A regional personal injury firm with offices in Denver, Colorado Springs, and Fort Collins asked ChatGPT who the best injury attorneys in Fort Collins were. The firm has practiced in that market for a decade. It did not come up. A two-attorney shop three blocks from their Fort Collins office did. The partners were baffled. They rank on the first page for the same search in a browser. They have more reviews, more verdicts, more of everything. And the AI skipped them anyway.
This is the specific trap of the multi-location firm, and it catches strong firms more often than weak ones, because strong firms assume brand reputation travels. It doesn't. Not the way you think.
Why AI Treats Your Branch Like a Stranger
When someone types "best injury lawyer in Fort Collins" into a browser, they're searching. When someone asks Gemini or Perplexity the same thing, they're asking a system that first has to decide which entities in the world qualify as an injury lawyer in Fort Collins, then rank the ones it's confident about. That confidence step is where multi-location firms get hurt.
AI models don't take a firm's word for anything. They look for information that shows up the same way across more than one independent source, because agreement between sources is what separates a real fact from a marketing claim. A model that keeps finding the same address, the same attorney names, the same practice areas tied to the same Fort Collins office grows confident enough to recommend it. A model that finds your Fort Collins office described three different ways, or barely described at all outside your own website, starts to hedge. And a hedging model recommends someone it's more sure about. Usually that's the small local shop whose single location has been described the same way everywhere for years.
Your Denver headquarters might be an extremely strong entity. Deep review history, dozens of verdict pages, citations across the legal web. None of that authority automatically attaches to the Fort Collins office, because to the model those are two different places, and it needs independent confirmation that the Fort Collins office is real, active, and staffed before it will put it in front of a Fort Collins client. Brand strength at the headquarters and entity clarity at the branch are separate problems. Solving the first does nothing for the second.
The State Line Nobody Optimizes For
Here's a wrinkle unique to law that most local-SEO thinking ignores entirely. Attorneys are licensed by state, not by firm. An attorney admitted in Colorado cannot represent a client in a New Mexico matter just because the firm has a Santa Fe office.
AI assistants have gotten noticeably better at catching this. Ask one of the seven major assistants — ChatGPT, Claude, Gemini, Perplexity, Google AI Overviews, Grok, or Copilot — for an attorney in a specific state, and a well-built model tries to surface attorneys who are actually admitted there. If your Santa Fe location page lists your Denver roster without making clear which of those attorneys hold New Mexico bar admission, you've handed the model a reason to be uncertain about whether your Santa Fe office can actually take the case. Uncertainty at that layer is expensive. The model would rather name a firm whose New Mexico admissions are unambiguous than risk recommending one it can't verify is licensed to practice where the client lives.
What "Consistent" Actually Costs at Scale
Consider the arithmetic a single-office firm never faces. Five offices, each appearing across a dozen relevant places — the firm site, the local business profile, the state bar directory, the legal directories, the review platforms. That's sixty separate records describing five real things. Every one of those records is a place where the address, the phone number, and the office name either match the others exactly or introduce a small contradiction.
Small is the operative word, because the contradictions that hurt are almost never dramatic. Nobody lists the wrong city. What happens is one profile writes "Suite 400" and another writes "Ste. 400" and a third writes "#400," and the local number is on the firm site while a tracking or toll-free number sits in an old directory listing. A human reads all three and sees one office. A model reading them as structured data sees records that don't cleanly reconcile, and every failure to reconcile is a small tax on how confident it can be that this office is one coherent, real entity. Sixty records means sixty chances to introduce that tax, and at a busy firm where different people set up different listings over different years, the drift is nearly guaranteed unless someone owns it deliberately.
Where Directory Authority Actually Lives
Take Martindale-Hubbell as a worked example, because it shows why not all citations carry equal weight to a model. Martindale-Hubbell has been rating lawyers since the 1800s, and its peer review rating — the AV Preeminent designation and the rest — is built on evaluations from other attorneys and judges rather than from the lawyer being rated. That provenance is exactly what makes it valuable to an AI system.
A model weighing whether to trust a claim asks, in effect, who's asserting it and whether they have reason to inflate it. A firm describing its own excellence is the least trustworthy possible source, because of course it's flattering. A rating assembled from the assessments of opposing counsel and the bench is a fundamentally different kind of signal, one the firm can't simply write for itself. So when a Martindale entry for your Fort Collins managing partner corroborates the credentials your own site claims, and does so with independent peer input behind it, the model gets to move a claim from "asserted by the firm" into "confirmed by a source with no incentive to flatter." That shift is the whole game. It's not that the directory is a box to tick. It's that the directory carries a form of verification your own website structurally cannot produce, and a model treats verified and asserted very differently when it decides whom to name.
Notice what that means for the branch office specifically. If the credentialing behind your firm's reputation is all attached to Denver attorneys and Denver profiles, the Fort Collins partner may be practicing under a strong brand while remaining, to the model, an under-verified individual in an under-described location. The authority has to live where the client is asking.
Case Outcomes Belong to the Office That Won Them
There's a habit strong firms fall into: every verdict, every settlement, every win gets published under the firm brand. The seven-figure result goes on the main site's results page attributed to "our firm." It reads well. It also quietly strips your branch offices of the thing that would make them recognizable entities in their own markets.
When a significant outcome is attributed to a named attorney at a named office, it builds authority at the location level. The model starts to associate real, verifiable results with your Colorado Springs office and the specific attorneys there, which is precisely what it needs to feel confident recommending that office for a Colorado Springs query. Attributed only to the firm brand, that same win reinforces the headquarters entity you already had and does nothing for the branch that's losing local queries. The result was real either way. Only one version helps the office that needs help.
The Client You're Actually Losing
That Fort Collins prospect who asked ChatGPT and got the two-attorney shop instead of you didn't comparison-shop and reject you. They never saw you. There was no bidding war you lost on price or reputation. You were simply absent from the answer, and the person hired the firm that got named while you never entered their consideration. That's not a slow erosion of traffic. That's a specific human being with a specific case who is now someone else's client, and it happened in the length of one conversation with an assistant.
Multiply that across every branch, every day, in every one of the seven places people now ask instead of searching. The firm you built over a decade is being quietly re-sorted by systems that don't know your Fort Collins office exists as clearly as they know your competitor's does. RankCommander's AI Visibility Index tracks how attorneys actually surface across those assistants, evaluated across thousands of real answers, and the recurring pattern for expanding firms is stark: the brand is strong and the branches are invisible.
The Part That Doesn't Generalize
What separates the multi-location firms that win these queries isn't a longer checklist than everyone else's. It's that they treat each office as a real entity a machine has to independently believe in, rather than assuming the brand's gravity pulls the branches up with it. Where any given firm should start depends entirely on which offices are already recognized and which are ghosts, and that answer is different for a two-year-old branch than for a flagship that's been described a hundred ways over fifteen years.
You can't fix what you can't see, and you almost certainly can't see this by hand across five offices and seven assistants. Right now, in the markets where you've invested years building a reputation, a smaller competitor may already be the name the AI gives when a client asks — and every week that goes uncorrected, that's your client walking into their office instead of yours. Run your firm's AI visibility scan and find out which of your offices the assistants can actually see, before the next Fort Collins prospect asks and hears someone else's name.
For further reading on the mechanics behind this, see what the top AI-recommended attorneys have in common and our deeper look at NAP consistency and AI search.