Shoppers are asking ChatGPT and Perplexity for cars now. Not in huge numbers yet, but the trend is one-way. If a shopper asks “who has a used CR-V for around $28K in Modesto,” the answer engine picks a source to cite. Right now, most dealer platforms aren’t giving the assistants anything to work with.
Four things go wrong. Fix these and the picture changes.
1. No structured data
The Vehicle, Offer, and Dealer schemas from Schema.org exist for exactly this — machine-readable claims about who you are, what you sell, and at what price. Most dealer platforms ship none of it, or ship stale versions. Assistants can’t cite what they can’t parse.
2. No llms.txt file
This is the emerging standard for large language models — a plain-text file at the root of your site that tells assistants what’s on your site, what’s important, and where to look. It’s ten minutes of work. Almost no dealer sites have one.
3. Content buried behind JavaScript
Some dealer platforms render vehicle detail pages entirely client-side. Assistants read the initial HTML response. If your prices and vehicle details only appear after JS runs, the assistant sees a blank page.
4. Stale inventory
Assistants weight recent, active pages. If your VDP was last updated three weeks ago on a vehicle you sold two weeks ago, the assistant knows. Freshness signals — real ones, not just a last-modified date — matter.
The fixes aren’t hard. They’re just not on the roadmap for platforms whose customers haven’t asked yet. Which is exactly why the dealers who ship them now are going to have the citation moat when shoppers do start asking at scale.