AI visibility for law firms

For twenty years, legal marketing meant winning a Google results page: rank for the practice area keyword, buy the ads above it, collect the click. That funnel still exists, but a growing share of clients now starts somewhere your rank tracker cannot see. iLawyer Marketing's 2026 consumer survey found that Google usage for attorney research dropped 15 points while willingness to use ChatGPT climbed to 41.9%, and 9.5% of consumers said they would research attorneys using only AI sources, skipping Google, Facebook, Yelp and YouTube entirely.

The shape of the session changes too. A person with a legal problem rarely opens ChatGPT and types a lawyer's name. They describe the problem: a crash, a termination, a custody dispute, an estate that needs settling. The assistant explains the legal landscape, and somewhere in that conversation it names firms worth contacting. Clio's 2025 Legal Trends Report found that among consumers who asked AI a legal question, 28% were told to contact a lawyer. That handoff moment is the new referral, and it goes to whichever firms the engine can name with confidence.

The trust curve is moving the same direction. In the iLawyer survey, 53.5% of consumers said they trust AI more than they did a year ago, against 16.4% who trust it less. That is better than a three to one ratio in favor of rising trust, in a purchase category where trust is the entire product. Firms that treat AI answers as a fringe channel are reading last year's map.

The lazy assumption is that AI research is a Gen Z habit and the clients with real matters still call the firm a colleague recommended. The 2026 data says the opposite. In iLawyer Marketing's survey, consumers aged 45 to 60 lead AI adoption for legal research: 57% of that bracket would use ChatGPT to research lawyers and 76% would use AI to research law firms, higher than any other age group including 18 to 29 year olds, who came in at 35% and 48%.

That bracket is where the valuable matters live. Estate planning, business disputes, serious injury claims, divorces with property at stake: the clients who can pay full rates are the ones most likely to ask an AI assistant who to hire. Clio's consumer data points the same way, with 14% of consumers already having asked AI a legal question and another 43% saying they would. The pipeline of AI-first legal consumers is more than half the market and it skews toward the demographic that signs larger engagement letters.

For a managing partner, the practical question is simple. If a 52 year old business owner in your city asks ChatGPT for a commercial litigation firm, does the answer include you? That is an empirical question with a checkable answer, and most firms have never checked it. Reachroller's free checker on the homepage runs three of those questions against ChatGPT in about a minute, which is usually enough to establish whether there is a problem.

AI engines answer from two layers. The first is training data: what the model absorbed about your firm from the public web before its cutoff. The second is live retrieval: when ChatGPT runs a web search mid-answer, pulls current pages, and composes a response from them. Training data moves on retraining timelines nobody can schedule. Retrieval moves in weeks, because it reads pages you can publish and sources you can influence today.

For legal queries, the retrieval layer leans on a recognizable diet: legal directories such as Avvo, Justia, FindLaw and Super Lawyers, bar association listings, review platforms, local news coverage, and firm websites that answer questions in plain language. An engine composing an answer to a personal injury question in Phoenix will look for pages that connect a firm name to that practice area and that city with corroboration from more than one source. Firms that exist as a consistent entity across those sources get named. Firms whose footprint is a homepage and a phone number get skipped, however good their trial record is.

There is also a structural quirk worth knowing: an engine can only recommend what it can disambiguate. If your firm's name is three surnames that also belong to a firm in another state, or your site describes services in insider language a consumer would never type, the engine may fail to connect the question to you at all. Entity consistency, meaning the same name, address, practice list and attorney bios everywhere, is unglamorous work that directly moves AI answers.

Legal marketing operates under real constraints. ABA Model Rule 7.1 and its state equivalents prohibit false or misleading communications about a lawyer's services, and state bars regulate claims about results, specializations and comparisons. None of that disappears because the surface changed from a billboard to a chatbot answer. Content you publish to win AI citations is attorney advertising in most states, and it should go through the same review as anything else the firm publishes. That is a compliance conversation for your own counsel, and it is worth having early rather than after fifty pages ship.

The constraint cuts both ways, and the second edge favors diligent firms. Legal questions sit squarely in what search quality guidelines call YMYL territory, where engines hold sources to a higher bar. AI assistants hedge on legal topics, disclaim that they are giving general information, and prefer citing sources that look authoritative and verifiable. A firm page written in plain language, attributed to a named attorney with bar admissions listed, with no outcome guarantees and honest scope, is precisely the source profile engines prefer for legal answers. The bar rules that stop firms from overclaiming happen to describe citable content.

One more compliance-adjacent reason to monitor AI answers: engines sometimes get firms wrong. Wrong practice areas, attorneys who left years ago, offices that closed, or a confident recommendation of your firm for work you do not take. You cannot correct what you have not seen. Tracking what AI actually says about your firm, in stored answers you can read, is the only way to catch errors before a prospect or a disciplinary complaint does.

The pages that earn AI citations in legal categories share a shape. They answer one question a real client asks, in the first paragraph, in language the client used. They carry specifics: fee structures explained, timelines by matter type, what happens at each stage, when a case is worth pursuing and when it is not. The original Princeton GEO research found that adding statistics, quotations and cited sources lifted visibility in generative answers by up to around 40%, while keyword stuffing performed below baseline. For law firms this means the classic thin practice-area page, five hundred words of keyword-laced generalities, is the exact format engines skip.

Fee transparency deserves special mention because it is the question clients ask AI most freely. People who feel awkward asking a lawyer what a divorce costs feel zero awkwardness asking ChatGPT. A firm that publishes an honest page on how its fees work, what a contingency percentage typically covers, or what drives an hourly estate plan quote, becomes the source an engine quotes when that question arrives. Most firms refuse to publish this, which means the few that do collect an outsized share of the answers.

Then there is the off-site record. Engines corroborate. A firm claimed as excellent only by its own website is a weak signal; the same firm with consistent directory profiles, a steady review base, bar association involvement and occasional local press coverage is a strong one. Digital PR for a law firm does not require national headlines. A quote in a local business journal about a new employment law, a bar journal article, a community sponsorship with a news mention: these are the citations that teach engines your firm is real, active and local.

The naive check is to ask ChatGPT about your own firm by name. It will say something flattering, because engines mention the brand you asked about by construction. Honest measurement uses unbranded questions, the kind a stranger with a legal problem actually types: best employment lawyer for a wrongful termination case in Denver, how to choose a probate attorney, whether a contingency fee is worth it. If your firm appears in those answers, you are winning demand you never had to buy.

AI answers are also probabilistic. The same question asked twice can name different firms, so a single run proves close to nothing. Sound methodology asks a fixed set of questions on a schedule, stores every answer, and reads the trend rather than any single result. Reachroller was built around exactly this: it tracks ChatGPT live through the official API with web search, with Claude, Gemini, Perplexity and Grok built and rolling out, and its scoring is evidence-grounded, meaning your firm only counts as mentioned when its name literally appears in the stored answer text. There is no inferred credit and nothing you cannot verify by reading the answer yourself.

For a firm, the tracked question set writes itself from the intake log. Take the matters you most want more of, phrase each as the question a prospect would ask an assistant, add your top competitors for comparison, and run the set weekly. Within a month you have something legal marketing has rarely had: a defensible baseline of where referrals-by-AI go in your market, and a list of the specific questions where they go to someone else.

Legal keywords are famously among the most expensive clicks in paid search, and every click stops the moment the budget does. AI visibility compounds differently: a page that starts getting cited keeps getting cited, and a corrected entity footprint keeps paying without a daily spend. The cost of finding out where you stand is now trivial next to a single practice-area campaign.

Reachroller's free checker answers three questions at no cost, and the 3-day trial includes 50 credits with no card required, where one credit buys one stored AI answer. The Starter plan is $29 per month with 400 credits and 25 tracked questions, which covers a solo or small firm tracking its core practice areas weekly. Growth at $99 per month tracks 75 questions with 1,500 credits and API access, and Agency at $249 per month adds 200 questions, 4,000 credits, 10 workspaces and white-label reporting for the marketing agencies that run legal accounts. When a tracked question shows you losing, a generated fix page costs 10 credits and gives you a compliance-reviewable draft targeted at that exact question.

The sequencing matters more than the spend. Run the baseline first, because it converts AI visibility from an anxiety into a punch list. Most firms discover they are visible for two or three questions they never optimized for and invisible for the ten that drive their best matters. That list, plus the playbook below, is the whole program.