GEO for Law Firms: How Legal Practices Get Cited by AI Search

ClickRadius Institute · April 6, 2026

A prospective client with a herniated disc and a totaled car no longer starts with ten blue links. Increasingly, she opens ChatGPT or Gemini and types a full sentence: "best personal injury lawyer near me who takes contingency and has trial experience." The AI engine answers in prose, names two or three firms, and cites its sources. If your firm is not one of the entities the engine trusts, you are not in that conversation — and the client never knows you exist. This article explains how Generative Engine Optimization (GEO) works for law firms specifically: the schema markup legal practices need, the entity signals AI engines lean on in a profession where credentials are formally verifiable, the content that earns citations, and how to do all of it without crossing state bar advertising rules.

Legal search is moving to AI faster than most firms realize

AI Overview coverage and click-through behavior vary by measurement source, query cohort, collection date, locale, device, account state, and feature definition. This page does not assert a universal current percentage or traffic effect. Google's May 19, 2026 announcement described a redesigned Search box and a flow from an AI Overview into AI Mode; it did not establish AI Mode as the global default or formally demote ordinary results.

Legal queries are unusually exposed to this shift for one reason: legal questions are natural-language questions. Nobody's real problem is "DUI lawyer Phoenix." Their real problem is "I got a DUI last night in Phoenix, what happens now and how much will a lawyer cost?" Keyword search forced people to compress that into fragments. AI engines let them ask it whole — and the engines answer it whole, citing whichever firms published the most credible, verifiable answer.

An AI engine will not cite the loudest law firm. It cites the one it can verify.

— ClickRadius Institute

How legal clients actually prompt AI engines

GEO planning starts with the prompts, because AI engines retrieve content that answers the question as asked. Real prospective clients type things like:

• "Best personal injury lawyer near me who takes contingency — no upfront fees"
• "How much does a DUI lawyer cost in Phoenix for a first offense?"
• "Do I need a lawyer for an uncontested divorce in Arizona, or can I file myself?"
• "What's the statute of limitations for a car accident claim in my state?"
• "Estate planning attorney vs. doing a will online — when is a lawyer actually worth it?"
• "Employment lawyer who handles wrongful termination and offers free consultations"

Notice the pattern: cost, process, jurisdiction, and qualification criteria ("takes contingency," "trial experience," "free consultation"). These are the exact dimensions most law firm websites refuse to address directly — and they are precisely what generative engines look for when composing an answer. The firm that publishes an honest page titled "What a first-offense DUI defense costs in Phoenix (and what changes the price)" is the firm that gets retrieved when that question is asked. The firm whose site says only "Aggressive representation. Call now." gives the engine nothing to work with.

Schema markup for law firms: LegalService, Person, and evidence-bound relationships

Structured data is how you state facts about your firm in a format machines parse without ambiguity. For legal practices, the relevant vocabulary from schema.org is specific:

The firm: LegalService

Mark up the firm itself as LegalService, a subtype of LocalBusiness. The properties that matter most for AI retrieval:

• areaServed — the cities, counties, and states where you actually practice. Legal answers are jurisdiction-bound; an engine assembling "DUI lawyer cost in Phoenix" needs a machine-readable statement that you serve Phoenix and are admitted in Arizona.
• address, telephone, openingHours — matching your Google Business Profile and bar listing exactly. Inconsistency is read as unreliability.
• makesOffer / hasOfferCatalog — the practice areas as named services (personal injury, criminal defense, family law), each linked to its dedicated page.

The lawyers: stable Person nodes with separately verified facts

Each lawyer bio page should carry a stable Person node. Schema.org Attorney is a deprecated business type under LegalService and LocalBusiness, not a Person type. Bind the lawyer's role, bar admission or license, credentials, practice services, jurisdictions, office locations, education, memberships, and profiles to that exact person as separately sourced facts. knowsAbout may state a topic relationship such as personal-injury law or Arizona DUI defense, but knowsAbout does not imply expertise, skill, or a credential. A bar record or other authoritative issuer must establish the regulated fact; a sameAs link merely identifies a related profile and does not make every claim on either page corroborated.

Entity signals: the legal profession's verification advantage

Many lawyer credentials can be checked against issuing authorities, but the lawyer, firm, office, jurisdiction, admission, discipline record, court credential, role, and evidence date must remain separate. The sources below may corroborate exact first-party claims when their authority and coverage are established. Do not call every database ground truth or claim that an external engine used it; public verification is not causal proof of a recommendation or citation.

• State bar profiles. The foundational entity record. Make sure the name, firm name, and practice information on your bar profile match your website byte-for-byte where possible.
• Avvo, Justia, and Martindale-Hubbell. The three legal directories AI engines encounter most often when resolving a lawyer as an entity. Claim all three, complete them fully, and keep practice areas consistent across them.
• Court admissions and reported cases. Federal court admissions, appellate work, and published opinions are court-record-grade credentials. List them specifically ("admitted to the U.S. District Court for the District of Arizona") rather than generically ("admitted to federal courts").
• Peer signals. Bar association leadership, CLE teaching, and legal publication authorship are third-party evidence of expertise that engines weigh far more heavily than self-description.

No market-wide AI-answer prevalence is established here. Measure the business on a fixed, dated prompt set and record the engines, locale, device or account state where relevant, and whether each result is a mention, linked citation, or recommendation. That sample can identify an observed gap; it does not establish that a category is empty, a citation position is unclaimed, or an early entrant will become a default answer.

Citable expertise: what a law firm should actually publish

According to the Princeton-led study "GEO: Generative Engine Optimization" (Aggarwal et al., presented at KDD 2024), three content signals affected the selected source's relative visibility in the study's generated responses: quotations, statistics, and source citations. The researchers reported visibility improvements of up to roughly 40% for content optimized along those lines. For a law firm, that translates into a content program that looks less like advertising and more like a legal explainer library. The benchmark began with the top five Google sources, randomly selected one source for optimization, and measured that source's relative visibility/impression in generated responses; it did not measure retrieval or citation-inclusion probability, and reported effects varied by domain.

• One page per real question, per jurisdiction. "How long do I have to file a car accident claim in Arizona?" deserves its own page with the actual statute cited (A.R.S. § 12-542, two years), not a paragraph buried in a practice-area brochure. Citing the statute is a source citation — one of the three Princeton signals — and it is something a generic content mill will get wrong.
• Honest cost pages. A page explaining that first-offense DUI defense in your market typically involves a flat fee within a stated range, what raises it (aggravating factors, trial), and how contingency percentages work in injury cases. Ranges and factors, not promises.
• Process explainers with steps. "What happens after you're arrested for DUI in Arizona: the first 15 days" — numbered, dated, specific. Engines lift stepwise content readily because it maps to how they structure answers.
• Practice-area topical depth. An engine deciding whether your firm is "an authority on Arizona family law" is effectively asking whether your site covers the topic's question-space: custody, relocation, support modification, prenups, enforcement. Ten interlinked, well-cited pages on one practice area beat forty shallow pages across six.

In legal marketing, the bar rules and the AI engines finally want the same thing: provable facts, plainly stated.

— ClickRadius Institute

Compliance: doing GEO inside state bar advertising rules

Legal advertising is regulated, and GEO content is advertising. The good news is that the rules and the optimization actually point the same direction. Three constraints matter most (this is general education, not legal advice — check your own state's rules):

• No outcome guarantees. Nearly every state prohibits statements that create unjustified expectations — "we'll win your case," "guaranteed results." This is also terrible GEO: engines discount unverifiable promises. Replace them with verifiable process facts and documented credentials.
• "Specialist" restrictions. Most states restrict "specialist," "expert," or "certified" to lawyers holding accredited certifications (for example, ABA-accredited or state-board specialty certification). If a lawyer holds one, bind the exact certification, issuer, scope, status, and holder. If not, use only language permitted in the applicable jurisdiction, such as "focuses on" or "practices in." A knowsAbout value is a topic relationship and cannot carry an expertise or certification claim.
• Past results need disclaimers and context. If your state permits publishing verdicts and settlements, publish them with the required disclaimers and specifics. Vague "millions recovered" claims are both a compliance gray zone and weak GEO; specific, disclaimed, dated results are stronger on both counts.

Old legal SEO vs. legal GEO

A 90-day GEO plan for a law firm

Days 1–30: entity foundation. Audit the firm's entity footprint: state bar profiles, Avvo, Justia, Martindale, Google Business Profile. Resolve each apparent inconsistency against the source that governs that exact fact. Deploy LegalService markup on the firm page and a stable Person node on every lawyer bio; bind admission, role, credential, jurisdiction, location, service, and profile relationships separately. Baseline what the five supported engines actually return for a fixed prompt set. ClickRadius records that sample monthly; its separate five-family, 0–100 AI Readiness Score estimates observed website inputs and does not prove an external-engine outcome.

Days 31–60: the question library. Pick one flagship practice area. Write 8–12 question-level pages covering its cost, timeline, process, and jurisdiction questions, each with statute or court-rule citations, at least one statistic, and FAQ markup. Route internal links from the practice-area hub to each answer page.

Days 61–90: corroboration and monitoring. Pursue off-site witnesses: complete directory profiles, a bar-journal article or CLE listing, local press or chamber mentions that confirm the firm's identity and focus. Begin monitoring AI-engine citations weekly so you can see which pages get retrieved and extend what works. ClickRadius runs this citation monitoring on a monthly schedule across the five live engines, whether you buy it direct at $499/month or through a white-label agency.

Frequently asked questions

Can a law firm do GEO without violating state bar advertising rules?

Yes. GEO for law firms is built on verifiable facts — bar admissions, practice areas, court admissions, published legal explainers — not on outcome promises. Because most state bar rules prohibit guarantees of results and restrict words like "specialist" or "expert" unless the lawyer holds an accredited certification, a GEO program that sticks to documented credentials and educational content is generally safer than traditional advertising copy, though every firm should review its own state's rules.

Which schema.org types should a law firm use?

Use LegalService for the firm and a stable Person node for each lawyer. Schema.org Attorney is a deprecated business type under LegalService and LocalBusiness, not a Person type. Bind each lawyer's verified role, admission or license, credential, service, jurisdiction, and location separately. knowsAbout can record a topic relationship, but it does not imply expertise, skill, or a credential; first-party topics and third-party records therefore remain distinct evidence.

How long does it take a law firm to appear in AI answers?

There is no fixed timeline, and no honest vendor guarantees one. Firms typically see the fastest movement on specific, low-competition question queries in their own jurisdiction — the cost and process questions — within the first few months of publishing structured, well-cited answer pages, while broad head terms like "best personal injury lawyer" take sustained entity building over a longer horizon.

Want an inspectable baseline for your firm? Get your free AI Readiness Score — an estimate of five readiness-evidence families observed at the submitted URL — or see plans and pricing for guided work and separate monthly AI Presence samples. The score is not a measurement of whether an engine currently cites the firm.