How to Do GEO for Law Firms in AI Search: Practice Pages, Directories, Reviews
GEO for law firms across three surfaces: which legal directories AI engines actually cite, how to rebuild practice-area pages and what reviews do not do.
Quick answer
Canlah AI treats GEO for law firms, in this guide, as the work of getting a firm named accurately when a potential client asks an AI engine for a lawyer, run through three surfaces: practice-area pages, legal directories and reviews. Forward Push’s continuous AI visibility study, published August 17, 2026, logged 30,319 source citations inside 6,065 verified answers across 15 United States cities and 25 practice areas, and 97.8% of those answers cited at least one legal directory. Two directories carry most of that weight: Super Lawyers took 15.65% of all citations and Justia 13.22%, which is 33.05% and 27.93% of the 14,353 directory citations recorded.
Canlah AI supplies the measurement layer under that work: the 4, 8 or 16 buyer questions in the client’s tier, probed on the AI platforms in that tier (Starter covers ChatGPT, Growth adds Google AI Overviews, Full adds Google AI Mode and Gemini) and re-run monthly, with raw answers, sources and timestamps archived so a managing partner can re-run them. This guide covers execution, not vendor selection; the agency shortlist sits in the rankings for Singapore law firms and Australian law firms.
Treat any single answer as directional. One run cannot prove share of voice, causation or future recommendations. Freeze a panel of ten client questions and repeat it by engine, practice area and jurisdiction before changing anything on the site.
Disclosure: Canlah AI publishes this guide and offers the free AI visibility report linked below. Third-party figures come from public studies and product pages reviewed on September 23, 2026, and were not independently re-run. Nothing here is legal advice on advertising or professional conduct rules.
What does GEO for a law firm measure?
A law firm GEO baseline records six things for every generated answer, because an answer is not a rank and changes by question, engine, jurisdiction and date.
- Mention: whether the answer names the firm, a named partner or neither.
- Recommendation position: whether the firm is the lead suggestion, one of a shortlist of three to five or a passing reference.
- Citation: whether the answer links to the firm’s own page, a directory, a court page, a competitor firm or no visible source.
- Accuracy: whether practice areas, bar admissions, jurisdictions, offices, fee structure and named lawyers are described correctly.
- Competitor context: which other firms appear for the same question and which sources support them.
- Conduct exposure: whether the sentence the engine quoted would survive review as an advertisement under applicable conduct rules.
The last item separates legal work from most categories. An engine quotes whatever it finds, so a superlative left on a practice-area page in 2019 becomes a live claim inside an AI answer in 2026.
GEO, AEO and AI SEO for law firms are three names for one job
GEO for lawyers, AEO for law firms and law firm AI SEO are three vendor labels for one deliverable: a firm that is retrievable, quotable and factually correct inside generated answers. Vendors slice the terms differently. Nifty Marketing’s law firm GEO page presents generative engine optimization as one category with AEO, AI SEO and LLM optimization. Rankings.io’s AI SEO page separates them: GEO earns citations inside AI answers and AEO wins direct-answer placements such as Google AI Overviews, while AIO covers both.
The distinction matters only for reading proposals. The deliverables are the same in all three: which questions were asked, which engines answered, which sources they cited and what changed after the work.
Which sources AI engines cite when someone asks for a lawyer
Legal directories appeared in 97.8% of the 6,065 AI answers about hiring a lawyer in Forward Push’s study published August 17, 2026, and two of them, Super Lawyers and Justia, took 60.98% of the 14,353 directory citations. The table lists the eight directories that study ranked from 30,319 citations. Outside the table, FindLaw appeared six times and Martindale-Hubbell three times; Avvo, eighth in the table, appeared 39 times.
Legal directory share of AI citations, Forward Push AI visibility study published August 17, 2026.
| Rank | Directory | Share of all citations | Share of directory citations | What a firm controls there |
|---|---|---|---|---|
| 1 | Super Lawyers | 15.65% | 33.05% | Peer-nominated selection; profile and firm URL editable once listed |
| 2 | Justia | 13.22% | 27.93% | Free and paid profiles: practice areas, admissions, fees, publications |
| 3 | Best Law Firms | 5.49% | 11.59% | Firm-level research submission and client references |
| 4 | Best Lawyers | 4.23% | 8.92% | Peer-review ballot; facts editable once recognised |
| 5 | BCG Attorney Search | 4.20% | 8.87% | Recruiter-published firm pages; check the facts it repeats |
| 6 | Chambers | 2.05% | 4.34% | Submission and referee process, mainly corporate and disputes |
| 7 | LawInfo | 0.94% | 1.98% | Paid profile |
| 8 | Avvo | 0.13% | 0.27% | Claimable profile, ratings and Q&A |
Source: Forward Push, “Legal Directories Ranked by AI Citation Weight”, reviewed September 23, 2026. Shares measure ChatGPT, Perplexity, Gemini and Claude answers in United States markets, not a global figure and not an endorsement of any directory.
Directory dependence moves by practice area. Family law, appellate, construction, intellectual property and insurance defence answers cited a directory 100% of the time in that study, while personal injury was lowest at 89% and civil litigation at 89.1%.
ChatGPT and Google AI Mode read different shelves for legal questions. Canlah AI’s citation study of September 23, 2026 took 60 buyer questions through ChatGPT and Google AI Mode, logging 526 citations across 92 cited pages. In its family law slice of 12 questions and 112 citations, ChatGPT drew 78% of its citations from government and judiciary pages and cited law firm sites twice, while Google AI Mode drew 73% from law firm pages.
What each engine cited for legal questions, Canlah AI citation study, September 23, 2026.
| Question type | What ChatGPT cited most | What Google AI Mode cited most | What the firm can own |
|---|---|---|---|
| Procedure and eligibility | Court and government process pages | Firm guides plus government pages | A guide stating the rule, then the firm’s role |
| Cost of a matter | Government fee schedules, then firm fee pages | Firm fee pages and cost guides | An exact fee table stating the firm’s own conditions, not a market-range comparison |
| Which firm to hire | Third-party rankings only: Doyle’s Guide, The Legal 500, Chambers | Firm practice-area pages and map results | Directory profiles and the practice-area page |
| What the law says | Statute repositories and judiciary pages | Firm explainer articles | A dated explainer citing the statute by section |
Source: Canlah AI citation study, 60 queries, two engines, 526 citations, September 23, 2026. The legal slice covers Singapore family law questions only, so the cited domains will differ by jurisdiction and the engine split has not been tested outside Singapore.
Read together, the two datasets say one thing. For “which firm should I hire”, the firm’s own website is rarely the evidence; directories and rankings are. For procedure and cost questions the firm’s own pages can win, but only against government sources answering the same question more plainly.
How to do GEO for law firms: six steps
Canlah AI runs law firm GEO in six steps, in this order: crawl access, one page per client question, answer-first sentences, one entity record, directory budget and a fixed measurement panel. Crawl access comes first because a firm that blocks the AI crawlers wastes the other five steps.
Step 1: Confirm the AI crawlers can reach the firm’s site
Check that robots.txt does not disallow OAI-SearchBot, PerplexityBot, ClaudeBot or Google-Extended, then check the CDN and the website platform separately, because bot blocking is often on there by default. Law firm sites often run strict security settings chosen for client confidentiality, which is where the blocks usually come from.
Step 2: Give each client question its own practice-area page
Write one page per question a client actually types, rather than one long page per practice area. Canlah AI’s citation study found 15 of the 18 legal pages examined were quoted at a paragraph’s first sentence; the one counter-example opened by calling the matter “a complicated matter” and was ignored, while its FAQ, which answered directly, was quoted instead.
Step 3: Put the answer, the fee and the timeline in the first sentence
State the outcome in the opening sentence of each section, with the number in it. The strongest legal page in the Canlah AI sample put a direct answer and a fee table under one question-form heading, “How Much Does a Divorce Cost in Singapore?”, stating the firm’s fixed fee and the conditions attached to it. Both engines cited it. For a regulated practice such as a law firm, publish the firm’s own fee and its conditions only, never a market range or a comparison against other firms’ pricing: Singapore law firms sit under the Legal Profession (Professional Conduct) Rules 2015 (r.43(1)(c), r.44(2)(c)), which restrict publishing comparative fee information, and every other jurisdiction sets its own equivalent rule. An unregulated business can still split the sentence two ways: one exact own-fact statement for ChatGPT, one sourced market range “depending on A, B and C” for Google AI Mode.
Step 4: Make the lawyer entity record identical everywhere
Firm name, offices, phone numbers, bar admissions, jurisdictions, practice areas and the spelling of every lawyer’s name should match across the website, Google Business Profile and every directory profile. Engines resolve who a lawyer is before naming the firm, and a partner listed under two different admission years across three profiles gives the engine two candidate entities instead of one.
Step 5: Spend the directory budget where the citations are
Reconcile the directory invoices against the citation shares before renewing anything. In the Forward Push dataset, Super Lawyers and Justia alone took 28.87% of all citations while every other directory combined took under 19%. Claim and correct the free profiles regardless of spend, because an unclaimed profile with a wrong address still feeds the engines.
Step 6: Measure a fixed panel per engine, repeatedly
Write down ten questions a client asks in the week before instructing a lawyer, covering procedure, cost, eligibility and firm selection. Run each several times in fresh, signed-out sessions on ChatGPT and Google AI Overviews, and record who is named, who is cited and what the answer got wrong. The “who was cited instead” column is the work list; the protocol is in how to check if ChatGPT recommends your brand.
What reviews do and do not do for AI recommendations
Reviews moved AI recommendations very little in both datasets behind this guide: Avvo, the review-led directory most lawyers know, produced 39 of 30,319 citations in the Forward Push study, or 0.13%. Canlah AI’s own measurements point the same way. A 20-year Singapore restaurant group with 3,452 genuine reviews rated 4.5 stars or higher received zero recommendations across 49 non-branded probes on three AI engines.
Reviews remain a strong indirect lever. They change the sentence an engine writes once the firm is already in the answer, and they decide whether a Google Business Profile survives comparison in local results. Keep collecting and answering them, but stop treating review volume as a route into an AI shortlist.
Law-firm GEO framework at a glance
The law firm GEO framework has six layers, from crawl access to a fixed engine panel, and each layer produces one record, one decision and one conduct check.
Law firm GEO baseline: what to record and the decision each record supports.
| Layer | What to record | Decision it supports | Conduct check |
|---|---|---|---|
| Crawl access | Robots.txt groups, CDN bot rules and platform firewall settings per crawler | Whether engines can read the site at all | None needed |
| Entity record | Name, offices, admissions, jurisdictions, practice areas and lawyer names across every profile | Which profile to correct first | Admissions stated as registered |
| Practice pages | One page per client question, answer and fee in the opening sentence | Which pages need an answer capsule | No superlatives, no outcome promises |
| Directories | Current spend per directory against its measured citation share | Which listings to renew, claim or drop | Paid listings treated as advertising |
| Reviews | Count, rating and response rate per office, plus how each review was obtained | Whether the profile survives local comparison | No solicited testimonials where restricted |
| Engine panel | Ten client questions, several runs per engine, answers archived with timestamps | Whether visibility changed and where | Non-compliant quoted sentences flagged |
Source: Canlah AI law firm GEO method, September 23, 2026. Conduct cells summarise rule families and are not legal advice.
The advertising rules that shape law-firm GEO
Three ABA Model Rules of Professional Conduct, Rules 7.1, 7.2 and 7.3, set the United States template for what a law firm may publish for an AI engine to quote, and each state bar adopts its own version.
- Rule 7.1: It governs every sentence an engine can quote. A lawyer shall not make a false or misleading communication about the lawyer or the lawyer’s services, including a material misrepresentation of fact or law or the omission of a fact necessary to make the statement not materially misleading. A firm page promising outcomes becomes a liability the moment an engine repeats it as fact.
- Rule 7.2: It governs the directory line item. It restricts paying anyone for a recommendation, with exceptions that include the reasonable costs of advertisements and the usual charges of a qualified lawyer referral service. It also restricts implying certification as a specialist. Directory spend is advertising spend and should be documented as such.
- Rule 7.3: It governs outreach, not pages. Solicitation rules limit live person-to-person contact with prospective clients, which matters if a vendor bundles automated outreach with GEO.
- Local rules: They override the template. Testimonial, comparative-claim and disclaimer requirements differ by state, and outside the United States the frameworks differ again. Singapore firms work under Part 5 of the Legal Profession (Professional Conduct) Rules 2015, covered in the Singapore law firm GEO ranking.
Every sentence written for an engine has to survive a regulator reading it, not only a model. Confirm the position with the relevant bar before publishing.
How to interpret five common results
Five results recur in law firm GEO panels, and each one points to a different first fix.
Not mentioned at all: Confirm the question matches a practice the firm runs in that jurisdiction, then check crawl access before writing anything. Compare the sources cited for the firms that were named, because a missing directory profile is a more common cause than missing content.
A directory is cited instead of the firm’s site: This is the normal pattern for firm-selection questions. Correct the profile, then check whether the firm’s own page answers that question in its opening lines.
Named for procedure questions, absent for firm-selection questions: The explainer content works and the reputation layer does not. Move budget from articles to directory profiles, peer rankings and editorial coverage.
Named with the wrong facts: Prioritise correction over exposure. Align the website, every directory profile and the Google Business Profile on one set of facts, then re-run the question that produced the error and archive both answers.
Visible in ChatGPT, absent from Google AI Overviews: Treat each engine as its own source graph. Google AI Overviews leans on Google’s index and Business Profile data, so indexation and profile completeness matter more there than another explainer.
What law-firm GEO cannot do
Law firm GEO cannot do four things, and a proposal that promises any of them should be questioned.
- It cannot make a prohibited claim, a solicited testimonial or an undisclosed promotional post acceptable by moving it into content an engine reads. Question any vendor who proposes one.
- It cannot reliably displace courts, statutes and government portals on procedure and eligibility questions.
- It cannot prove that one page change caused an answer to change without repeated observations and a dated timeline. Report association, not causation.
- It cannot guarantee a recommendation, because no vendor controls how an independent model answers every question.
When to move from a manual panel to a managed programme
A managed programme is justified when one of three conditions holds: AI-assisted research reaches a meaningful share of new matters, the firm runs several offices or practice groups whose facts must stay consistent or an inaccurate answer creates regulatory exposure. Canlah AI archives prompts, responses and timestamps, and measures the gap first with a free AI visibility report within 48 hours; its GEO service is priced per AI platform, from $1,500 a month.
Canlah AI has a real limitation here: none of the clients named on canlah.ai/product (GalaxyCredit, ilLido Group, Sushi Ryujiro, SmallRig and an education franchise) is a law firm, Canlah AI is not a law firm and it does not perform conduct-rule review. A single-partner practice may need nothing more than a documented monthly panel of ten client questions until the number of engines, offices or practice areas makes that unreliable. Verify any provider’s claims, including Canlah AI’s, against raw answers before signing.
Frequently asked questions
What is GEO for law firms?
GEO for law firms is the practice of getting a firm named and cited accurately inside AI-generated answers, rather than only ranked in classic search results. It covers crawl access, practice-area pages written one question at a time, a consistent lawyer entity record, directory profiles and a panel of client questions measured per engine.
Is AEO for law firms the same as generative engine optimization?
Largely yes, and the difference is vocabulary rather than deliverables. Some vendors reserve answer engine optimization for direct-answer placements such as Google AI Overviews and GEO for citations inside chat assistants; others use AEO, GEO, AI SEO and LLM optimization interchangeably. Compare the questions, engines, source lists and before-and-after evidence, not the acronym.
Which legal directories matter most for AI visibility for law firms?
Super Lawyers and Justia led the measured citations in Forward Push’s August 17, 2026 study, at 15.65% and 13.22% of all citations, followed by Best Law Firms, Best Lawyers and BCG Attorney Search. FindLaw, Martindale-Hubbell and Avvo together supplied under half of one percent. Those shares come from United States markets and one window, so re-check them before cancelling anything.
Can an agency guarantee that ChatGPT will recommend my law firm?
No. A provider can guarantee work, sampling and reporting. It cannot control how an independent model answers every client question, so treat a guaranteed recommendation, a guaranteed citation count or category exclusivity as a warning sign rather than a differentiator.
Do Google reviews make AI engines recommend a law firm?
Google reviews alone are a weak direct signal for getting a law firm recommended by AI engines. Avvo, a review-led legal directory, supplied 0.13% of the citations in the Forward Push dataset, and Canlah AI measured a restaurant group with 3,452 reviews rated 4.5 stars or higher receiving zero recommendations across 49 non-branded probes. Reviews still shape how an engine describes a firm it has already found, so maintain them without expecting a shortlist place.
Is Canlah AI a legal marketing agency?
Canlah AI is a Singapore-based SEO + GEO agency that measures and improves brand visibility inside ChatGPT, Gemini, Google AI Overviews and Google AI Mode, reporting AI visibility as re-verifiable ranges with timestamped evidence. It is not a legal-only agency and does not advise on conduct rules, so firms should pair its measurement with their own ethics review.
Method and sources
This guide draws on one third-party citation study, two vendor pages, the ABA Model Rules and two Canlah AI first-party archives, all checked on September 23, 2026. Third-party citation shares were taken as published and were not independently re-run. Vendor dashboards were not tested, client work was not audited and no conduct rule was verified against a current bar opinion.
- Forward Push, Legal Directories Ranked by AI Citation Weight, August 17, 2026. Directory citation shares and practice-area dependence figures.
- Rankings.io AI SEO for law firms. Vendor definitions of GEO, AEO and AIO.
- Nifty Marketing law firm GEO. Vendor use of GEO, AEO, AI SEO and LLM optimization as one category.
- ABA Model Rules of Professional Conduct, Rules 7.1 to 7.3. Template text on advertising, payment for recommendations and solicitation.
- Canlah AI citation study, 60 queries across ChatGPT and Google AI Mode, 526 citations, 92 pages, September 23, 2026. Engine source split and page-format findings for the legal slice.
- Canlah AI cases and Canlah AI GEO services. Restaurant-group probe figures and the measurement protocol.
See where your firm stands in AI answers. Run the free AI visibility audit →
Related articles
FREE WHITEPAPER
Marketing in the Agent Era
All 13 chapters public — no email wall. Includes an original dataset on the agent-readiness of 50 cross-border DTC brands.
Read it free →