“AI Can’t Find You” — Why AI Is Skipping Your Dealership Website | Wade Kerzie, Founder Kerzie AI
This is one of those episodes where you'll want to pull up your own website halfway through.
Michael sits down with his neighbor, Wade of Kerzie.ai, to talk about why AI keeps skipping right past dealership websites. Wade's been tracking how fast search is shifting. A year ago, 6% of people were using ChatGPT to search. Now it's 45%. When one of those people asks ChatGPT to find them a vehicle, AI doesn't read your homepage. It looks for a page written specifically for it. If that page isn't there, it pieces together an answer from older information, and that's when your hours and your inventory come back wrong.
Wade walks through his front cover, back cover way of thinking about it, a simple test you can run on your site in under a minute, and how his team keeps a dealer's inventory current for AI every single day. If you're stuck on an OEM-mandated website platform, he's got good news for you there too.
Connect with Wade at Kerzie.ai. His calendar and email are both there.
Michael's takeaway: We spent twenty years learning how to get found by Google. This round is faster and cheaper to adapt to, but only for the dealers who actually do it.
Timestamps
(00:00) AI Findability Matters
(01:26) Podcast Sponsor Break
(02:52) AI Upside For Dealers
(05:15) Search Shift To ChatGPT
(09:46) Structure Beats Content
(11:26) Inventory Data For AI
(12:36) Author Your AI Profile
(20:18) Future Of Websites
(25:18) Human Value And Upskilling
(29:17) Wrap Up And Contact
(29:43) Final Outro
Episode Brought To You By FlexDealer
Need Better Quality Leads? FLX helps car dealers generate better quality leads through localized organic search and highly-targeted digital ads that convert. Not only that, they work tirelessly to ensure car dealers integrate marketing and operations for a robust and functional growth strategy.
Watch The Full Episode
Prefer to watch episodes? Be sure to subscribe to our YouTube Channel and turn on notifications to get full access to our content.
Episode Transcript
(00:00) INTRO
Wade Kerzie (00:00): You're not in the game if AI can't find you. This is a real opportunity to do what we call authoring. You get to author what AI reads about you. Websites are not built for LLMs, or most are not, because it has to live at the base, what we call the root of your website. That's what AI is looking for when it does a search.
Michael Cirillo (00:21): I need to emphasize what you're saying for those that are listening, because a lot of the soup that's being sold is more content, more content, more content. And I think that does have a place. However, if it's not married to the structure of what you're saying, then it gets the hand wave.
Michael Cirillo (00:41): Now, my question from an infrastructure perspective. A dealer's listening to this and they're saying, "Okay, this all makes sense, and I like this idea of controlling the authorship of what's said about me. But I'm in an OEM program, and I have to use this website provider or that one, and they all have this limitation." What do we do about that? What's the question I should be asking, or what's the first step I should be taking?
Wade Kerzie (01:05): I have great news for you here.
(01:09) A Message From Michael
Michael Cirillo (01:09): One of the things I enjoy most about producing The Dealer Playbook is hearing from you. The messages I get from people who are getting so much value out of the podcast, applying it to their day-to-day workflows, and finding a thriving career right here in the retail auto industry. It means the world to me.
One of the ways we make this possible is through my agency, FlexDealer. And in the spirit of providing value, I think this is a perfect time to head over to www.flexdealer.com to show even further support, my beloved DPB gang. Right now, if you go to flexdealer.com, you can get a full free PDF of my number one best-selling book, Don't Wait, Dominate.
The reason I think it's so special is that a lot of the topics discussed in this book are even more relevant today, with this surge in popularized AI and people wondering, "What can I do next? How can I have a competitive advantage?" Well, that's all here in this book. I'd love to offer you a free copy if you go to flexdealer.com. It would mean the world to me, because that is how we continue to produce this show for you.
(02:31) What AI Actually Means for a Dealership
Michael Cirillo (02:31): Okay, I think I want to start here. I was recently at an event in Istanbul, sitting in a room filled with Russian and Saudi Arabian car dealers who are, and I don't mean this in a demeaning way, just an observational one, probably 18 months behind on the conversation about AI.
They're really trying to reconcile it. They have all the same access we do: Claude and GPT and all of these things. They see all of the headlines. You know, "Is this going to kill mankind in the next five years?" They're seeing all of this stuff. And I think as a result, they're asking, "What is the actual operational upside of all of this? And is it worth paying attention to?"
I know you're very involved in this space, so I want to start with you. What are you seeing in the landscape, and how could it apply directly to a dealership's operation by way of efficiency, revenue, and so on? And if so, what are the things you're thinking about when it comes to AI?
Wade Kerzie (03:48): Yeah. I appreciate you starting broader like that, because of the reality of AI and individuals using it. You've said on previous episodes that AI scales the human, right? And what we're seeing is when an individual gets educated on AI, they have this blast radius within whatever organization they're in.
If they're in a department inside a corporate office, or they run a function in a dealership, for example, they have this blast radius for what work they can get done. Their productivity goes up if they're using it right. Their quality of work goes up, their speed goes up, and their time to deliver goes down.
(04:35) The 6% to 45% Search Shift
Wade Kerzie (04:35): So we're seeing these little pockets of blast radius in the world where people are educating themselves. There are a lot of free courses out there. People are taking them and doing their own thing. Our focus is on where people are on the ground. How does it impact me?
And the reason I reached out was this trend we're seeing in searching. You're an expert. I've listened to several of your episodes. You're an expert in SEO and website optimization, all the things they've beaten into our heads. Twenty years of what I call chasing the Google algorithm. How do we rank high? How do we draw the number of eyeballs we need, and how do we create the calls to action that get someone, in the dealership world, onto the lot?
What we're seeing now, and this number absolutely floored me, Michael: a year ago, 6% of people were using ChatGPT to search. That number is now 45%. That's the rate of change in human behavior. Now, you and I could have a long conversation on a different episode about how that's the most inefficient way to actually find what you're looking for on the internet. But you've got to play the game that's being played.
Michael Cirillo (05:57): Right.
Wade Kerzie (05:58): You've got to put the right uniform on. The reality is 45% of people are going to ChatGPT now, in most cases the free version, and searching for the criteria of whatever it is they're looking for. We'll use the dealership: "I'm looking for a car."
(06:13) The Page AI Is Looking For
Wade Kerzie (06:13): Now, this can work really well, because Google doesn't ask you questions back when you do a traditional search, but ChatGPT does. It asks you questions to find out more about what you're looking for.
Google calls this an emerging convention, which is a step before a standard. There's a convention out there known as LLMs, and that's short for large language models. Most websites are not built for LLMs right now, because it has to live at the base, what we call the root of your website. It needs to be yourwebsite.com/llms.txt. That's what AI is looking for when it does a search.
And this is a huge opportunity for dealers, because they now get to control the message for what those bots are either crawling and storing in memory, or reading in real time while someone's doing a search in ChatGPT for a specific vehicle. The AI drills down into what they're actually looking for and says, "Ah, I found one."
That's what we call the back cover. I use the analogy that you're an author. The back cover of your book usually has either an "about the author" or what the book is all about. Through the years, we go to the bookstore, we pick a book up, we look at the front cover. That looks interesting. We flip to the back and read about what's inside.
That's what AI is looking for, and it's looking for it in a way that we as humans wouldn't really want to read. It's in markdown language. It looks like code. It's not really code, because you can read it in English.
Michael Cirillo (08:10): It almost looks like plain text, for those wondering what markdown is.
Wade Kerzie (08:15): Yeah, it's black text on a white screen. I can share some examples with you. But in the AI's world, that's what it's looking for. On the front cover of the book, we've graduated over the years to videos and pictures and all kinds of JavaScript that loads on the front page and looks really good to human eyeballs. That's the behavior we've been driving: trying to get people onto our website so we can give them enough information to make their next decision or move along in the buying cycle.
LLMs don't read the front page at all. They look at it, and if there's some text to be captured, they might. But in Google's own language, and I'm using Google as the gold standard here for how this works, they say the AI basically won't spend much time digging on the front cover of your book for information. It hand waves, as we like to call it in AI, and moves on. But if it sees a page that's written for it, it grabs all of that information.
(09:25) Why More Content Gets the Hand Wave
Michael Cirillo (09:25): Can we pause for a mind-blower moment with what you're saying? I need to emphasize what you're saying for those listening, because a lot of the soup being sold is more content, more content, more content. And I think that does have a place. However, if it's not married to the structure you're describing, then it gets the hand wave.
I think that's important in the context of the auto industry, because a lot of people are going to talk about the first half, and they're not going to talk about exactly what you're describing. This is where we say, "Wade, you're dropping power bombs." Because if people don't know that, it makes us susceptible to false or half-accurate information.
We've had a lot of conversations about this internally with my team. Bad data in is bad data out. Or a lot of data in without the right structure means essentially no data out. And I'm thinking about the implications. Take our tech company, an infrastructure layer for inventory data that normalizes and standardizes the data and makes sure it's accurate and enriched. It's only good if it feeds into the infrastructure layer you're talking about. Otherwise, who cares? You're basically leaving it to the agent to search for it on its own, and you're none the better.
(10:57) Keeping Live Inventory Readable Every Day
Wade Kerzie (10:57): Right. So we emphasize that the front cover and the back cover of the book need to match one another. Specifically in a car dealership, you have a very dynamic component of your website. It's called your inventory. You have to have a representation that lets AI read the live inventory on a daily basis, because it changes every day.
It would be a very heavy manual lift to scrub 300-plus units between new and used, figure out what changed overnight, and put that new information out there. This is where AI handles structuring data very, very well. So the back cover is simply a mirror image, in the legible language AI wants to read, and it can follow instructions to the referenced URLs.
There's no standard we know of yet for ChatGPT, Claude, or Perplexity delivering back the URL. That's part of this emerging convention: what will it deliver back? But what I say is, you're not in the game if AI can't find you.
This is the findable thing. We spent all this time on SEO to be findable when a human does a Google search. The same effort needs to go into giving AI what it wants to read, so it can deliver accurate information.
I think there's been this underlying assumption, and I probably had it for a while too in the early days of ChatGPT, that it was looking at websites the way humans do. What we've come to discover is that's not true at all. In fact, if it can't read the website, most of the time it defers to Google search results in the cloud. And what we find every single time, Michael, is that it's six to nine months old or it's wrong.
This is where AI can get off the rails, in what we call hallucinations. It can go wrong because it wants to deliver an answer, so it makes something up to give something back to the person searching in ChatGPT.
This is a real opportunity to do what we call authoring. You get to author what AI reads about you. Whether it stores that in memory for later reference or fetches it in real time during a search, either way, you control what AI is saying about your business: where you are, your hours, and your inventory.
From the car dealership perspective, there's a static component of the llms.txt page that needs to stay the same, because your address isn't going to change every day. Your inventory is the dynamic component. And for the gold standard on how to structure this, which is what we do for our customers, we followed the way Google says to present data so AI can read it, and for the structure, we follow Carvana's method.
There's a daughter page that's updated every single day, and that's actually llms-full. The first page says, "Hey, here's all my inventory." It's really that simple. The AI says, "I've got a page I can read, because I'm looking for llms.txt," and boom, here's a daughter page with deeper data about this dealership and its live inventory.
Michael Cirillo (14:32): Wow, you're blowing my mind right now. What did they used to say? The nerds make all the money. And here we are again. I'm loving this, because I identify as a nerd more than anything, and we're talking about something so nerdy that has real, tangible downstream operational impact for a dealer. Most dealers are not going to know this, and by no fault of their own. They're running a business. Without conversations like the one we're having, Wade, most won't know about it.
(15:10) The 30-Second Test for Your Website
Wade Kerzie (15:10): So the math supports that. I don't have the direct statistics in front of me, but I can confidently say that 85% of what we've searched for... And by the way, this is a simple acid test. In one of your recent episodes, you talked about Google Business Profile and something you can action on today. I want to take that same approach here.
Anyone can go to the root, their main page. If you're not familiar with the term, the root is where your website lives. Michael, in your case, that's thedealerplaybook.com. After that, in the address bar, type /llms.txt. Eighty-some percent, and again, I don't have the exact numbers in front of me, but we know it's the large majority, are going to throw what the internet calls a 404 error. In other words, page not found.
In some cases, because of the SEO plugins they're using on the front cover, a page has been populated and there might be some info there. But it's most likely an advertisement for the SEO plugin. It doesn't contain the actual information a user is looking for. And in a lot of cases, it's still not crawlable by the robots, so it throws them a 403.
We do this with our AI all the time. I say, "Go look at this website and tell me the status of the page as it relates to llms.txt." It'll come back and say, "We got a 403. It's not crawlable." It might be visible to human eyes, because of that plugin where someone checked the box and said, "Here you go." But it really isn't doing the job it's designed to do.
The good news there, Michael, is that it means the website hosting company itself is built to support llms.txt. Some are holding that back, and I don't want to take us too far off track.
Michael Cirillo (17:03): How come? Why would they hold it back?
Wade Kerzie (17:05): Well, they call it "supportable," right? It's an emerging convention, as Google calls it, and they're not really sure what it is. So they're not letting you put a daughter page on your website that they don't control or don't have a template for. A lot of these websites are built with a template. Drag a picture in here, drag a text box in here. So it's a page format they're unfamiliar with.
A good friend of mine is a photographer. Photographers tend to host on one of three main platforms, and he's on one of them. They will not support the llms.txt page. Not because they can't. They just don't know what it is yet.
Again, it's emerging, but this is coming so fast. I'm so happy you had me on, because this is what dealers need to hear. This is your chance to own the authored page of what AI reads about you. And for the time being, there are no ad dollars associated with it. We all know the Google ads, the Meta ads, all of the motion that's been created. This is your chance to author what AI reads about you.
(18:21) Stuck in an OEM Website Program
Michael Cirillo (18:21): Now, my question from an infrastructure perspective. A dealer's listening to this and they're saying, "Okay, this all makes sense, and I like this idea of controlling the authorship of what's said about me. But I'm in an OEM program, and I have to use this website provider or that one, and they all have this limitation." What do we do about that? What's the question I should be asking, or what's the first step I should be taking?
Wade Kerzie (18:48): I have great news for you here. Dealer.com, DealerOn, and Dealer Inspire are the three big daddies that I'm aware of. I'm not in the car business, but I've learned a lot about it doing this. Those are the three major platforms, and they all support an llms.txt page, meaning you as a user can populate that page at will.
That's what we do for our customers. We take that user access on a daily basis and scrub the inventory to repopulate the page. And we give them a dashboard where they can see the number of vehicles that came onto the lot and off the lot every single day, with a running tally of what's going on their page.
So in most cases, if you're covered by one of those three, it's something you can do. If you're on Wix or other platforms, most of them are supporting it, or they're very quickly coming around to, "We have to support this type of page, even though we may not understand it or how people are going to use it."
(19:56) Are Websites as We Know Them Going Away?
Michael Cirillo (19:56): Okay, you need to talk me off a ledge right now. Controversial opinion of the day. I'm allowed one, and I'm going to use it on this. I love the way you spelled this out. The front of the book and the back of the book need to be in alignment.
Do you think there's a world in the next five years where the concept of websites as they exist today, which is all speed and user experience and quick navigation, actually moves into a much more agentic repository of information? You talk about a swing from 6% to 45% in user adoption on GPT. What was the timeframe, a year? [Wade Kerzie: 12 months.] Twelve months. That's a huge uptick in user adoption. Are we moving into a world where the way we've been thinking about websites is going to be obsolete, and we need to think about them as repositories of clean content and inventory data?
Wade Kerzie (21:01): I'm not even going to express just my own opinion, because the short answer is yes, and we're already seeing the evidence of that. Cloudflare said last month that we've crossed over the 50% threshold, where more traffic on the internet is machine to machine than human. It's in the low 51%, 52% range. That's now agentic AI across the internet, and fewer humans.
Michael Cirillo (21:30): You know why this makes me laugh? Because eight years ago, five years ago, three years ago, 12 months ago, we were saying, "What's with all this bot traffic on my site? I'm seeing all these bots in my analytics." And it was like, okay, create a filter that filters out the bot traffic so we actually know what's going on. And what I'm hearing you say, and what I've been feeling, is, yo, go limp. Embrace the robots. You're going to see a lot more robots doing the car shopping.
And I think it's predicated on one thought process for me. I'm still working through the words to articulate it, but I'd be curious about your thoughts. We have this once-every-30-years cycle. About 35 years ago, before websites really hit, you had to shop for a car by opening the newspaper or buying the magazine. You had to call all the stores, you had to go visit them, you had to test drive everything.
Then websites came out and said, "Hey, good news, you don't have to do that anymore. You can shop from the comfort of your couch." Now here we are saying, "Hey, you don't have to do that anymore either. You can tell your agent, which knows everything about you, understands more context, and will ask you follow-up questions, like you said earlier. And it will go do all the shopping for you and present back the most viable options that match your lifestyle."
In other words, we're worried about AI replacing our jobs. Well, there are aspects of jobs, or tasks, that it does replace. Just like websites replaced the need to go to 15 different dealerships on a Saturday. Websites legitimately replaced the thing we should never have had to do to begin with. And now AI is replacing the idea of, "Now I've got to spend my whole morning going to seven different dealership websites, searching their inventory, doing a chat, calling them." The thesis for me is, well, you should never have had to do that either. Now AI can do it for you. Thoughts, feelings? What do you think?
Wade Kerzie (23:47): Yeah, no, absolutely right. With my career in technology sales, I like to compare it to the continuum that is a buying cycle. What you just articulated is the early part of the buying cycle, the part that's really full of friction for the user.
As for our AIs, we're going to have them with the newest iPhone or the newest Apple Watch, or who knows what OpenAI is going to introduce as the next device we carry. It's fair to assume we're all going to have an AI that's personal, that really knows me.
So it can go off and do the discovery and get me further along in the decision cycle. Then maybe it points me to videos on the dealer's YouTube channel, and I say, "Yes, that's one of my top two, and I'll spend a couple of hours Saturday afternoon driving to those two dealerships." Because I've gotten through the decision cycle without friction, by having my AI agent do the work that used to take the eyeballs and clicking effort of the human.
(24:57) Where the Human Still Wins
Michael Cirillo (24:57): This brings me to where I want to take the conversation as we wind down, which is my plea for people to get fully immersed in this whole AI thing. That way you're in a position to understand where the value of the human will actually increase, and where the value of the robot is.
And it's funny, Wade. Yesterday we're having lunch at a steakhouse. LongHorn Steakhouse, okay? Nobody think I'm all posh and going to some $80 or $100 steakhouse. No. LongHorn, lunch special. The server hears us talking about marketing, and he comes to our table and says, "I couldn't help but overhear. I was actually going to go into marketing, but I heard all the stuff you guys are saying about AI, and it made me want to go into nursing. I'm in the nursing program now." And I'm like, "Hey, that's awesome. Good for you."
To the point about increasing human value, I said, "Hey, can I share something with you?" He said, "Sure." I said, "Focus on the things in your practice that only a human can do. AI is not going to hold my hand when I need my hand held. The value of that exchange actually goes up." Claude has never complimented me, Wade, and made me go, "Aww." But it's from full immersion that I gained that clarity.
And you know what's so funny about this, and so heartbreaking at the same time? Right after that, he asked if I wanted my soda refilled. I said yes, and he forgot to do it.
Wade Kerzie (26:46): Humans make errors.
Michael Cirillo (26:50): And I'm like, well, it didn't land. I think the reason I'm so passionate about this idea of full immersion, to understand the line between what interactions technology will enhance and what scale can come from it, is because I'm worried most people are going to miss the point.
Wade Kerzie (27:10): Yeah. I've written about this in a couple of my essays, and I'm passionate about helping people get through what I call the blast door. Think back to the Millennium Falcon. Those doors closed really fast. In a split second, the door was closed. I'm helping as many people as I can upskill, because the judgment layer is still human.
The instruction layer, meaning telling AI what to go do, judging its outputs, and then deciding its next actions, is the human layer you're talking about. That's truly an uplift. We're enhancing our human capabilities instead of getting caught in the minutiae of changing the spreadsheet, or in your case, shopping for the car. Go have something else do that, then layer your judgment and your further instructions on top of it.
That's where I come at it from. I truly believe there's a lot of opportunity for people to do just that. Lift yourself up and understand it to the degree you need to, but don't get caught too much in the weeds of how it works. It just works if it's used properly. I always go back to that.
So yeah, I think the opportunity is almost unfair. And I want to leave your dealer listeners with this. It's almost unfair that for 20-some years we've been trained to chase the Google algorithm, as I said earlier, and now the game's changed. The good news is it won't take as long to adapt, and it's nowhere near the expense, to get the AI side of search addressed the same way you've spent all these years making the front cover of your book discoverable.
(28:55) Connect With Wade Kerzie
Michael Cirillo (28:55): I love it. What a fun conversation. My neighbor, everybody. How about this? My neighbor, Wade. How can those listening or watching connect with you?
Wade Kerzie (29:04): Well, my name has one of those funny spellings: Kerzie, K-E-R-Z-I-E, dot AI. That's me. You can connect with me there. My calendar's there, my email's there. Kerzie.ai. Thank you.
Michael Cirillo (29:17): That's awesome. Thanks so much for joining me on The Dealer Playbook Podcast.
Wade Kerzie (29:20): Yep, thank you.
Michael Cirillo (29:22): Hey, thanks for listening to The Dealer Playbook Podcast. If you enjoyed tuning in, please subscribe, share, and hit that like button. You can also join us and the DPB community on social media. Check back next week for a new episode of The Dealer Playbook. Thanks so much for joining.
