Car dealerships
Before the buyer drives to your lot, an assistant may already have said who has the car.
A buyer with a model in mind asks ChatGPT or Perplexity which dealer nearby has one, what it costs and whether the service department is open Saturday. The answer is built from the pages the assistant can fetch and read. Where the inventory is drawn by a script, the inventory is not one of them.
Whether a dealership gets named comes down to whether its vehicles exist as text. An assistant asked for a certified used hybrid within thirty miles can only name the dealer whose vehicle page states the year, trim, mileage and price in HTML it can fetch. An inventory drawn by a script after the page loads, or served from a provider's domain, gives it nothing to name.
How your customers ask
What buyers ask before they choose a lot
The buyer arrives with a model, not a dealer
By the time someone asks an AI assistant about dealerships, they have often settled on a model or two and want to know who has one, at what price, and whether it is worth the drive. The assistant answers by retrieving pages. If it finds a vehicle page that names the trim, the mileage and the price, it can say that you list one. If it finds a homepage that says 'huge selection' above an inventory that loads in a frame, it has nothing to say about you, and the dealer whose pages it could read is the one it can name.
Four questions decide most of these answers. Is the specific vehicle in stock. What are the financing terms, as published. What happens to the trade-in, including one with a loan still on it. Is the service department open when the buyer is free. Each is answered on your site or it is not; an assistant does not call your BDC to find out.
What a dealership site has to state in plain HTML
An assistant can only repeat what it can read, and for a dealership that is a longer list than for most trades, because the buyer's questions span three departments. The right-hand column is where dealership sites usually fail: the fact exists, but in a form a fetcher without a script engine never receives.
| The buyer's question | What your site states | Form |
|---|---|---|
| Do they have the car | Each vehicle on its own page: year, make, model, trim, mileage, price, stock number or VIN, and whether it is new, used or certified | HTML text, not a script-drawn widget; Product with an Offer carrying price and availability |
| Is it certified | What your certified program includes: the inspection, the warranty term, what is excluded | One page of text, linked from every certified vehicle |
| Can I finance here | Whether you finance in-house, which lenders you work with, what a buyer brings to the desk | Text on a financing page, in the words your finance manager uses |
| Will they take my trade | Whether you appraise without a purchase, whether you pay off a loan balance, how the number is set | Text; the process, not a promise |
| Are they open | Sales, service and parts hours, each labeled, and the address once | Text on the contact page plus OpeningHoursSpecification per department |
| Who are they | One legal name, one address, one number per department, identical on every page | AutoDealer, the LocalBusiness subtype, in JSON-LD |
| Are they any good | Customer reviews quoted on the page, with the model bought and the department | Quoted text, with the platform and the date |
The business type is AutoDealer, one of the LocalBusiness subtypes on schema.org, and it carries the address, telephone, opening hours and priceRange that Google's local business documentation lists. Each vehicle is a Product with an Offer stating the price, the currency and availability. Google recommends JSON-LD and asks that structured data describe only what the visitor can see, so the vehicle in the markup is the vehicle on the page, at the same price, on the same day.
Where dealership sites lose the assistant
In our audit of 9,483 small business websites (40 trades, 90 metros across the US, Canada, Europe and the UAE, late August to early September 2026), an AI crawler of our own fetched each site's homepage, contact page and about page without executing JavaScript, since nothing in the assistants' crawler documentation promises a rendering step. 3.0% of the sites rendered their content only in JavaScript, and 7.2% sat behind bot protection that served a challenge page instead of the site. Dealership inventory is where we expect both to bite hardest, because vehicle pages are so often drawn by a script or served from a provider's domain.
36.9% of the sites published more than one phone number across their own three main pages, and 43.1% were inconsistent about who and where they are. A dealership has reasons for several numbers: sales, service, parts, a tracking line per campaign. The assistant does not know the reasons; it sees four numbers with no machine-readable label and either picks one or gives none.
Reviews were unreadable to a crawler on 64.7% of the sites, and 8.9% carried any rating schema. 17.8% had no schema.org markup at all, and most of what the other 82% carried described the website rather than the business. A site built on a manufacturer's template is worth checking for exactly that: markup that says a web page exists, and nothing that says a dealer does.
Fix the inventory first, then the numbers
- Open a vehicle page with scripts disabled, or view its source, and look for the trim, the mileage and the price. If they are not there, they are not in what an assistant reads. Ask your inventory provider for server-rendered vehicle pages on your own domain: the one item here you may not be able to do yourself, and the one that matters most.
- Run the crawler access check on your domain. It reads your robots.txt the way each AI crawler token does and tells you whether a rule, yours or your CDN's, keeps the answer-side crawlers out. A challenge page from bot protection blocks them as completely as a Disallow line, and nobody at the dealership decided it.
- Settle on one number per department and use it on every page the crawler reads. If a tracking number has to stay, it has to be the same tracking number on the homepage, the contact page and the about page.
- Mark up the business as AutoDealer with sales and service hours, and each vehicle as Product with an Offer. Check that every value in the markup is on the page a visitor sees.
- Write the certified, financing and trade-in pages as answers: the question a buyer types as the heading, the answer in the first sentence, the conditions after it.
- Put reviews on the page as text, with the platform named, and leave them out of your review schema; Google's review snippet rules say not to aggregate reviews from other websites. Then measure the same buyer questions before and after, the way our methodology describes.
Go deeper
Articles for car dealerships
Fundamentals
AI visibility for car dealerships: measuring a lot that changes every day
Technical & Structure
The used-car page as a record: one vehicle, every fact a buyer checks
Content & Answer Pages
Dealer service pages: booking, warranty, recalls and the makes you service
Measurement & Reporting
An afternoon's check of AI answers against your dealership's lot
Local & Industries
What people ask AI before buying a car, in the order they buy
What dealers ask
Our inventory comes from a provider's widget. Can we change how it renders?
Usually not the widget itself, but you can ask the provider for server-rendered vehicle pages on your domain, or a feed your site builds pages from. The test is the page source: if the year, trim, mileage and price are in it, a fetcher receives them. If they appear only after a script runs, the pages exist for a browser, not for an assistant.
The manufacturer controls most of our website. What is left for us to fix?
More than it looks. On a templated site the dealer typically still edits the hours, the phone numbers, the staff page, the reviews, the financing and trade-in text and the about page, and those are most of the table above. Ask the template's support team whether AutoDealer markup with department hours is available; if not, keep the text right, because the text is what gets read.
Buyers still come from the big listing sites. Why does an assistant's answer matter?
Because the assistant's answer is mostly not a listing site. In the probe attached to our audit (September 2026), we put one buyer question each to Claude for 9,240 businesses; of every name it gave, 99.4% were actual businesses and 0.6% were directories or aggregators. Whether a dealer is among the names depends on what its own pages say. One assistant, one question per business, one day: a direction, not a rate.
Do we have to publish prices when they change daily?
The page can change daily too. What matters is that the price on the page and the price in the markup agree when an assistant reads them, and that the page carries one at all, because a vehicle page without a price cannot answer 'what does it cost'. Where you cannot publish one, publish what the buyer gets instead: the fees included, the certification, the warranty term.