We’ve heard it all about GEO: It’s snake oil, it’s SEO with a new acronym and higher price tag, that it’s too early to measure, and that nobody is buying a car from a chatbot anyway.
Some of that skepticism is fair. There are lots of big claims being made about AI search right now, with very few firms adding any data to the conversation. When a channel is difficult to measure, asking for proof is a reasonable place to start.
The “too early to tell” argument is the one we wanted to test. Ask ChatGPT about a dealership today, and you may get a different answer next week. Change the city in the prompt, and the result can shift again. With a small sample, it’s hard to know whether you’re seeing a pattern or just variability. So we went bigger.
Our GEO study analyzed nearly two million dealership-related prompts across ChatGPT, Google AI Overviews, and Perplexity. We built those prompts around the kinds of questions consumers ask across their customer journey, including new and used vehicles, fixed operations, financing, trade-ins, EVs and hybrids, and dealership reputation. Then we measured them across the top five cities in each dealership’s market area. We tracked three things: whether a dealership showed up, what AI said about it when it did, and whether the dealership’s own website was one of the sources building the answer.
Visibility varied by market, platform, and question type, but geography varied the most. Dealership showed up in Google AI Overviews 94.6% of the time in the primary city but dropped to 64.1% by the fifth city out.
Citations dropped even more than inclusion did. Take one dealership from the study: In its primary market, Google cited its website in 87% of responses. In its most distant market, that dropped to 49%. On ChatGPT, the same store went from 58% to 16.8%, which means that in a city it actively advertises in, its own website is informing roughly one AI answer in six.
The cause is unglamorous and even less surprising. Most dealership content gets built around the home market, and surrounding cities get a page or a passing paragraph. That surfaces the moment a prompt gets geographically specific.
What the shopper asked about also made a difference. Regional and new-vehicle prompts produced average inclusion rates of 79.2% and 78.2%, with hybrid and EV searches close behind. Fixed operations came in much lower, at 52.1%.
That fixed ops number is worth sitting with. Service is a major part of the dealership business, and it is more than 25 points behind new vehicles. Dealerships were absent from nearly half of the applicable prompts we studied.
A dealership can have hundreds of pages and still have gaps around the questions shoppers are actually asking. A page explaining what a multi-point inspection actually covers, how long a brake job typically takes, and what’s included in a maintenance package gives an AI platform something to work with. A page listing hours and a phone number does not.
The search lessons and fundamentals we already learned still matter. Technical SEO, site architecture, schema, and strong content all play a role. GEO just adds another question to the mix: can the information on the page actually support the answer a shopper is asking for?
Visibility told us whether a dealership appeared. We also wanted to know what happened when a shopper asked AI about the dealership itself, so we tested reputation-focused questions around pricing transparency, service experience, trade-in fairness, and what someone should expect when working with a store.
The responses rarely placed dealerships into clean “good” or “bad” categories. Sentiment scored between 60 and 70 on every platform we tested, with AI giving even answers with both strengths and concerns. The result was balanced, but it also meant dealerships could start to sound pretty similar. Not negative, not positive, just undifferentiated. When a shopper is asking AI to help them choose, that is a competitive problem rather than a net neutral outcome.
Poor communication and follow-up accounted for 36.3% of negative reputation themes, followed by pricing transparency concerns at 31.3% and trade-in dissatisfaction at 21.4%. On the positive side, friendly and knowledgeable staff accounted for 63.1% of positive themes. These are patterns assembled across many sources over time, not the last bad review someone left.
A dealership can appear in an AI response without their website being used as a source, which is why we tracked citation frequency alongside visibility and sentiment. A mention and a citation are two different outcomes. If a dealership appears but the answer is built from other sources, the store has visibility without much influence over the information behind it.
In the study, Cars.com accounted for 24% of third-party citations, DealerRater 19%, and CARFAX 11%, with Yelp and Google at 8% each. The platforms don’t agree on weighting either. ChatGPT leans harder on DealerRater and BBB, while Perplexity puts nearly half its third-party citations into Cars.com and DealerRater.
Most dealers haven’t audited those profiles in years. It is not unusual to find something like “2014 Dealer of the Year” still sitting on a listing as a credibility signal, and AI platforms will pass it along to a shopper in 2026 because it’s what exists on the profile.
Instead of asking whether dealers are publishing enough, it’s worth asking whether they are publishing information worth citing. Generic content doesn’t give an AI engine much to work with, and copy that could live on any dealership website in the country gives it even less.
This is where one of the skeptical arguments deserves real credit: GEO does not mean dealers should throw everything they know about search out the window.
Across a full year of Google core updates and expanding AI Overviews, organic clicks held flat for the dealers we track. The first eight-week average was 26,064. The last was 26,059. And 29.2% of Google searches still end in a click to a website, against 1.3% for ChatGPT. AI search is not draining dealership traffic. It is shaping what gets said about a store before anyone clicks anything.
Google AI Overviews also cite dealership websites two to three times more often than ChatGPT does, and those citations track the same signals good SEO has always run on: crawlable architecture, real content depth, schema, and pages that answer a question directly instead of circling it. The dealerships showing up well in AI answers are usually the ones that did the unglamorous work first.
What’s now changing is how that work is used and shows up. A shopper may still click a traditional search result, and they may also get an AI-generated answer before they ever get to a link. The job is no longer only to rank. It’s about understanding how a dealership is being represented in the answer itself.
You don’t need nearly two million prompts to get a feel for this. Ask ChatGPT or Gemini for the best dealership for your brand near your primary city, then ask the same question for another city in your market, ideally one farther from the store. Compare the answers and look at which sources are being used. Then try the same exercise with service.
A handful of prompts isn’t a measurement system, but it makes the idea tangible in about two minutes.
From there, come back to the three questions we kept returning to. Do you show up? What does AI say about you when you do? And is your own content helping inform the answer? That, more than any new acronym, is the part of GEO dealers should be paying attention to.
The full study, including the platform-by-platform breakdowns and market proximity data, is available at reunionmarketing.com/reunion-resource-center/.
V20 Group LLC
P.O. Box 10066
Raleigh, NC 27605
Join the conversation
Log in to your V20 Group Marketplace account to leave a comment. We'll bring you straight back to this post.