Here is the uncomfortable shape of the problem. Somewhere in your system there is a franchisee who has held page one in their city for two years and is watching inbound calls decline anyway. They call the field marketing rep. The rep pulls the report. The report says everything is fine. The report is not lying. It is simply measuring a race that fewer and fewer customers are running.
Roughly 68% of Google queries now end without a click, up from 58.5% in earlier clickstream research. The customer got the answer on the results page. They never arrived. And increasingly they never even opened Google, because they asked an assistant, got three recommendations and picked one.
The question that should keep franchise marketers up at night is not whether your brand shows up in that answer. It’s whether the specific location closest to the customer does.
The discipline that has grown up around this is called generative engine optimization, or GEO, and the one-line version is that SEO competes for a ranking while GEO competes to be cited inside the answer itself. There is no page two of an AI response. You are in it or you are not. But generative engine optimization for franchises carries a second problem that single-location brands never have to solve, and that problem is the subject of this piece.
A study published July 1 by Arobis AI looked at 100 SaaS brands across ten categories and found that first-page Google positions and inclusion in generative engine answers move independently. In some categories they moved in opposite directions. The practical translation is that a marketing lead can read a strong SEO dashboard and still be entirely absent from the shortlist an AI assistant builds before the customer ever sees a search result.
Two honest caveats, because this is where a lot of the current GEO commentary gets slippery.
First, that study looked at software companies, the subscription tools you log into through a browser rather than businesses anyone drives to. Salesforce, Slack, Mailchimp. Nobody asks an assistant for the best CRM near them, so proximity, map pins, opening hours and walk-in reviews play no part in how those brands get surfaced. Local service and food categories run on exactly those signals. The finding is a warning flare, not a franchise benchmark, and anyone handing it to you as proof of what will happen to your locations is overselling it.
Second, the evidence genuinely conflicts. A meta-analysis published in May by Cyrus Shepard synthesized 54 experiments, patents and case studies into 23 scored citation factors, and the two highest were URL accessibility at 9.5 out of 10 and traditional search rank at 9.4. That is close to the opposite conclusion. Shepard’s own summary is that winning at SEO wins you AI citations most of the time, with extra steps.
A third data point breaks the tie in an uncomfortable direction. Ahrefs found that only 38% of pages cited in Google AI Overviews also ranked in the traditional top ten, down from 76% eight months earlier. Rank still helps. It is helping less every quarter.
What you should take from the disagreement is not a tactic. It’s a posture. The relationship is real but decaying, nobody has a stable model of it yet, the people selling you certainty are selling you something and the correct response is to measure AI visibility as its own thing rather than assuming your existing dashboard already covers it.
Every brand has an AI visibility problem right now. Franchise systems have a worse one, and it’s structural.
When a customer asks an assistant for the best option near them, the assistant has to resolve a question your org chart has never had to answer cleanly. What is the entity here? The brand, or the location?
Your national content is excellent. It lives on a strong domain with real authority, it gets cited and the brand shows up in the answer. You have probably invested real money in keeping that voice consistent across every location. That is a genuine win and it is also, for the operator watching their calls dry up, close to worthless. The customer did not ask for a brand. They asked for a place to go tonight.
Meanwhile the thing that represents that operator in the machine’s view of the world is some combination of a thin location page on the brand domain, a Google Business Profile that a part-time manager last touched in March, a handful of reviews and possibly a separate franchisee-run site with a different phone number on it. That is the raw material an assistant has to work with. It is not enough, and it is not the brand team’s dashboard, which is precisely why nobody owns fixing it.
The zor looks visible. The zee starves. Both are looking at accurate reports.
The fixes here are not glamorous and most of them are things your system already half-knows it should be doing. If you want the ground-level version, we have already covered the local SEO fixes every franchisee should make. What follows is why those fixes now matter for a reason nobody was talking about a year ago.
Make every location machine-readable. Consistent name, address and phone across every surface. LocalBusiness schema on every location page with hours, services and geo coordinates. If an assistant cannot parse which entity a given location page describes, it will not risk recommending it.
Give each location page a reason to exist. Most franchise location pages are the national page with a city name swapped in. That was survivable under classic SEO. It is fatal when a model is choosing which of your 400 pages to actually pull a sentence from. Staff names, real photos, local specifics, actual service area detail. Thin pages do not get cited, they get skipped.
Reviews are now source material. Not a reputation nicety, an input. When an assistant answers a question about your location, it is pulling from what other people have written about you, and the language customers use in reviews is the language a model reaches for when describing you. A location with 30 reviews saying “fast” and one with 300 saying “the technician explained everything” are different entities in the machine’s understanding, and the second one wins queries the first will never see.
Audit the answers, not just the rankings. Someone in your organization should be running the twenty queries your customers actually ask, in the assistants they actually use, for a sample of markets, once a month. Not brand queries. Local intent queries. Then write down what came back. That is a real report and almost nobody in franchising is producing one.
Track AI recommendation frequency as its own metric. If the Arobis finding holds even partially, then rank and citation carry little shared signal, which means one number cannot stand in for the other. Two columns on the dashboard, not one.
Here is where this stops being a marketing conversation and becomes a governance one.
Location-level AI visibility is unit-level work that produces system-level benefit. The brand team owns the domain and the schema. The franchisee owns the reviews and the photos and the local detail. The ad fund pays for national. Nobody’s budget line says “make sure unit 214 exists to a language model,” and so it does not happen, and everyone’s dashboard stays green while the calls dry up.
That is not a technology problem. That is the same coordination gap that has always sat between national marketing and local execution, wearing a new outfit. It is the gap that makes national ads quietly fail the local units they are supposed to serve, and the same one underneath the fight over who actually owns the customer data. The systems that fix it will be the ones that treat unit-level machine readability as an operational standard, the way they treat store hours or signage, rather than as a marketing project that lives with whoever has spare capacity.
The brands that win the next two years will not be the ones that cracked some prompt-shaped trick. They will be the ones whose four hundredth location is as legible to a machine as their homepage is.
