Google will put your store first if a buyer asks it to

Google added a setting that lets a person choose which sites Search puts in front of them. It is called Preferred Sources, it applies to Top Stories and to AI Mode and AI Overviews, and the only way a dealer group ends up chosen is by asking its own customers to choose it.

That last part is what makes this different from every other AI visibility lever. There is no markup to deploy, no threshold to clear, and no budget that buys a better position. Which also means no platform vendor can sell it to a dealer group as a line item, and no aggregator can outspend a store for it. The work that gets covered on the generative engine optimization side is about earning a citation from the engine. This is about a buyer overriding the engine on your behalf.

A buyer tells Google which sites to put first

The mechanic is a preferences page. A person opens Google's source preferences tool, searches for a site, and adds it. The selection sticks for future searches rather than applying to one query. From then on, that site's content is surfaced more prominently in Top Stories, marked with a preferred badge, and eligible content receives similar highlighting inside AI Mode and AI Overviews.

So there is no qualification step to fail. A dealer group's domain is already selectable in that tool today, whether or not anyone at the group has heard of the feature.

Two of the three surfaces matter to a dealer

Google lists Top Stories and the AI surfaces. Being straight about which of those is realistic matters more than counting them.

• Top Stories is news-shaped. A dealer group's blog is unlikely to surface there for the queries that sell cars, and treating it as the payoff would be selling something that will not arrive.
• AI Mode and AI Overviews are the real prize. These are where model research, cross-shopping, and the plain-language questions buyers ask before they ever type a dealership name increasingly resolve. A store that is preferred by a household is a store the engine has been told to weight for that household.

The second bullet is the entire business case. Everything a dealer group does to get cited in an AI answer is competitive work against a market. A preference is not competitive at all. It applies to one person who already chose you.

The subdirectory rule decides your unit of account

One restriction changes how a multi-rooftop group should think about this. Google is explicit that eligibility stops at the domain and subdomain level, and names a subdirectory as the counterexample.

For a group whose rooftops each run their own domain, every rooftop is separately selectable, and a service customer in one market can prefer the store they actually use. For a group running every rooftop as a folder under one corporate domain, there is exactly one thing any customer can choose, and it is the corporate domain rather than the store down the road from them. That is a structural consequence of an architecture decision most groups made years ago for unrelated reasons, and it is worth knowing before anyone builds a campaign around this. It sits alongside the other cross-rooftop problems a group inherits from its URL structure.

Dealer groups own the one asset this rewards

Preferred Sources converts on audience contact, not on authority. On that measure a dealer group is in an unusually strong position, and most of them have never counted it as a search asset.

A single rooftop touches more local households in a month than most regional publishers. The service lane sees customers on a schedule. The sales follow-up sequence already exists. There is an email list, a text campaign, an invoice that goes out after every oil change, and a thank-you page after every test drive. Each of those is a place where a store can ask a person who already likes it to spend fifteen seconds on a preferences page.

Compare that to the two competitors actually taking share in AI answers. An aggregator has enormous scale and no relationship with a specific household. A regional group three markets over has budget and no reason for your service customers to have heard of it. Neither can reach the people a store services every ninety days, and that is precisely the population this feature is built around.

The realistic ask is small and repeatable. A line in the service reminder email. A card at the cashier desk with a short link. A sentence in the delivery follow-up. None of it requires a campaign, and none of it competes with anything else the store is already sending.

Two ways to ask, and one costs a third-party script

Google documents a JavaScript button and a deeplink. The choice between them is a technical decision with a real cost on one side.

• The button. Two lines of HTML that load publisher.js from a Google origin and render a branded control, with theme and language options. It is familiar and it is the version Google promotes.
• The deeplink. A plain URL of the form google.com/preferences/source?q=yourdomain.com. It works in a link, a button, an image, a social post, an email, or a text message.

On a site running a Content-Security-Policy that restricts scripts to its own origin, the button is not free. Adding it means widening that policy to admit a Google host, on every page the control appears. Weighed against a feature Google itself says needs no implementation at all, a plain anchor is the better trade for most dealer sites. Following a link is not governed by CSP, so the deeplink needs no policy change whatsoever. That is the version running in the footer and on every article of this site.

For a dealership the deeplink also travels further. The highest-intent moment is rarely someone reading the website. It is the service customer holding an invoice, or the buyer reading a delivery follow-up on a phone, and a URL goes into both of those where a script cannot.

What Google has not said

Three things are genuinely open. Google has published no figure for how much prominence a preference confers. It has not described how a preference interacts with the ranking and retrieval that decide AI Overview citations in the first place. And there is no public adoption data on how many people have used the tool at all.

Any vendor attaching a percentage lift to this is inventing it. What holds up without a number is narrower and still worth acting on: the opt-in persists once set, the cost of asking is close to zero for a business that already emails its customers, and a competitor cannot take a preference away from a store by outbidding it.

That combination is rare enough in search to be worth a place on the list, somewhere below fixing the schema a platform shipped incomplete and well above most of what gets pitched as an AI visibility product. The technical layer still decides whether an engine can read a store at all, and no volume of preferences compensates for a site the engine cannot parse.

FAQ