Someone in Austin decides on a Sunday night that the marriage is over. Before they call anyone, they open ChatGPT and type "best divorce lawyer in Austin who does flat fees." The engine gives them two paragraphs: a warning that it can't give legal advice, a plain summary of how Texas divorce works, and then three firm names with a sentence about each. One firm gets described as "publishes flat-fee packages for uncontested divorce." That firm gets the consultation. This is now an ordinary way to find a lawyer: 45% of US consumers used AI tools like ChatGPT for local business recommendations in the past year, up from 6% the year before (BrightLocal, 2026), and legal questions are a natural fit because the asker is embarrassed, anxious, and wants a neutral answer before they talk to a human.
Legal is also the vertical where engines are most careful. They hedge on advice, they add disclaimers, and when they do name a firm they reach for what they can verify: bar admissions, years in practice, the practice areas a site actually describes, the fee model it states, and what reviews say about communication. That caution is good for a firm that publishes specifics and bad for a firm whose site is a stock photo of a courthouse and a phone number. Most firm websites are the second kind.
What law firm questions do people ask AI?
The pattern is practice area plus city plus fee model, in some order. "Personal injury lawyer near me no fee unless we win", "how much does a DUI lawyer cost in Ohio for a first offense", "estate planning attorney in Scottsdale flat fee will and trust", "immigration lawyer near me who speaks Spanish", "employment lawyer in Chicago for wrongful termination free consultation". The fee question is attached to almost everything, because the biggest fear a prospective client has isn't losing; it's not knowing what the lawyer costs until the bill arrives. Firms that state a fee model in text get described by it. Firms that don't get described by whatever the engine can find, which is often nothing.
Before the "who" question there is usually a "do I need one" question: "do I need a lawyer for a first offense DUI in Georgia", "can I file for divorce without a lawyer in California", "is it worth hiring a lawyer for a small claims case", "how long do I have to file a personal injury claim in Florida". The engine answers these with general information, adds a disclaimer, and then, frequently, names a local firm whose site explained the same thing clearly. That's the quiet win: a firm that publishes a plain-English page on the statute of limitations in its state gets named in answers to the question that precedes hiring. And then the diligence family: "is [firm] legit", "reviews of [attorney]", "how do I check if a lawyer is licensed in Texas". Engines answer those from bar records and reviews, so both need to be in order.
Who does AI name for law firms?
The firms with verifiable credentials and specific practice-area pages. When an engine names an attorney it likes to say why: "licensed in Texas since 2009," "focuses exclusively on family law," "offers flat-fee uncontested divorce." Every one of those phrases has to come from somewhere it can read: the state bar directory, Avvo, Justia, FindLaw, Super Lawyers, Google reviews, or your own site. The firms that are consistent across those sources get named with confidence. The firms whose bar record says one office address and whose website says another get hedged or skipped. Large consumer firms with heavy directory presence and thousands of reviews also show up often, for the same reason chains do in every vertical: they give the engine more to quote.
How often does a small firm make the shortlist at all? SOCi's 2026 Local Visibility Index found ChatGPT recommended just 1.2% of local brand locations, versus 11% for Gemini and 7.4% for Perplexity. That study measured multi-location brands, not law firms, so treat it as a signal about how short the list is rather than a legal-specific figure. Yext's analysis of 6.8 million citations found 86% came from brand-managed sources, about 44% owned websites and 42% managed listings, but its sample covered retail, financial services, healthcare, and food service, not legal. Nobody has published a clean measurement of which directories AI cites for law firm queries or how often engines name a solo practitioner versus a large firm. What we see in daily scans is qualitative: engines lean on credentials and reviews, quote fee language when they can find it, and hedge everywhere else.
What should a law firm publish to get cited?
Pages that answer the practice-area, fee, and credential questions in text, on your own domain, in language a model can quote without adding a disclaimer of its own. Check everything against your state bar's advertising rules before it goes live. Six pages carry most of the weight.
- One practice-area page per area you actually take, written for your state: what the matter involves, the typical timeline, what the client has to do, and the questions you hear in the first call. "Family law" as a bullet on a services page is invisible; a real page on uncontested divorce in Texas gets quoted.
- A fees page that states your model per practice area: flat-fee packages with the price and what's included, contingency percentage and who pays costs, hourly ranges, whether the consultation is free and how long it runs. If your bar limits what you can say, say what you can and explain how the rest is set.
- Attorney bios with the facts engines cross-check: bar number, state and federal admissions, year licensed, law school, practice focus, languages spoken, board certifications. Match your bar directory and Avvo entries word for word.
- A plain-English guide per practice area answering the "do I need a lawyer" and "how long do I have" questions for your state, with the statute of limitations, the court process, and when self-representation is reasonable. Honest guides get cited; hard sells don't.
- A results page within your bar's rules: representative matters, outcomes where you're allowed to state them, and the required disclaimers. Engines treat this as evidence of what you've actually handled.
- Review responses on Google and Avvo that address communication, responsiveness, and billing surprises without discussing the matter. Engines quote review exchanges as evidence of how a firm treats clients, and legal reviews are heavy on exactly those themes.
How does a law firm measure AI visibility?
With a fixed set of your market's legal questions, by practice area and city, asked repeatedly on every engine and tracked as the share of answers that name your firm. One search proves nothing: the same prompt can name you on Monday and not on Thursday, and each engine keeps its own list. Profound found ChatGPT and Perplexity overlap on only 11% of the domains they cite, and Ahrefs found only 12% of AI-cited URLs rank in Google's top ten for the original prompt. A firm that spent years winning "Austin divorce lawyer" on Google can be absent from every AI answer to the same question, and the firm in position eight on Google can be the one ChatGPT names.
Measure per engine, over time. Tibly tracks seven: ChatGPT, Perplexity, Gemini, Google AI Overviews, Google AI Mode, Microsoft Copilot, and Claude. On Claude, be honest with yourself: nobody has measured how it recommends local businesses and it has no local-business surface, so watch it as a check rather than a channel. For the rest, the loop is the same one that works in every vertical. Generate the questions clients actually ask in your service-area cities, scan them daily, watch which firms get named and which sources get cited, and publish the page that closes the widest gap. Then watch the share move. If the fees page went live and the "flat fee divorce" prompts still miss you, the page isn't saying what the engine needs. If they flip within a couple of weeks, you know what to write next.
