GEO for car dealerships: generative engine optimization without the fluff

Generative engine optimization is the practice of making your content something an AI model will cite when it composes an answer. For a dealership the good news is that the engines are consistent about what they read. The bad news is that almost no dealer site gives it to them.

What generative engines read

When ChatGPT or Google AI Mode answers a local automotive question it does three things. It resolves the entities (which stores exist near this place), it retrieves pages it can parse, and it composes an answer from sentences it can quote or paraphrase. Sliders, hero videos, and JavaScript inventory grids fail the second step. Thin pages fail the third.

Our citation-share data shows the same pattern in every market: the engine cites the dealer's own domain when it can, and falls back to DealerRater, Carfax, Cars.com, KBB, and Yelp when it cannot. Falling back means someone else is describing you.

Four pages to rewrite first

1. The store page. Not the homepage. A page that reads like an answer: "Smith Honda is a Honda dealership at 100 Main St in Naperville, Illinois, open Monday through Saturday 9 to 8, serving Naperville, Aurora, Plainfield, and Bolingbrook, with new Honda inventory, certified pre-owned, and a 24-bay service department." Every clause is a fact an engine can lift.

2. The service page. Same structure, for service. Name the towns. Name the services. State the hours. State whether walk-ins are accepted. Service prompts are where dealers lose most, because most service pages are a form and a photo.

3. The FAQ page. Ten to twenty questions in the shopper's words, each with a two-sentence answer and FAQ schema. Financing with a 580 score. Whether you take out-of-state trades. EV charging on site. Loaners during service.

4. The inventory page. Expose year, make, model, trim, price, mileage, and drivetrain as text on the page, not only in a widget. Then go further and publish a machine-readable feed. See MCP for car dealers.

What not to do

Do not publish a hundred "best Honda dealer in [town]" pages. Engines pattern-match that as spam and so does Google. One strong store page that names the towns beats fifty thin ones.

Do not keyword-stuff the llms.txt. It is a description for a model, not a meta tag from 2009.

Measure it

Same rule as everything we publish: non-branded prompts, multiple engines, geo-pinned and geo-neutral, repeated, with the raw answers kept. Methodology.

How is GEO different from SEO?

SEO earns a link on a results page. GEO earns a sentence inside an answer. The technical hygiene overlaps; the content style does not. GEO content is written to be quoted.