A prospective client in Sacramento, weighing a DUI charge at 11pm, doesn't open ten tabs anymore. They open ChatGPT and ask who handles this. What comes back isn't a page of links to sort through — it's a name. Confident, specific, and delivered like a recommendation from someone who already did the homework.
That's the part worth sitting with. A referral used to come from a person — a friend, a colleague, a front desk at another firm. Now a growing share of that first referral comes from a model that has never met the client and never will, and it's making a specific call anyway. ChatGPT, Claude, Gemini, Perplexity, Google AI Overviews, Grok, Copilot — seven different systems, each capable of being asked the same question and each capable of giving a different answer.
The shift from search to a single answer
Google search always kept the decision with the user. Ten blue links, a map pack, a few ads — the client did the comparing. An AI assistant collapses that into one turn. Ask it who's good, and it tells you who's good. There's no scrolling past the answer you don't like.
For a law firm, that's a strange kind of exposure. You don't get to be the third option a stressed-out prospective client compares against two others. You're either the name that came out of the model, or you're not part of the conversation at all. Most attorneys have never actually run this test on their own name and practice area — asked the assistants directly and watched what came back.
Why a model trusts some firms and hedges on others
Here's the actual mechanism, and it's simpler than it sounds. AI models don't take a firm's word for it. They look for information that shows up the same way in more than one independent place, because agreement across sources is what makes a claim trustworthy rather than just asserted. A model that finds agreement grows confident, and confident models make specific recommendations. A model that finds contradictions, or finds nothing corroborating a claim at all, hedges — and a hedging model tends to name someone else instead.
Take Martindale-Hubbell, for instance. It's been running peer-review ratings for attorneys for more than a century, and the AV Preeminent designation that comes out of that process is a genuinely interesting case study in how this plays out. The rating isn't something a firm writes about itself — it's other attorneys vouching for the work, compiled into a credential a model can point to. When an assistant is deciding how much confidence to put behind naming a particular estate planning attorney in Tampa, a peer-conferred rating like that is exactly the kind of independently-sourced fact that tips the model toward a specific, confident answer instead of a vague, hedged one. It's not that this one credential is the whole story. It's a clean illustration of the underlying pattern: information that came from somewhere other than the firm itself is what a cautious system leans on.
What the numbers actually look like right now
This isn't a hypothetical dynamic. RankCommander's AI Visibility Index continuously scans a real panel of businesses across the seven major AI platforms — more than 35,000 individual platform answers evaluated so far — and law firms are one of the tracked industries. Across a scanned sample of 144 firms, the median AI Visibility Score is 27 out of 100. A quarter of firms in that sample outright block major AI crawlers from reading their site at all, whether deliberately or by accident.
Sit with that for a second: most firms in that panel aren't losing this contest by a little. They're barely on the board, and roughly one in four never had a chance because the front door was locked. The full breakdown, updated as the panel keeps scanning, is public at the AI Visibility Index.
What's quietly at stake
Here's the part that should bother you. This is happening right now, whether your firm has looked at it or not. Somewhere in your market this week, someone asked an AI assistant for exactly what you handle — and it named a firm. Maybe yours. Maybe the one three blocks away, handling the same caseload, no more qualified than you are. Either way, that person is now somebody's client. Not because they compared you and chose the other side. Because you were never in the running at all.
You didn't lose that case. You never knew it existed. Do that math across every AI conversation happening in your market this month, and what you're looking at is a referral channel running at full volume with nobody from your firm in the room — after years spent building the reputation that was supposed to earn you exactly that referral.
Get on the right side of it
A twelve-year litigator with a strong track record and a newer solo practitioner building a name for themselves aren't losing ground in the same place. Guess wrong about which one you are, and you spend a quarter chasing a fix that was never the problem — while whoever's actually getting named keeps getting named.
RankCommander skips the guessing. The scan checks how your firm currently appears across the directory landscape, your site's own content, the technical signals AI crawlers read, and, most directly, what each of the seven major AI assistants actually says when someone asks about attorneys in your practice area and market today — the same measurement behind the Index numbers above, run specifically for you.
You've put in too many years to let a conversation you'll never see decide where the next client goes. Get RankCommander on your side, see exactly where you stand, and close the gap before the firm down the street does. Free, in under a minute, at /attorneys.