How visible is a small business in AI Overviews and ChatGPT?

0:00 AI Visibility Tracking Overview and Commitment
Hey, I’m Jeremy Ashburn with PushLeads, helping small businesses get more leads without struggle and frustration. And we’re making a commitment to show our ongoing AI visibility online directly here on YouTube, on LinkedIn, and other places. And so here’s an update on this. So we have our AI visibility updated every Sunday and Monday, and so this is going to be updated information here. So let’s take a look at it. Let’s dive in. So the short version here is that we have two numbers that tell the whole story. So between June and mid-September of 2026, we ran a weekly AI visibility audit on our own domain, PushLeads.com. And this is across four different services, and we ran it ourselves because, obviously, if the measurement is going to be wrong, we wanted to be the first ones who ate it. And the point is that we’re being transparent here, right? Because so many different companies out there that do what we do are scams or they’re not transparent. It’s really the old school versus the new school way, right? So I have white hair, but why not be old, be new school, not old school, right? So let’s get the bottom line here.

1:05 GSC Data 62% Impressions Flat Clicks
We saw 62% more search impressions from April through August through Google Search Console. And we saw 2% less clicks, right? Impressions are up, but the clicks are flat. That’s sort of the name of the game, the AI Visibility Game. And then we saw an increase of 41% of AI Overview citations. And this is tracking a set number of keywords over three weeks. And we saw 147% more AI referral sessions from May through August. So these trends, the impressions climbing, but the clicks are flat, are what’s happening everywhere in America and probably worldwide right now. And almost nobody is measuring it. So, pretty wild stuff, AI is changing the game, and it’s either change and pivot or die, right? The real question is, why do we run this on ourselves first? The question that every client has been asking is, my traffic look, my, So the question that all of our clients are asking is, hey, my traffic look, look, So the question that all of our clients are asking is that, hey, my traffic looks weird. Am I showing up in ChatGPT? And until 2026, the honest answer from most agencies was like, oh, I don’t know. I mean, it’s ChatGP, it’s AI. We can’t really figure out a way to do this. And so what they would really do is have someone open up ChatGP and have, you know, and search for the client’s service and city and then screenshot the result and call it a report. That’s not really a measurement. That’s a single die roll, right? Because AI searches change dramatically.

This is part of our AI search visibility guide.

2:35 Why We Measured Our Own Domain First
But three things are happening in the background that’s changing the game here. So the first thing is that Google answers more queries itself. AI Overviews are appearing on a large and growing share of commercial searches. And the click that used to come to your site frequently does not, right? Because of AI Overview. The second thing is buying research starts inside an assistant, right? Claude and Perplexity. And they’re asking questions like who’s the best water damage company near me before they even go to Google. And we do know that AI searches for service based businesses have gone from six to 45% in one year for 25 to 26. So that’s crazy. So the question really is how much is that going to go up between 26 and 27? It’s going to go from 45 to 60 or 80%. I mean, the AI wave is here. You’re either going to jump on a surfboard and ride the wave, or you’re going to get smashed in the surf. Either way is fun, but obviously right in the way is more fun. I probably was a surfer in a previous life. They don’t even exist. But anyway, none of this is showing up in analytics. AI influence mostly produces a memory, not a referral click. And so the dashboard looks flat. However, though, AI is referring some people. I mean, somebody found me on ChatGPT back in October of last year, and that person became a client. And when they approached me, they were like a restoration company in Northwest Arkansas. And they’re like, hey, I really need to show up in Google. And their current agency wasn’t doing any work, right? So I hear this all the time. And so we started doing the work and now we’ve helped them grow from 450 to 650k in revenue in the first three months of work. So obviously agencies that are doing the work is making a difference. But the reality is many companies out there aren’t doing work or they’re just scams.

4:33 Four AI Services We Track and How
So what do we actually measure? We measured four distinct services and we looked really basically they all, all these services behave differently. They’re cited differently and they need different things to move to the top to be visible, right? So Google AI Overviews is obviously the generated answer at the top of Google results. This is the highest volume, highest stakes service, and it sits directly at the top of the search demand. And we’re tracking 190 keywords there. And the short answer is that YouTube video is the thing of it’s a single thing that drives AI Overviews from what I can tell. It’s YouTube. In fact, I did a study internally and looked at, looked at all of my 50 plus clients and I saw that YouTube was showing up half of the time. So it’s YouTube. The second thing is Google’s AI Mode. And this is the conversational search service. And we started collecting data really on this at the beginning of September. And so we have a very simple, you know, we have a very thin sample of 22 prompts, But we’re going to flag that up front and say, hey, we’re jumping on the AI visibility. My company PushLeads here in Asheville, North Carolina, where I’m based out of, and I can see I went from 20 to around 56% visibility. This is the regular visibility. I’ve got 95 keywords at the top of Google. I can see what my traffic looks like. We provide this to all of our service, all of our clients, but I can see what my AI search looks like. My AI visibility is 25.4, and I can see exactly how many AI citations I’m showing up for Perplexity. I can barely say that, I’m so perplexed. ChatGPT, Google’s AI Mode and Google’s AI review and Co-Pilot, and I can see what my AI visibility looks like over time. And then I can also see exactly what prompts am I showing up for, for AI Overview, right? And then I can also see what my, and then I can also see what prompts I’m showing up for, for other searches and which AI is showing up here, whether it’s AI Overview, chat GDP or Perplexity or whatever. So I can see all of it there. So we all have it now tracked, but this is all brand new stuff, right? That we are just rolling out because a lot of us haven’t been tracking any of this, but having this tracking data is super critical. And we’re also tracking chat GVP and chat, and we’re also tracking chat GVP. And Perplexity, which is with fixed buyer and print, and this is with fixed buyer intent prompts. And this is with fixed buyer intent prompts. We’re running that weekly and we record the brand name. And we record brand name, domain cited, competitors named, and also the position in the list and named and cited or different outcomes with different causes. And then of course, AI referral traffic sessions that actually land on site from an AI service. Split by an engine, tracked in ChatTDB, tracked in Google Analytics 4, and pulled nightly. And this is the smallest number, and in some cases, the least important one. So let’s get into the meat and potatoes here.

7:41 Finding Two AI Overview Citation Rankings
Finding number one is that impressions are up 62% while the clicks were flat. And this is what we call the AI tax. So this is the number that every business owner needs to internalize. And this is from our own Google Search Console data, April through August of 2026, and we did have a random spike of 287 from a single viral outlier. So we’re calling that whether than, rather than variant. And so the honest read here is if you strip out June’s outlier, the pattern is stark. Impressions are grew 62% from April through August. And the clicks went from, you know, around 55 to 54. So these are real searches where Google decided our page was relevant enough to surface. And the click didn’t sort of, you know, and the click didn’t come because increasingly there’s no click to give, right? Because people just see the answer right in AI review. And Google answered the question right there. So if you’re only looking at your Google analytics dashboard. This period looks like failure, and many business owners have fired their agency over a chart that looks exactly like this one. The visibility looks like it’s going up, but the clicks are about the same. And here’s finding number two. We’re the second most cited domain in our own AI Overview set. So across the last three weekly sweeps, we’ve recorded every domain cited in AI Overview and with our own keywords. And then we counted and the results surprised us. We basically were 61 citations, YouTube had 141 citations and we had 61, whereas Reddit had 46 and on down. So we were actually above Clutch and Thrive and some of these other major Industries and we were cited more than every and so the odd thing is that and so the crazy thing is that we were cited more than any competing agency in our vertical, more than Wikipedia, more than WebFX, BlueCorona, or even Semrush’s agency directory. The only thing ahead of us is YouTube. And that’s the aggregate of every video. And that’s the aggregate of every video on the internet. And of course the caveat here is this is our own keyword set. And the caveat here is this is our own keyword set. But any domain measured against queries chosen for its own relevance is going to look good. So this is not like PushLeads is the number one. So this is not like. Like a search like PushLeads is the number two SEO agency in the world. It’s within 190 queries that describe our business. And we have built the deepest citation footprint. So no one else needs to win every search on the internet, but you do, but you do need to win the 190 searches that describe what you sell. And that’s the whole idea here. And then let’s get into the citation trend.

10:30 Citation Rate vs Absolute Footprint Growth
This is the growth rate versus the absolute footprint. And this is the distinction between rate and absolute count. And most reports hide this. So what we basically had around August 30th was 166 keywords checked and we showed up for 85 AI Overview. And we were cited 17 times, for 85 over and for 85 overview. And so starting at the beginning here from August 30th, we tracked 166 keywords, AI Overview present with on 85, and we were cited on 17. So our citation rate was 20% on September 2nd. On September 6th, we had 191 keywords checked and there was 101 AI overview present. and we were cited in 20, that’s a 19%. And then September 13th, we checked 205 keywords and we were present on 113 and we were cited 24. So it was 21.2% citation rate. And so our absolute citations grew 41% in the last three weeks from 17 to 24, whereas the rate barely moved because we expanded our track keyword set at the same time. We added 39 harder, newer terms. So growing your absolute footprint while holding your rate is, is steady. So growing your absolute footprint while holding your rate steady against a harder keyword set is a generally good result. So if you claim the rate grew 6% that would be dishonest against a 1.2% move inside the sampling noise. But the exciting thing here was going from 17 to 24 citations while adding more keywords. So that’s really exciting. That’s showing me that the YouTube videos that we’re doing as well as all the, all the content we’re publishing on the site is really making a difference. And so now on to finding number three, we were named one out of.

12:20 Finding Three ChatGPT and Perplexity Prompts
So now on to, so now on to finding number three, we were named one out of every five AI answers with zero negative sentiment. The honest reading is the rate did not improve. It started around 47-58% with a tiny 17-24% sample and settled at 18-23% on a 100% sample. Well, that’s not performance climbing. That’s simple bias correcting. Early weeks tested prompts that we already knew that we’re strong in, right? So this is the sentiment classification with 333 mentioned. We had 18 positive and 177 neutral, and we had 138 false positives, and zero negatives. So what does all this mean? The queries where AI names and cites us on Google’s AI Overviews were We’re showing up for disaster restoration, SEO, and SEO. And of course, that’s one of the verticals we’re moving into, disaster restoration. We’re showing up for local SEO firm, local SEO service, SEO companies, SEO firms, and also the professional SEO service, search engine marketing expert, and also pest control, GPT optimization, pest control keywords, plumbing keywords, top plumbing keywords, pool builder, SEO, and then how to start a restoration company. And of course, with things about, and of course, we’re showing up for SEO client dashboards and small business SEO services. So this is the mix of generic high competition head keywords and specific vertical operator terms that match exactly what the buyer we’re looking for. Now in ChatDBT, We showed up for these buyer intent pools, how to choose a water damage restoration company, how much does it cost for someone to clean mold, how much does it cost to encapsulate a crawl space. Fire damage restoration costs, sewage backup, cleanup costs, and how much does water damage repair costs, and IICRC certification lookup, and, and almost all cost questions, the decision answers, the moment buildings, the moment something goes wrong before anyone calls anyone else, mostly commercially viable real estate. And this is the most commercially viable real estate in AI search. And then of course the search, how to start a water damage restoration company is a search for someone who’s going to become a buyer. in about eight months, right? The AI views are surfacing us at the top of that funnel. And this is a position in the organic search that we really want and we’re excited to have. So

14:40 Finding Four AI Referral Traffic by Engine
on to finding number four, AI referral traffic is real small, but growing fast. From May to August, we went from 30 to 74 sessions. And these are measured sessions, right? From Google analytics. And that was up 147%. So September is tracking to roughly 71 on a half a month of expellation. And so September is tracking on only a half a month of up to 71. And that’s a plateau that’s considered a new level, whether whether and that’s. A plateau at a new level rather than continued growth. So if we scroll down here in May, we had like 16 ChatGPT, seven Gemini’s hardly anything Perplexity. And June, we had nine Copilot, three Perplexity, 13 Gemini and 10 Claude and 30 ChatGPT. And then in July we had two Copilot, seven perplexities, 17 Gemini, 11 Claude, and 20 ChatGPT. And then look at Claude. Claude went basically from 11 to 36. So apparently Claude really likes us. I’m not sure. Hey, thanks Claude. I love you too, man. But Perplexity is 11. Gemini is 16. Claude is 36 and then chat is 11. So the thing here is that we don’t want to chase the volume. 74 sessions a month is not a business. The value of AI visibility is almost entirely in the unclicked impression. The buyer who’s reading your name and remembers it three weeks later when the pipe burst, right? Of course, the engine mix is very volatile. Claude tripled in August. We went from 11 to 36 impressions, becoming our single largest AI referral. So building a strategy around one asset is a bad, it’s a bad vet because everything is so volatile and the, and the search has changed so dramatically. And of course, engagement is weak. Out of 74 AI in sessions, only 25 were engaged around 34%. So AI referred answers. So AI referred visitors already got the answer. They were verifying, not researching and we’re proving the claims fast. So finding number five, what actually moved the needle?

16:48 What Actually Moved the Needle Content Depth
90 new long-form pages and 179 optimizations in four and a half months. August was a heavy month. We did 28 blogs, 16 silo pages, and 84 optimizations. And Google did Google’s AI Overview citations began climbing in the, in the last week of August and rose every sweep since. So depth beats for, so what are the findings? Finding number one is the depth beats frequency. Our floor for new articles is around 2,500 to 4,500 words per piece. Shorter pages get indexed and ranked fine, but they get cited a lot less. AI systems prefer long form content around 25 to 4,500 words that fully cover a question over sources. And this tells me, and this brings us back to what I was just saying to one of my new clients about AI search. I was saying that websites are evolving past where we have just websites for humans and we have websites for humans, but also a hidden aspect of the website for AI or hidden inner hidden pages that are in the menu. Who’s going to read a 4,500 word page and AI will read the page, a human might read the page. So you’re not going to necessarily put those pages in the menu. So the websites for human, the web, the part of the website for human is going to be the design, the look and feel, the things we respond to, whereas the inner hidden pages are still going to have design and videos and stuff like that, but it’s really going to be long-form content design for the AI. Finding number two is that you want to answer in the first hundred words. Every cited page has a direct extracted answer near the top, no preamble, not a story with why the topic matters, the answer stated plainly in a sentence a model can lift. Then of course, the third finding is that we have the FAQ schema in the body. The FAQ structured data is injected directly through the SEO plugins The problem here with using SEO plugins is that the SEO plugins metafield is fragile, so what we do is we inject the FAQ schema directly under the page bottom. So if your FAQs aren’t getting picked up, check to see if you’re trying to use it for a plugin versus you’re putting the JSON directly on the page. And of course, specific numbers earn citations. Mold remediation typically runs around 1500 to 6k depending on a square footage is quotable. Mold remediation costs vary is not.

19:19 What Did Not Work Honest Failures
So it’s important that you have specific numbers within specific time frames. AI’s are like Spock and they care about that. They don’t care about salesy lingo or PushLeads is the best agency ever, right? No one cares about sales content or the AI’s don’t care about that. And we’re really just writing content for AI’s at this point, right? I mean, here. It’s humans, of course, but AI’s are taking over. And so it’s important that you always have costs with cost ranges or concrete figures. They get cited at a much higher visible rate. So what did not work? The other column, an article that only lists wins as marketing and here’s the honest accounting of what failed. So we first tried to optimize already ranking pages and that’s net negative. So, running an AI-era optimization path on a page that’s already in the top 10 in traditional Google produced worse outcomes on average than leaving it alone. On pages that are positioned 21-plus, the same path was clearly positive. And so, don’t optimize pages that are already on page one. Google already trusts those pages and the more you change those pages, the worse it gets. The second thing that did not work was pre-built page templates and it had zero measurable wins. So we had an automated play that published pre-structured pages at volume and under, and our honest outcome measurement is that it has a win rate of zero. So we stopped doing that. Volume without substance does not get cited. Brand name and the third thing that did not work is brand name rate in assistance. So three months of heavy publishing did not work. We not move our name mention rate. Our coverage grew, but at a rate which our, but, but at the rate which a model chooses to name us among our competitors held flat. So our hypothesis is this is driven by third party signals like directories, review platforms, and Reddit mentions more than our own publishing. And then the last finding is that one Google AI Mode sweep is not a trend. We have 22 prompts on September 13 showing a 9.1% brand name rate. We’re not reporting that as a finding, it’s a baseline. Ask us in November or wait for November’s video. So we do have systems in place that publish all this content. Now, this video for you, long form content. And what my system, my push bid system is going to do is take this video, remove all the ahs and ums and produce that and upload that to YouTube directly with chapters and titles and descriptions and captions and all the things. And then it will shorten this whole video down to a 15 minute video for LinkedIn, because I’m still posting on LinkedIn. So you don’t need my systems though. You just need discipline.

21:58 How to Replicate This Measurement Yourself
So here’s how to replicate this for yourself. You want to have a fixed number of prompts. Lock in a 20 to 40 question set. You want to record four things. You want to capture the prompt, service, citation, and citation rank. You want to have separate services that analyze search answers and widgets. You want to track the rank. You want to measure citation position, not just presence. And you want to watch impressions and clicks. Monitor the impression to clip gap. And these are fixes you want to do that if you haven’t done already or are just getting started, you want to fix your prompts set and never change it casually. You want to pick 20 to 40 buyer questions, including cost comparison, and how do I choose, and your brand name, and you want to run that same set every week. When you add prompts, note the date and report before, after, separately. Number two, you want to record four things per prompt, not one. The first thing is, Did the AI service appear? Were you named in the chat? Was your domain cited as a source? Who else was named? Being named and cited are different outcomes with different causes and different fixes. Number three, separate the services. AI Overviews, AI Mode, ChatGPT, Perplexity, they’re not all one number. Averaging produces a metric that moves for reasons you can’t diagnose. So you want to have separate services. That’s why our, that’s why our dashboard I showed you has separate services. And then four, you want to mat, and then four, you want to watch the impression to click gap. You want to pull, you want to pull Google. Search Console monthly, impressions climbing with fix, impressions climbing with clicks flat is the signature of AI absorption, not failure. You need citation data to prove it, or it looks identical to failure.

23:43 What This Means for Local Service Businesses
So what does this mean if you run a local service business? Number one, your traffic report measures the wrong thing. Flat clicks with, flat, flat clicks with rising impressions. it’s the signature of AI absorption, not decline. If you’re not measuring citations, you can’t tell the difference between we’re winning and it doesn’t show, and we’re losing. Secondly, the cost tractions, secondly, the cost questions of the new front door, a clear, specific, a, a clear, specific answer. First page answering how much does X cost and your city is the single highest leverage asset you can build right now. And that’s what your buyers ask before they call anyone. And then third, you compete with Reddit and YouTube. Now in our AI Overview set, YouTube, Reddit, and LinkedIn, all three take three of the four citation slots, ignoring them because that’s not where our customers are, is a 2019 position. Don’t do that. And then fourth. And then lastly, being named is worth more than being clicked at our volume. AI referral traffic is around 74 visitors a month. The value is in the hundreds of AI answers where a prospect reads your name in a short list and doesn’t click at all because that impression, because that impression compounds it’s brand advertising you earn instead of buy. And so the depth here is the moat thin content still ranks, but it does not get cited. 800 word blog posts published twice a month. Through the strategy built around a search engine that’s being replaced.

25:11 Bottom Line and Next 90 Days
The bottom line here and what’s next. Our core finding is the click is being taken and the citation is replacing it. Our impressions rose 62% while our click stayed flat. Over the same period, AI Overview citations rose 41% and we became the second most cited domain in our keyword universe ahead of every direct competitor in our vertical. If we had only been watching Google Analytics, we concluded the last five months were wasted. The second scorecard the second scoreboard set the second score the second scoreboard says the opposite. Most businesses are running on the first scoreboard alone. alone. Most businesses are running on the first scoreboard alone. That’s a gap, not a tool gap, a measurement gap. So over the next 90 days, here’s our approach. Attack the flat brand name rate. Directory presence, review platforms, and third-party sources are our competitors set. Our own publishing has taken us as far as it’s going to. We’re going to build out the AI, we’re going to build out the AI, we’re going to build out the AI Mode based on. We’re going to have a full prompt set so we can have a real trend by November instead of a single sweep. We’re going to publish long-form videos like this because YouTube is the number one-sided domain in our set and we’re underweight. So long-form uploaded transcript, question-based titles, not shorts. And then we’re going to publish the next 90-day update, including whatever doesn’t work. So, all figures are pulled from PushRank, which is our SEO operating system, measuring PushLeads in all of our 50 plus clients. The small businesses get more leads without struggle or frustration. Talk soon.