clicks in the SERs.
Takeaway: Google prefers predictable
fact-based questions where it can
confidently summarize a consensus
answer. That makes a lot of sense.
Another reason why going after compact
keywords is so good because you are
target with my method of SEO you are
targeting undertargeted keywords. That's
also the reason that why my compact
keywords landing pages which I made in
2019 still rank number one for their
target keywords because other people
don't target them. So they get all these
conversions year after year and then
they they continue to go undertargeted
and because not that many people are
targeting these keywords, Google doesn't
have as much to to summarize. So there's
not a clear consensus. Back to Semrush
AI overviews are showing up most on low
CPC, low volume queries with nearly 60%
of keywords having under 100 monthly
searches. A significant share, over 65%
fall into the 21 to 60 keyword
difficulty range, indicating that Google
is targeting queries that are moderately
competitive in organic search, but still
offer little ad revenue potential. But
here's the deeper insight. 95% of
keywords with AI overviews either have
no ads or extremely low CPC. Wow. That
means Google is deploying AI overviews
primarily on queries that are difficult
to monetize, deliberately avoiding
disruption to its ad revenue model. Wow.
Key takeaway. AI overviews are spreading
across the entire customer journey, but
Google is starting with the safest,
least profitable queries. So, while user
experience is changing, the business
model behind search remains intact for
now. Where AI overviews are showing up,
SER feature overlap. Are AI overviews
replacing existing SER features or
appearing alongside them? We examined
which features most often sit next to AI
overviews in the SERs to find out. AI
overviews are being layered on top of
existing formats rather than replacing
them, at least for now. Visual and
structured content types, videos,
reviews, FAQs look more likely to
survive. That's great. Do continue to do
video SEO. Take your target keywords,
put them at the beginning of your video
title, the beginning of your video
description. Take your video transcript
if it if it is a short video, and then
use that as the rest of your video
description. Back to Semrush. Classic
monetization features, ads, shopping,
news are rarely paired with AI overviews
under 1%. Category trends. Which
industries have been most affected by AI
overview so far? To find out, we looked
at the percentage growth from September
2024 to March 2025 in the number of SERs
featuring AI overviews across different
verticals, comparing early and current
rollout periods. Key observations:
Sciences at the AI overview frontline,
22% up. Science content is rich in
structured, welldocumented, consensus
answers, making it easy to synthesize,
and that has put it on the AI overview
front line. Two, health, people and
society, law and government. Google is
weighing in on sensitive subjects. On
one hand, it's not surprising that
health, people, and society and law and
government have seen a huge rise in AI
overview coverage. All three industries
are dominated byformational queries. On
the other hand, it's surprising to see
Google roll out AI overviews so
aggressively in three industries that so
often lack consensus answers and come
with more than their fair share of
regulatory pressure and misinformation
risks. That's true. That's a good point.
The key takeaway this suggests Google
believes its models have become more
reliable and fact anchored. News and
sports is Google shying away from
real-time information. News and sports
are two categories that are almost
entirely made up of searches for
up-to-dateformational content. The fact
that these categories are lagging behind
the average growth in AI overviews
suggest Google is still wary of covering
recent events with AI generated content.
The key takeaway is Google is treading
lightly around searches for breaking
news and real-time information where
there isn't enough information to
present a consensus answer to a query.
That makes sense, too. That's also
brilliant if you're if you're doing SEO
with breaking news and I've done that
very often. You might be less at risk.
Food and drink and home and garden
surprisingly slow growth. That is
surprising. Intuitively, you'd expect
food and drink and home and garden to
have seen a similar growth in AI
overviews as industries like pets and
animals and autos and vehicles. All four
categories have similar have a similar
mix of
transactionalformational keywords. After
all, however, food and drink and home
and garden queries are triggering
significantly fewer AI overviews than
industries with a similar search intent
distribution. So, these are two
categories to keep an eye on over time.
How about real estate and shopping?
Noticeable lagards. Real estate and
shopping have seen the slowest growth in
AI overviews, which makes sense given
these two industries are made up of a
much higher percentage of commercial and
transactional queries than others. Real
estate's lack of growth also suggests
that location-based content is safe for
now. So, here's a quote from Morty
Oberstein about Pizza NYC. I think of AI
overviews like a less visual local pack.
If I were searching for Pizza NYC and I
saw Joe's Pizza in the local pack, there
is a good chance I would not go to that
website. I would look at the location,
some pictures, maybe some reviews, and
just go buy a slice. In the case of
Google's AI overviews or AI mode, I
might not have all of that context,
pictures, reviews, etc. So, I might have
to search for the site. Being listed in
the AI output can and should trigger
navigational searches. The only
difference here, well, not the only
difference is that to get into the AI,
you don't need to optimize your Google
business profile. You need to optimize
your entire web profile. And that's a
much broader comprehensive complicated
and holistic process than optimizing for
something. As I read that, my mind went
to if you were listening to this right
after listening to this, just go submit
your site to a bunch of directories.
Please do that. Submit your brand to a
bunch of directories. Submit your site
to a bunch of directories. Make sure you
are in places. So here is strategic
takeaways for marketers and SEO
strategies based on industries and the
risk level. So we'll start with science,
health, people and society, law and
government. Risk level is very high. The
strategy is move beyondformational
content. Prioritize distinct POVs,
expert commentary and interactive
formats like tools and video or just go
after transactional keywords,
notformational ones. Travel, jobs and
education, electronics, internet and
telecom. The risk level is moderate. The
strategy strengthen authority signals,
authorship, expert sourcing,
credentials, and linked references. Or
my advice, go after transactional
keywords. Next one, pets, autos,
finance, beauty, forums. Risk level is
moderate strategy. Prepare for hybrid
SERs. Optimize now for blends of AI,
organic, PAA, and multimedia. And it
doesn't say this, but Edward's advice.
Just go after transactional keywords.
Last one. News, real estate, shopping,
games, food and drink. The status, the
risk level is safer for now. Strategy,
focus on eat, experience, expertise,
authoritiveness, trust, but monitor
closely. These categories are lower on
the rollout curve, but may rise fast.
And Edward's advice, continue to go
after transactional keywords. We're
almost at the end of this thing now.
We're almost at the end of this of this
mega article. You know, this article
said that it was a 13minute read. I
don't know about that because I've been
recording this so far for 39 minutes.
Yeah, I've been taking breaks to talk
about stuff and to make my coffee with
some nice milk. But it's I don't know. I
don't know if this is 13 minutes. If I
was sitting and reading this by myself,
would it take me 13 minutes? I think it
would take me longer. So cool that I get
to read the stuff that I'm going to read
normally like this and I have people who
want to listen to me read it. It's so
neat. It's like I'm doing my work and
I'm not by myself and everyone gets to
gets to hear my thoughts aloud. Okay,
I'm I'm I'm done indulging. Back to this
back to this article. Which sites have
been most affected by AI overviews? To
find out, we looked at how the average
number of a domain's keywords triggering
AI overviews has changed from January to
March 2025, revealing which industries
are accelerating and how rapidly. Across
nearly all industries, February showed a
dip or mild decline, but March exploded
with most sectors more than doubling
their AI overview triggering keywords
compared to February. Yeah, just huge
increases across across all industries.
Which industries receive the most links
from AI overviews? Not every AI overview
links out to a domain. So, we also
looked at the average number of keywords
for which a domain is actively included
within an AI overview summary. In other
words, sites that don't just show in the
traditional organic links when AI
overviews appear, but as part of the AI
generated answer. This is a critical
distinction. While many search engine
results pages may display an AI
overview, not all domains earn
visibility inside of it. By comparing
February to January and March to both
months, we can observe how Google's
model evolved in terms of trusting
domains to power it summaries. A dip in
February suggests a recalibration phase
likely related to quality signals data
testing or policy refinement. Another
quote from Cindrum. What we have to
remember is that LLMs and their search
utilities don't have sophisticated
crawlers with rendering. They don't own
the most popular browser in the world
like Google does. This will change
optimization strategies. LLMs and new AI
search utilities will likely have to
lean on feeds, databases, and APIs more
than we are used to. You can already see
that with the addition of product feed
ingestion for chat GPT. Different
methods of discovery could offer SEOs a
whole new list of potential things to
optimize. That's interesting, too. Just
make sure your brand is submitted to a
bunch of different places. And actually,
that's the whole article. That's that's
the end. We're done. Huh. The rest is
promotional stuff for Seamrush, which is
great and everyone should check out
Seamrush. Thank you Seamrush for writing
this crazy crazy study. What takeaway
did I get? Go after transactional
keywords. That was my that was my
takeaway. And make sure your brand is
distributed across different
directories. Make sure you are active
just like putting out your brand in
other in social media and other places.
But that's what that you should have
been doing that anyway. If you have a
website, you should be doing that
anyway. You should be submitting your
website to different places, promoting
it in different places, doing normal
marketing for your brand because that
build link that builds links. It builds
links while getting attention. So much
of of link building doesn't actually
result in attention. It results in links
that very few people click or pay
attention to. My favorite link building
is is link building that also results in
attention. And so you do that which
distributes your brand in other places.
and then you continue going after bottom
offunnel transactional keywords. That's
my takeaway. Thank you again so much to
Seamrush for writing this. This was
written by Jana Ganco and actually I
think I've spoken to Jana before. I
can't remember where. I'm positive that
we've met. Anyway, this was such a great
article. I hope everybody enjoyed it.
This is episode 671 of my daily digital
marketing podcast, The Edward Show.
Thank you to everyone who made it
through all of this. I learned a lot.
Hope you learned a lot too and I will
talk to you again tomorrow.
AI Overviews appearance frequency (growing fast). Industries most affected. Zero-click search trends. Informational content at highest risk
Nikhil Kamath
