A frightened defendant is going to ask an artificial-intelligence system which lawyer to hire. The system will answer. The only unresolved question is whether your firm will be in the answer.
Start with the audit → Find Out What The Machines SayRanking on Google's results page and being named inside an AI answer are two different fights, and most firms are only equipped for the one they already know.
Appearing on page one of Google still matters - it's still where a lot of searches end. But a growing share of them don't end there anymore. They end with a question typed into ChatGPT, or asked out loud to whatever's on someone's phone: "who's a good OUI lawyer near me," "who handles clerk-magistrate hearings in this county," "is this lawyer any good." The AI doesn't hand back ten blue links for the searcher to sort through. It hands back a name, or three names, in a paragraph that sounds like advice from a knowledgeable friend. If your firm isn't one of the names, you don't rank tenth. You don't exist in that conversation at all.
Ranking well on Google is necessary and still worth doing. It was never sufficient, and now it's less sufficient every month. The two battlefields require different evidence, and most firms have only ever built for one of them.
An AI's answer is only as good as what it's been able to read about you, from sources it trusts.
These models weren't trained on a private interview with you. They were trained on what's publicly written down - and they lean hardest on sources that look independently verifiable, not on anything that reads like an ad. That means the raw material for whether you get recommended already exists, or doesn't, whether or not you've ever thought about it in these terms.
Part of the audit is simply asking the machines the questions your prospects are already asking them, and reading the answers closely - not just whether your name comes up, but why.
Run those four across four different models and you'll usually get four different answers, built from four different sets of sources - which is itself useful information. The overlap tells you who the machines agree is legitimate. The gaps tell you exactly what's missing from the public record.
The tempting fix - and the wrong one - is to generate a large volume of new content and hope the algorithm notices.
Ten thousand automatically generated words about "what to do if you're charged with OUI in Massachusetts" do not replace externally verifiable authority, because the models are specifically trying to filter for content that sounds like it came from a real, specific, checkable person - not a content mill. More words that could have been written by any firm, about any client, in any county, don't move the needle. They can actually hurt, by diluting a site with material that reads exactly like everyone else's.
The rule I use for my own content applies here directly: if it could have been written by any lawyer, any consultant, or a particularly cautious toaster, it is not yet personal enough to count as evidence of who you actually are.
What actually moves an AI's answer is the opposite of generic: real courtroom observations, a specific county's filing data, an actual case-type niche, a voice that sounds like one identifiable person rather than a marketing department. That's a smaller amount of content than the volume approach, and it works because it's the kind of thing a content mill can't fake.
Not a lecture on how AI works. A specific, prioritized answer for your firm, in your counties.
Exactly what each model says for each of your practice areas and courts, side by side.
Who's being recommended instead of you, and what they have on the record that you don't.
Which specific pages, reviews, and citations each answer is actually drawing from.
What to fix first, in order, instead of a vague mandate to "do more content."
This is either the most on-the-nose proof of the whole page, or the least serious. Possibly both.
Somewhere in the writing of this page it became obvious that a bluegrass song titled "Artificial Intelligence" is, itself, a small piece of exactly the kind of public, specific, unmistakably-made-by-one-actual-person content that an AI model treats as evidence a real firm is behind the name. It's absurd. It's also true. Publicly discoverable authority doesn't have to be solemn to count - it has to be real, attributable, and impossible for a content mill to produce by accident.
The audit tests ChatGPT, Claude, Gemini, and Grok against your actual practice areas and courts - and tells you exactly what to fix, in what order.
Start with the audit → Contact Russell Matson