Nobody's first move is "find me an attorney." It's "can my ex take the house if it was mine before we married" or "do I need a lawyer for a first-offense OWI." The assistant answers — and then, unprompted, often names who handles that kind of matter locally.
That name is the referral. It happens before anyone visits a website.
See whether your firm gets named — freeNo account, no card.
The Structural Problem
This is the difference that separates legal from every other local category we work in. Ask an assistant for a plumbing company and a company name is a complete, satisfying answer. Ask it for a divorce attorney and a firm name is only half an answer — the implicit question is which lawyer there, and are they actually licensed to handle my matter in my county?
Most small-firm websites can't answer that. There's a Practice Areas page covering five areas in three paragraphs, an Attorneys page with headshots and two-sentence bios, and a contact form. Nowhere on the site is it machine-readable which named individual handles which matter type, in which courts, admitted since when. So the engine has a firm, a general subject area, and a city — and no way to connect them into a recommendation it can stand behind.
The firms that get named have solved this at the individual level, not the brand level.
Legal queries carry a shape you don't see in other verticals: they combine a situation, an eligibility test, and a jurisdiction. The jurisdiction is usually a county or a state, not a city, because that's the unit that actually governs the answer.
Note how few of these contain the word "best." Most are informational first. The firm that gets named is the one whose own site answered the informational half well enough to be the source.
The arithmetic is unforgiving but simple. One practice area in one jurisdiction is one page. Five practice areas across two counties is ten pages, before attorney bios, fee pages, or process explainers. A single Practice Areas page is one page competing for ten separate questions — and it loses all ten to whoever wrote the ten.
Corroboration
In the trades, the directories are lead marketplaces: commercial, pay-to-play, and treated by AI engines with corresponding caution. Legal is different. Avvo, Justia, FindLaw, Martindale-Hubbell, and Super Lawyers are built on bar admission records and peer review. A Martindale rating or a state bar record is a licensing fact, not an ad buy, and engines weight it accordingly.
That cuts both ways. A firm whose directory records agree with its own website gets identity verification most local businesses can never assemble. A firm whose records disagree — a since-departed partner still listed, a former office address, a name that changed after a merger and updated in three places out of six — hands the engine a reason to hesitate. In a one-answer format, hesitation is the same as exclusion.
Google Business Profile, Yelp, Bing Places, Apple Maps, Foursquare, the BBB, and Facebook — checked for whether the firm name, office address, and phone match each other and match your site.
One caution worth stating plainly, because it's a real failure mode: a directory a business pays to appear in is not independent corroboration in the way a bar record is, and we don't score it as though it were. Where a distinction like that matters to your result, the report says which kind of source a finding came from.
The Constraint
Every marketing vendor a law firm has ever hired has run into the same wall: state advertising rules restrict superlatives, restrict "specialist" and "expert" language absent certification, restrict comparative claims, and prohibit anything implying a guaranteed outcome. Firms tend to experience this as a handicap.
For AI visibility it's closer to an advantage. Unverifiable superlatives are exactly what an AI engine discounts — a page insisting it's the best is weaker evidence than a page stating a checkable fact. The claims that survive both the rules and the engine are the same set:
Those are facts an engine can restate without exposure, which is precisely what makes them the ones it restates.
Connexion AI is not your counsel and does not advise on legal ethics compliance. Everything produced for a firm is reviewed by that firm before it's published, and your own state's rules govern. We flag language that looks like a superlative or an outcome claim, because it's bad for both reasons — but the compliance call is yours.
We put your clients' actual questions to Claude, ChatGPT, Gemini, and Perplexity, record who gets named, and score the result 0–100 — 0–24 Invisible, 25–49 Emerging, 50–74 Present, 75–100 Dominant. What's specific to legal:
Whether the site declares Attorney or LegalService rather than a bare Organization block, and whether individual practitioners carry Person markup. Both are scored; most small-firm sites have neither.
How many of your practice-area/county combinations have a dedicated page versus a bullet on a shared page. This number is usually the largest single gap in the report.
Whether Avvo, Justia, FindLaw, Martindale-Hubbell, and Super Lawyers describe the same firm, the same attorneys, the same address, and the same phone as your website does — and which specific records disagree.
Whether the site carries FAQ content and FAQPage markup answering the informational half of a legal query. This is the highest-leverage missing piece on most firm sites, because the informational question is what gets asked first.
Usually from the legal directory layer rather than from firm websites. Avvo, Justia, FindLaw, Martindale-Hubbell, and Super Lawyers carry structured, per-attorney records tied to bar admission data and peer ratings, and engines treat that kind of licensed, third-party-verified record as unusually strong corroboration. A firm site that lists practice areas but never establishes which licensed individual practices them, in which court, since when, gives an engine nothing comparable. So the answer comes from the directory record — and the firm named is whichever one that record points at.
Those profiles are why the directory can answer the question — not why your firm gets named instead of the directory. Two things separate a listed firm from a cited one. First, the directory record and your site have to describe the same entity: same firm name, same address, same phone, same attorney names, because a mismatch reduces confidence that they're the same practice. Second, your own domain has to answer the underlying question in more depth than a profile page can, at the level of one practice area in one jurisdiction. We check the first mechanically and scope the second as content work.
Both, and the attorney-level pages are the ones usually missing. Legal questions put to an assistant almost always want a named person qualified to handle this matter in this place. Schema.org supports Attorney and LegalService as entity types and Person markup for individual practitioners, and we score all of them. An attorney page carries the details that make a recommendation defensible: bar admissions and the year of each, courts admitted to, the matter types handled, education. Those are facts an engine can restate without risk — which is what makes them citable.
They shape it, and here the constraint and the strategy point the same way. Advertising rules in most states restrict superlatives, specialist and expert claims without certification, comparative claims, and anything implying a guaranteed outcome — and those are exactly the claims AI engines discount, because they're unverifiable. What performs is verifiable specificity: jurisdictions, admission dates, matter types, published fee structures, consultation policy. Connexion AI is not your counsel and doesn't advise on ethics compliance; content is reviewed by your firm before publication and your state's rules govern.
In the model we work from, a practice area in a jurisdiction is one page — so five areas across two counties is ten pages, before attorney bios, FAQ pages, or fee and process pages. That sounds like a lot until you compare it to what a winning local business actually publishes: our April 2026 research found the consistently-cited business in a competitive local market carried more than twenty individual service pages plus separate pages per town. Depth is the mechanism.
The query grammar and the directory ecosystem change completely between categories. Three more we've written up:
Why the answer is Angi, and how the contractor who beats it is structured.
The three-identity problem, and the insurance question nobody answers.
Competing with Zillow for an answer Zillow can't actually give.
Already paying an SEO retainer and wondering how this is different? AEO versus traditional SEO, side by side. Agencies with several firms as clients can deliver this under their own brand — see the white-label program.
We run your practice areas and your county through all four engines and report who gets named. Free to start — no credit card.