Dealership facts
Is the store's name, address, hours, departments and brand list stated where machines read it, and does it match Google?
Shoppers are starting to ask AI assistants which vehicles fit and which stores have them. For your dealership to come up in those answers, and come up right, its information has to be understandable and current to machines as well as to people. Lehvel helps make that happen.
Shopping for a vehicle, 25% of new-vehicle buyers used an AI website such as ChatGPT or Copilot, or an AI-generated overview (19% of all buyers).
A dealership executive recently shared an email from Instinct, a general AI personal assistant. It was writing for a shopper who wanted a 2026 Volvo XC90: B5 or B6, no plug-in, six seats with two second-row chairs, Ultra or Plus, Onyx Black or Vapour Grey.
It asked the dealer for five things: trim, exterior and interior colors, VIN or stock number, availability or expected arrival, and lease eligibility. It said it was an inventory inquiry, not a request to hold a vehicle.
That is one email, not a market. But it shows the shape of what is starting: software that knew the exact configuration, asked what a good internet sales manager asks, said plainly that it was not a hold, and came in by email, the one door every dealership already has.
The questions were operational. Trim, colors, VIN or stock number, availability or arrival, lease eligibility. None of that is on a homepage, and most of it is not reliably on a vehicle detail page.
The shopper never visited the page. The assistant read what the store publishes and asked for the rest. The dealership's facts still decided the outcome. The page was where they started, not where the shopper was.
The existing internet was built for people reading pages. A vehicle detail page is written to be looked at. Software needs the same unit as a record: an identifier, the configuration, the price, the condition, whether it is on the lot, and when each of those was last confirmed.
Assistants today, agents next. ChatGPT, Gemini, Claude, Perplexity and Muse already help shoppers research. AI agents, systems that carry out several steps for a buyer such as writing to a store, are beginning to appear. They need the same information. What they are allowed to do with it is a separate question, and it is yours.
| Compared | What it asks | Where the answer lives |
|---|---|---|
| Dealer facts | Who are you, which brands, which rooftop, what hours, which department, what the service side can do | Website, Google Business Profile, OEM listings |
| Vehicle facts | Year, model, trim, body style, drivetrain, colors, condition, mileage, VIN, stock number, listed price | Inventory pages, inventory feed, DMS |
| Status | Available, in transit, allocated, on hold, sold, or unknown | DMS, OEM allocation |
| Timing | Expected arrival | OEM and logistics data |
| Eligibility | Lease, finance or other programs, where you've chosen to say | Your policies, lender programs |
| Next step | Inquiry, quote, callback, test drive | Your CRM and BDC |
Every dynamic answer needs the time it was last checked. If the system can't confirm a unit, the right answer is “unknown,” with a way to ask. Never a false “no,” and never a false “yes.”
The question is not whether your website has structured data. It is which vehicles you actually have, what is true about each one, where that information came from, and how current it is. Price, mileage, availability and location change quickly. A listing that is still in the page is not proof that the unit is still on the lot.
Every fact that can change carries where it came from and when it was last seen. Structured data on your own pages is one way to make that explicit and machine-readable. It does not create rankings or promise that an assistant cites you. It makes the information legible.
The first four are useful before any action authority exists. Lehvel builds for them now and keeps the fifth behind your explicit permission.
One record of the dealership. Rooftops, brands, hours, departments, service capabilities and contact routes, kept consistent across your website, Google and the assistants customers use.
Inventory truth. One record per unit, read first from the public inventory pages you already publish and then from the feed you already produce, with a source and a time on every fact that can change. We start with what is public, not with your DMS.
Website, search and AI representation. Your site, your listings and your structured data carry the same facts, so what a person reads and what a machine reads agree.
Observatory. We ask ChatGPT, Claude, Gemini and Perplexity about your store, your brands, your locations and specific units, and compare the answers with your own facts, with the date on every answer.
Business Check. Findings with evidence about what machines can and cannot understand about your dealership, in a fix-first order. Not a score.
Assistant and agent interfaces. Read-only questions first, like “is this unit available?” Then one structured inquiry that lands with your team the way a web lead does today. The same facts and rules serve your website, ChatGPT, Claude and whatever comes next.
Human-reviewed action where judgment matters. A person at the dealership answers every inquiry. Anything that commits the store waits for your opt-in and rules written with you.
Is the store's name, address, hours, departments and brand list stated where machines read it, and does it match Google?
Can a machine read your units as units, with a VIN, a trim, a price, a condition and a status on each one, from the public pages you already publish?
Can an assistant working for a shopper find the store, read a unit correctly, and tell whether it meets a request from current facts rather than from a guess?
Can a person be reached by phone, email and an inquiry page? Is the right department findable? What, if anything, is software allowed to do beyond asking?
When was each unit's availability and price last confirmed, and is that inside the window a shopper's assistant should trust?
Do your website, your feed, your Google listing and the assistants' answers say the same thing about the same unit and the same store?
Findings with evidence, ranked by what they cost the store, each with the fix named. No score. Nothing behind a login: we read what your pages declare, the feed you give us, and public listings.
Current. Websites, search presence and listings work; checks of what ChatGPT, Claude, Gemini and Perplexity say about a business against its own facts; and one managed record of the business behind them.
Next. An inventory record read from your public inventory pages and the feed you already produce, with a source and a time on every fact. The dealership dimensions inside the Business Check. Read-only questions an assistant can ask about your units. One structured inquiry that lands with your team. We want one dealership to build this with first.
Horizon. Actions beyond an inquiry: holding a unit, booking time, deposits, finance or lease steps, trade valuation. Each needs policy, identity, integration and human review, and your opt-in, before it is switched on. Lehvel never negotiates for either side.
We can't promise that any assistant will recommend your store or send you a buyer. What we can do is make sure that when one asks, your dealership has a correct, current answer and a way to take the request.
A paid working session for dealerships and automotive groups, built on your real public presence and inventory data rather than a generic presentation. Before it, we read your pages the way a machine reads them and ask the assistants about your store and specific units. In it, we answer seven questions: can they understand what you sell, tell which vehicles are actually available, separate stale inventory from current, understand your locations, departments, hours, brands and service, pick the right vehicle or store for a stated need, find the correct contact path, and what would have to exist before an agent could safely act for a buyer.
No. Search still matters and we do that work. This is about what happens after a shopper's assistant has found you: whether it can read your inventory, tell if a unit fits, and reach a person. Ranking is one layer; being read correctly is the next.
An AI assistant working for a shopper can find your store, read accurate dealer and inventory facts, tell whether you have what the shopper wants, and reach the right person. An AI agent, later, could send you a structured request your team answers. Nothing acts for the store without your say.
No. The first version reads the public inventory pages you already publish, then the feed you already produce. The website stays where it is. We don't assume your website provider exposes an API because it generates pages, and nothing needs a login to your dealer management system.
Not today. Questions and structured inquiries come first. Holds, lease and finance steps, deposits and trade valuations come later, if at all, and only with your opt-in and approval rules.
No. An inquiry from an assistant lands with your team like any other lead. The difference is it arrives with the exact configuration, and your answer comes from current inventory instead of someone looking it up by hand.
Yes, and it matters more there. Each rooftop is its own record with its own brands, hours and inventory, so an assistant doesn't send a Kia shopper to the Ford store, and the group sees every rooftop the same way.
A ChatGPT plugin is one way to hand your dealership's facts and request forms to ChatGPT. It's useful only if the facts behind it are right. We build the facts and rules first, then connect them to each assistant.
ChatGPT plugins for local businessRun Check My Business. It compares what ChatGPT, Claude, Gemini and Perplexity say with your own facts, and a person at Lehvel reviews the findings before they reach you.
Check My BusinessBook a time with Lehvel and we’ll go through what your business actually needs.