AI Search Cites 0.8% of Dealership Inventory Pages

00:00Imagine spending, I don't know, hundreds of thousands of dollars on your website. Or millions. Yeah. Easily millions over the years, right? You're building this massive online inventory for your business. Only to find out that the world's most advanced AI search engines are just completely ignoring like 76% of your entire website. Yeah, it's a harsh reality check.

00:21It really is. I mean, we've spent what two decades figuring out the traditional map for Google. You plug in the keywords, you build the landing pages, the search engine, crawls it, and boom, you show up. But now with the shift to AI Search and Generative Answers, it's basically like they've drawn a totally new map in, you know, disappearing ink. That is a great way to put it.

00:43Yeah. Yeah. Because it's a completely different architecture now. Right. Traditional SEO is, well, it's transactional, right? It's about matching strings of text. Keywords. Exactly. Keywords. But AI Search is about synthesizing concepts. Okay. The old tactic was basically handing a search engine. It's a massive raw spreadsheet of every single thing you sell. Here's my stuff.

01:04Right. Here are my ingredients. But an AI doesn't want a spreadsheet of ingredients. It's acting more like a, like a librarian who's looking for a completed book. Oh, I like that. Yeah. It wants to synthesize the experience, right? The original commentary, the context around the things you sell. And that's exactly the tension we are unpacking today in this deep dive. Our mission is to figure out exactly what makes an AI actually cite a local business.

01:30Yeah. And we have, I mean, this is the ultimate stress test to look at today. We are diving into a massive stack of research called the 2026 Dealership AI Search Study. It is published in August, 2026 by an independent group called A3 Brands LLC. And why dealerships? I mean, why is that the focus? Well, the automotive industry is just the perfect lens for this because of the sheer scale.

01:54Okay. Dealerships have these massive websites, right? Yeah. And they have constantly turning inventory. Plus they operate in fiercely competitive local market. Oh, yeah, dealer in every corner. Right. So if we can map how an AI engine parses a car dealership. I mean, we can apply that logic to almost any local business or retail operation. That makes sense. And the scope of this study is, honestly, it's what makes it so hard to ignore.

02:17It's huge. Yeah. They audited 661 US franchise dealerships. And they didn't just, you know, pick random lots off the highway. Yeah. Yeah. They specifically targeted the best reviewed dealer for every brand in every single state. The top performers. Right. Then they threw get this 113,094 AI prompts at them. Wow.

02:36Across four major engines. So they did Google AI Overview, Google AI Mode, ChatGPT and Gemini. Yeah. And over 100,000 prompts gives us a data set that's large enough to finally move past, you know, anecdotal SEO advice. Right. The gurus on Twitter. Exactly. The guy's just guessing. This lets us look at structural realities. And the first reality the data exposes, it just fundamentally challenges how retail businesses should be allocating their digital marketing budgets.

03:03Okay. Let's get right into that because the data flips the script on what pages actually matter. It really does. The study found that an editorial page, so meaning a research page, a blog post, a comparison guide, something like that, is 27 times more likely to be cited by an AI engine than a vehicle listing page. 27 times.

03:22Yeah. And the raw numbers really put that multiplier into perspective. Okay. So they audited over 21,000 published editorial pages and 21.1% of them were cited by an AI engine. So roughly one in five. Yeah. That's over 4500 pages. But when they looked at the vehicle detail pages, you know, the actual listings for the cars sitting on the lot. The ones with the pictures and the price.

03:47Right. Out of almost 392,000 of those pages, only 0.8% were cited. Okay, wait. I have to stop and play devil's advocate here for a second because if I'm running a dealership, my entire reason for existing is to sell cars to move metal. Yes. I want people to find the cars. And the report notes that these vehicle detail pages, the VIN pages, they make up about 76% of a typical dealer's website. Most of the site, yeah.

04:11So why on earth does an AI engine care more about like a generic blog post comparing towing capacities than the actual physical truck I am trying to sell to a guy down the street. I know it sounds counterintuitive, but it comes down to how large language models build their knowledge graphs. Okay.

04:30Those vehicle detail pages, they're generated automatically as physical stock turns over on the lot. Right. A truck comes in, a page goes up. Exactly. There is one unique URL for every single vehicle identification number. So when a car arrives, a page is born. When someone buys that car three weeks later, the page is deleted. Oh, okay.

04:49So it's essentially a massive rotating door of links, a rotating door. And the AI knows it's temporary. Oh, yeah. AI engines are incredibly resource intensive. They are not looking to index and permanently cite a fleeting piece of inventory that might, you know, throw a 404 error tomorrow. [unclear] the car sold. Right.

05:08So that 0.8% citation rate for inventory. It's not a failure of the website's code or anything. It's just an unavoidable denominator. The inventory count is dictated by the physical lot. But editorial pages, the buying guides, the maintenance breakdown, those stay up forever. Yes. Those are permanent layered information. That is the original commentary that the AI librarian is looking for.

05:32That librarian analogy makes so much sense now. And it's not just, you know, a few outlier dealerships skewing the data, either. No, it's across the board. The report points out that 89% of the individual dealerships showed a higher citation rate on their editorial content than on their inventory, which is staggering. Nearly nine out of 10 businesses found more AI visibility through their informational guides than their actual physical products. Exactly.

05:57And the practical takeaway here is that the path to AI visibility isn't through the inventory itself. Right. It's through the content surrounding the inventory. The AI uses that editorial layer to understand the authority of the business. Okay. So if we establish that these editorial pages are doing all the heavy lifting, the immediate question for any digital strategist listening to this is, how do I build the perfect editorial page? Right.

06:23What's the formula? There has to be an optimized formula. Like what kind of back end coding or review rating pushes an editorial page to the front of the line. Well, this is where the study acts as a massive reality check for the whole SEO industry. Oh boy. Yeah. They tested four major signals that everyone assumes dictate local AI visibility, which are technical AI readiness, Google star rating, review volume. And the website platform itself.

06:53Okay. And the data indicates none of them are reliable predictors of AI citation. None of them. None of them. Let's start with technical AI readiness because I know businesses spend absolute fortunes having agencies audit their site speed, their code structure, all their back ends. Oh, yeah, thousands of dollars a month. And the assumption is always that the cleanest code wins the citation, right. Sure.

07:14But the correlation between a dealerships technical readiness score and their AI citation rate. Was a positive 0.068 meaning in statistical terms that is virtually flat. It is entirely negligible. Wait, how is that possible? Because if an AI can't read your site, it obviously can't cite you. Well, you're right.

07:37It absolutely has to be able to read the site. Right. But the issue is that technical readiness is no longer a competitive advantage. It's simply become table stakes. Oh, I see. The study found that 97.9% of the audited dealerships were technically eligible for Google's AI features. Almost all of them, basically all of them. None of them were outright disqualified.

07:58100% had server rendered content. 97.1% had descriptive headings. So it's not that technical readiness doesn't matter. It's that almost everybody is already passing the test. Exactly. It's like if everyone in the room is wearing a bright red shirt, nobody stands out. Precisely. You're just part of the crowd. Yeah. And the report specifically calls out things like valid llms.txt files. Yes, I saw that in the report.

08:20I actually need you to translate that for me. Sure. What is an llms.txt file? And why were they even tracking it? So it's essentially a text file placed in the root directory of a website. It's similar to a robots.txt file. If you're familiar with that. Yeah, the one that tells search engines what not to crawl. Right. But this one is specifically designed to give large language models a clean, markdown-formatted map of the site's most important information.

08:46Oh, wow. Yeah, it's a way to basically hand feed the AI. So it's literally a cheat sheet for the AI bot. Exactly. That sounds like a massive advantage. I mean, why wouldn't everyone want that? It would be a massive advantage except over 80% of the sites they audited already had them present. Oh, yeah, when 60.3% of all dealerships are scoring a 90 or above on their overall technical readiness, there just isn't enough variance left in the data to explain who gets cited and who

09:14Doesn't. Everyone gets an A in the class so the grade doesn't mean anything anymore. What about the areas where they did fail the technical audit? Because the report mentioned something about structured data. Yes, that was really the one widespread technical failure. Only 4.3% of the dealerships had specific vehicle schema properly implemented. And schema is what again, schema being the hidden code that explicitly tells the search engine,

09:39You know, this specific number is the price, this text is the mileage, this is the trim level. Right, labeling everything perfectly exactly. But here's the kicker lacking that advanced schema did not negatively affect their citation rate. Wait, really? Really, the AI engines are smart enough now to understand the page contextually without all those manual tags.

10:00That feels like a massive relief for anyone managing a smaller website. It really is. You don't need a perfectly pristine codebase to compete. Nope. But okay, what about human reputation? Because surely a 4.9 star business with a thousand reviews has to have an advantage over a 4.2 star business with like 50 reviews? You would think so.

10:19The AI must prioritize authority, right? Human consumers definitely prioritize it. Obviously, yeah. But the data shows the AI engines in this study didn't. Wow. The correlation for Google star rating was actually slightly negative at negative 0.064. Wait, negative? Just slightly.

10:38And review volume was positive 0.001. Which is zero. That is essentially a zero correlation, yeah. And the website platform they used, you know, whether they were on a bespoke enterprise, CMS, or just a standard template provider. Like a Shopify or a WordPress kind of thing. Right. It also didn't matter. Okay, in the report, when they discuss the website platforms, it says,

11:00And I quote, the statistical p-value was 0.0226, which didn't meet the Bonferroni threshold of less than 0.0033. Yeah, that's mouthful. You are definitely going to translate that for those of us who haven't taken a graduate stats seminar recently. What does that actually mean in English? Fair enough. It's basically a measure of statistical rigor. Okay.

11:19Imagine rolling a 20-sided die. Like, in Dungeons & Dragons. Exactly. If you roll it enough times, eventually you'll hit a 20 just by sheer chance. Well, in statistics, when you test 15 different variables at the same time, like site speed, platform, star ratings, review counts. So rolling the die a lot. Exactly. You increase the risk of finding a false pattern that is actually just random noise.

11:41You'll hit a 20 eventually. I see. So the Bonferroni correction basically makes the die heavier. It forces the data to meet a much, much stricter threshold to prove that a pattern is actually real and not just a fluke. Got it. Because the P value for website platforms didn't meet that stricter heavier threshold, the researchers safely concluded any difference there was statistically insignificant.

12:03Okay. That makes perfect sense. So let me summarize this section. I can't outcode my competitors. I cannot review them. And the platform I use doesn't give me any inherent edge. Correct. Does it come down to the quality of the writing, then? Like, if I hire a fantastic copywriter to make my editorial pages 2000 words long with beautiful formatting, does that win the citation? Well, the researchers looked at this too

12:25By comparing cited editorial pages against uncited editorial pages that were sitting on the exact same websites. Oh, that's a good test. Yeah. And they looked virtually identical. In what way? The cited pages averaged 905 words. The uncited pages averaged 845 words. Wow. That's basically the same. Exactly.

12:44Cited pages had an average of 14 headings, uncited had 13. Even the density of industry-specific terminology was practically the same. On jargon. Yeah. It was six spec terms per 1000 words for cited and four for uncited. Out of 10 measured properties, none differed significantly. That is wild. The cited pages range from a brief 171 words

13:05All the way up to a massive 17,000 words. This breaks basically every SEO rule I've ever been taught. I've been told for years that long-form content is king that you need to hit a 1500 word minimum to even rank. Yeah, the old word count minimum. But the AI doesn't care if it's 200 words or 2000 words as long as it has the answer. Exactly.

13:24Right. The AI is extracting the concept, right? It's not grading the word count like an English teacher. OK. So if nothing we've discussed so far actually moves the needle, what actually just volume? Just volume. The single strongest signal in the entire 113,000 prompt study was simply publishing more editorial content, dealerships

13:47That published a higher sheer volume of editorial pages saw a positive 0.362 correlation with getting cited. That's way higher than the other numbers. Yes. Of all the signals tested, editorial volume was the only one that had a confidence interval completely excluding zero, meaning it was a definitively positive relationship.

14:07So it's really just brute force. You simply have to have more pages of information available for the librarian to pull from. It's a numbers game based on surface area. More editorial pages mean more contextual hooks for the AI to grab onto when a user asks complex questions. OK, that makes logical sense. The bigger the net, the more fish you catch. Exactly. But if I'm a digital strategist listening to this right now

14:27And I realize volume is the key, my next instinct is to look for a dashboard. Always looking for a dashboard. Right. I want a software tool that gives me a single AI visibility score, a blended number from 1 to 100 that tells me how I'm doing across all these different AI tools. Sure. But based on the behavioral data in this study, a blended score is practically useless.

14:49Wait, really? Why? Because it averages away a massive schism between how the different AI platforms actually operate. Oh, let's break that down. Because I think this is probably the most critical strategic pivot in the whole report. The study tracks two distinct outcomes, right? Being cited and being named. Yes.

15:09And that distinction is absolutely everything. OK, explain the difference. Being cited means the engine used your web page as a source in its back end processing. It might just give you a tiny hyperlinked footnote at the very bottom of the text or maybe a hidden link in a carousel. It used you, but it didn't really tell anyone. Exactly. But being named means the actual name of your business appeared in the generated text of the answer.

15:32Like it actually says, you should go to Smith Ford. OK, so it's being behind the scenes as a footnote versus center stage in the actual conversation. Exactly. Now look at the split between the engines. Google AI Overview cites a source in 98% of its answers. Which is almost always. Almost always. But it actually names the dealership in the text only 6% of the time.

15:53Oh, wow. So Google is using your homework to formulate the answer, but they aren't giving you any credit in front of the class. That's a perfect way to look at it. But contrast that with Gemini. OK. Gemini cites a source only 61% of the time. Lower. Much lower. But it names the dealership 51% of the time. That is a total inverse.

16:14Yes. Why is there such a massive gap? I mean, they're both Google products, aren't they? They are, but they serve entirely different foundational purposes and their underlying architectures reflect that. We have to think about the business model of each product. OK, break that down for me. Think about Google AI Overview. It's built right into the traditional search engine results page.

16:35Right. Right at the top. And Google search generates revenue by keeping you on that page, engaging with their interface, and eventually clicking on ads. Yep. So if Google AI Overview explicitly names the business right there in the text, Smith Ford has the best F-150s, then my search is over. Exactly.

16:54Your search journey ends. You don't need to click anything else. So by only providing a footnote citation, they force you to click through to see the source, which keeps you in their tracking ecosystem. Oh, man, that makes perfect sense. The citation is just the breadcrumbs trail. Exactly. And Gemini. Well, Gemini is a conversational assistant. It's a chatbot. Its goal is to provide a satisfying standalone answer

17:16In a conversational format, much like a human assistant would. Yeah. A human wouldn't hand you a footnote and walk away. I would say, I recommend Smith Ford. So Gemini's architecture is actively optimizing for direct naming. So if I'm tracking my success with a blended 50 out of 100 visibility score, I basically have no idea if that means I'm getting hidden footnotes on Google search

17:38Or explicit brand shout outs on Gemini. None at all. And those two outcomes require totally different business strategies to capitalize on. Right. A hidden citation might drive some traffic to your website that an explicit name drop builds immediate brand authority directly with the user. You have to decide which engine and which outcome you are actually trying to optimize for.

17:59Wow. OK. But before anyone listening pauses this deep dive, fires their technical web developer and just goes all in on mass-producing thousands of editorial blog posts, we probably need to do a reality check. We definitely do. Because data is only as good as the methodology behind it. And there are some clear boundaries to what this study is actually proving.

18:20Yes. The most crucial boundary to remember is that this study measures correlation, not causation. Which means just because two things move together, it doesn't mean one cause the other. Right. The data undeniably proves that having a larger volume of editorial content correlates with receiving more AI citations. The numbers don't lie there.

18:40They don't. But it cannot account for hidden variables. Like what? For example, a dealership that has the financial and operational resources to publish 500 high quality editorial pages probably also has a massive overall marketing budget. They probably have heavy SEO agency support and high local brand awareness already.

19:00Right. They're just a bigger player. Exactly. We can't isolate the content volume and definitively say it alone caused the citations. It might just be a proxy for overall digital maturity. That's a great point. And we also have to look at the sample itself. We mentioned at the start that they audited 661 dealerships. But they heavily skewed that sample by design, didn't they? Yes.

19:21They intentionally chose the best-reviewed franchise dealers. Right. The median rating of the entire group was 4.7 stars out of five. That's incredibly high. In fact, 94.4% of the dealerships in the study were sitting at a 4.5 star rating or higher. OK, so when the study concludes that star ratings don't matter for AI visibility, that's not exactly the whole truth.

19:42Right. It's the truth for that specific top-tier bracket. Because everyone is already highly rated. Exactly. There was so little rating variation available in the sample that these statistical models couldn't detect a review effect. OK, so we aren't saying a two star rating won't hurt you. No, definitely not. We are saying that if you are already a 4.5, fighting and clawing to get to a 4.9

20:03Isn't going to suddenly make the AI cite you more often. Right. The difference is statistically invisible at the top. Got it. And the researchers even note that their methodology would only detect a correlation of 0.112 or stronger. Which means. It is entirely possible that things like website platform word count or star ratings do have a very minor effect. But in this specific highly rated sample,

20:27Any effect those elements had was simply too small to meet the minimum detectable threshold. OK, that is incredibly helpful context. So bringing this all home, whether you are running one of these automotive dealerships, a local bakery in Brooklyn, or a massive B2B software company, the rules of AIO, AI optimization, are actively shifting under our feet. Technical perfection is no longer a competitive edge.

20:50Not anymore. It is the bare minimum requirement to enter the arena, because the AI can read everyone's code now. The baseline has absolutely been raised. So right now, based on this massive August 2026 data set, the volume of editorial content is what dictates visibility. It's what the data says. You have to provide the AI librarian

21:11With original permanent commentary. You have to publish the buying guides, the comparisons, the deep dives into your own products. You do, but honestly, that strategy leads us directly into a fascinating and, I think, slightly concerning future puzzle. Oh, how so? Well, the data clearly shows that right now, just having more editorial pages correlates

21:32With more citations, regardless of how beautifully written or beautifully formatted they are. Brute force volume wins today. Right. But imagine what happens when this report becomes the new playbook across every industry. Oh, boy. Every business owner realizing volume is the only lever left to pull is going to turn to generative AI tools to flood their own websites with thousands of generic,

21:53Synthetically generated editorial pages. Just to hit volume metrics. Just to get the numbers out. That's just a sea of synthetic words. Exactly. And if the AI search engines are citing businesses based on content volume, and that volume is being artificially inflated by other AI bots writing generic content, how quickly do the search engines realize they are eating their own tail?

22:16The whole system loses its utility for the human user if it's just machines citing machines. The AI librarian is going to realize the books are completely empty inside. Exactly. So the next frontier of digital visibility won't just be about who can publish the most content. It will inevitably be about how AI search engines evolved to measure and prioritize actual human authority

22:40And genuine human experience inside that synthetic sea. Which means eventually handing that AI a massive, artificially generated menu isn't going to work either. No. It's going to want to know if a human chef is actually in the kitchen. Precisely. Something to really think about the next time you hit publish.