GEO for Law Firms: AI Citations and Google Profiles in 2026

According to Sistrix (April 2026), 58% of Google queries in France now trigger an AI Overview, and divorce lawyer Paris is one of them. A potential client typing their question into ChatGPT or Google AI gets a synthetic answer with two or three firms cited, and that is it. For a managing partner investing yearly in SEO, the question is no longer am I ranking first but am I in the AI answer. This playbook covers legal-specific issues: professional conduct, Attorney schema, authority signals recognized by LLMs, and common mistakes firms make when transposing SEO reflexes without adapting to GEO. No first-result promise. A public methodology and a conduct-compatible framework.

The legal market is one of the most exposed verticals to the SEO-to-GEO transition. A search for divorce lawyer or employment attorney in 2026 rarely starts with a direct Google query. It starts with a conversational question in ChatGPT, Claude or the AI Overview-enhanced Google bar. The AI answer cites two or three firms, and the competition for those spots now runs on signals most firms do not measure.

Why legal GEO differs from legal SEO

Legal SEO optimizes for ranking. Legal GEO optimizes for citation. The difference is concrete: a firm can rank third on Google for divorce lawyer Boston and be absent from Google AI Overview on the same query. According to Vercel and MERJ (500M GPTBot fetches analyzed), AI bots crawl differently than Google bots and weight different signals.

SEO rewards position, dense internal linking and freshness. GEO rewards factual clarity, verifiable authority and answer structure. A firm publishing a 4000-word unstructured employment law page can dominate Google and never be cited by Claude, because the extractable passage simply does not exist.

Point 1: professional conduct rules (ABA Model Rule 7.1 in the US, RIN article 10 in France) prohibit false or misleading communications. This excludes some aggressive SEO practices but actually favors AI citability, which rewards factual sobriety. Point 2: the Attorney schema imposes verifiable fields (bar admission, certified specializations) that exactly match the authority signals LLMs seek to extract.

Attorney schema: the non-negotiable technical foundation

Schema.org Attorney type (or LegalService depending on the case) is the first criterion ScoreGeo checks on a lawyer page. According to Ahrefs (March 2026, 1885 pages tested), pages with valid JSON-LD are 5.8 times more often cited by AI than pages without. For a law firm, the ratio is likely higher, because LLMs explicitly seek to validate professional credentials before citing.

Critical Attorney fields: name, jobTitle (partner, associate, of counsel), worksFor (the firm), areaServed (jurisdictions), knowsAbout (practice areas), hasCredential (admitted to New York bar 2014, certified specialist in family law 2019). These fields build a structured profile ChatGPT and Claude can extract and restate without hallucination.

The common mistake is filling hasCredential with vague titles like 15 years of experience or recognized expert in criminal law. These formulations are both professionally risky and technically useless. An LLM will not cite recognized expert without a source. It will cite admitted to the California bar since 2008, certified specialist in employment law, because it is verifiable.

Answer-first format for practice area pages

A practice area page (employment law, family law, business law) must open with a self-contained 2 to 3 sentence answer to the main question a potential client types. The inverted pyramid format is the GEO norm: answer first, context second, development last. LLMs weight the first 200 characters disproportionately.

For an uncontested divorce page, open directly with: An uncontested divorce in California is available when both spouses agree on all terms (property, custody, support). The filing fee is approximately 435 dollars plus attorney fees averaging 1500 to 3500 dollars per spouse. Typical timeline 6 to 12 months from filing to final judgment, faster if no contested issues arise during waiting period.

This opening gives ChatGPT and Claude all the factual elements they will look to cite. Classic development (procedure, exceptions, case law) comes after. To go deeper on answer-first mechanics, our guide to common GEO mistakes details the pitfalls to avoid.

Google Business Profile and directories: consistency before quantity

A law firm must appear strictly identically on 4 minimum surfaces: Google Business Profile, the bar directory, Avvo or Justia (or Martindale-Hubbell), and the firm website. NAP (Name Address Phone) consistency is a strong authority signal for LLMs, which cross-reference these sources to validate firm existence.

Common divergence: the website lists 555 Main Street New York, Google Business Profile lists 555 Main St. Suite 1200 NY 10001 and Avvo lists Smith Law Firm 555 Main. For a human reader it is the same firm. For an LLM weighting authority signals, it is three potentially distinct entities, and citation becomes uncertain.

Point 1: adopt a canonical NAP format and deploy it everywhere, character for character. Point 2: add opening hours, main phone and contact URL inside the LegalService schema output. Point 3: verify Google reviews are coherent with displayed specialization (a firm presenting as employment law with 80% reviews on criminal cases sends a contradictory signal).

Structured legal glossary: a citation multiplier

A 30 to 50 entry legal glossary (civil procedure, employment law, criminal law, etc.) in FAQPage schema is one of the most efficient content assets for multiplying AI citations. Each definition is a short self-contained answer, ideal for LLM extraction, and each page naturally links to the corresponding practice area.

According to Semrush (analysis of 150k ChatGPT citations), structured FAQ pages represent a disproportionate share of AI citations in professional verticals. For a firm, this means a page defining wrongful termination can generate more AI visibility than a generic 3000-word employment law page.

Questions to prioritize are those a non-lawyer client would type: What is the difference between layoff and wrongful termination, How much does uncontested divorce cost, How long does a personal injury case take. These formulations match the conversational queries LLMs relay.

To structure this foundational work, a manual GEO audit of the 50 most citable questions of the firm is the high-leverage investment. That is exactly what our GEO accompaniment for law firms covers, including mapping the 30 priority queries and deploying FAQPage schema on target pages.

Common law firm mistakes in GEO

Three mistakes systematically recur when a firm transposes its SEO reflex to GEO. First mistake: overloading pages with keywords (cheap divorce lawyer Boston best firm) at the expense of readability. LLMs heavily penalize keyword stuffing in extraction. Second mistake: multiplying near-duplicate pages for each city (divorce lawyer Boston, divorce lawyer Cambridge, divorce lawyer Somerville).

Third mistake: using unverifiable promotional formulations (best firm, recognized expert, leader in employment law) that are both professionally risky AND technically counterproductive. An LLM prefers citing 12 years of experience in employment law (verifiable via the bar) over recognized expert (unverifiable and unsourceable).

For a firm starting in GEO, following the ScoreGeo methodology and auditing current citability is the prerequisite. The score on 100 points indicates precisely where the gaps are (missing schema, non answer-first format, NAP inconsistency, missing llms.txt, etc.) and which corrections to prioritize.

Monitoring and measurement: what gets measured gets managed

GEO monitoring for a law firm rests on 15 to 20 target queries tested monthly on ChatGPT, Claude and Google AI Overview. For an employment law firm in Chicago: employment lawyer Chicago, severance negotiation timeline, wrongful termination damages calculation, etc. The result tells you whether the firm is cited, in what position, and with which passage.

This measurement gives a raw but direct read on AI visibility. A monthly variation of plus or minus 20% is normal (LLMs have intrinsic variance). A persistent absence over 3 consecutive months signals a structural defect: missing schema, insufficient authority, or non-extractable content.

For a firm wanting to industrialize this monitoring, the ScoreGeo monthly tracking offer covers 15 queries tested on 3 LLMs, the citation report compared to identified competitors, and a monthly call to adjust priority content.

Frequently asked questions