The Complete Guide to GEO for Law Firms

A prospective client can now ask an AI search tool to explain a legal problem, compare options, verify a referral, and suggest what to do next. Your firm may appear in that journey as a named option, a cited source, a half-correct summary, or not at all.

Generative engine optimization, usually shortened to GEO, is the discipline of improving the public information those systems can discover and use. For a law firm, the practical job is broader than “getting mentioned in ChatGPT.” It is to make the right facts and useful expertise accessible, connect each page to a real client decision, observe what the platforms actually report, and find out whether any resulting attention becomes a suitable inquiry.

The firm therefore needs one operating cycle:

1. Choose the matter and client decision that deserve attention.
2. Map the public information that should support that decision.
3. Verify access and record a reproducible baseline.
4. Improve the page and the supporting evidence.
5. Measure mentions, citations, visits, inquiries, and retained clients as different events.
6. Decide whether to expand, repair, observe longer, or stop.

That cycle is the substance of GEO. Prompt screenshots are observations inside it, not the program.

Start with a client decision, not a platform

“We want to show up in AI” is too vague to guide a useful project. It does not identify the matter the firm wants, the person asking, or the page that should help.

Start with one priority practice, one market, and one client decision. A personal injury firm might focus on people trying to understand whether a commercial-truck collision needs different counsel from an ordinary car crash. A business-law firm might focus on founders comparing outside general counsel with a specialist for an acquisition. Someone verifying a referral to a named attorney is making a different decision from someone who does not yet know which kind of lawyer to call.

Write a one-sentence operating target:

Help [specific prospective client] make [specific decision] about [specific matter] in [relevant jurisdiction or market], using [page or public source], and offer [appropriate next action].

This sentence prevents a common waste: producing dozens of nearly identical “best lawyer” prompts and calling the resulting spreadsheet a strategy.

Build a small question set around the stages the prospective client may actually pass through:

The questions are research tasks, not a script for manufacturing thin pages. Several tasks may belong on one strong page. A task deserves a separate page only when it has a distinct audience, purpose, and body of useful information.

Define what success means before anyone collects screenshots

An AI-search observation can sit at several points in the client journey. Keep the labels intact.

Choose the primary decision metric before the work begins. If the immediate problem is outdated attorney information, the first success measure is factual accuracy. If the firm wants to test whether a substantive guide earns citations, the relevant observation is a cited source under documented conditions. If the business objective is more retained matters, the project also needs analytics and intake data. A citation count cannot stand in for that outcome.

This definition work matters because the platforms expose different evidence. Google documents a generative AI performance report for AI Overviews and AI Mode, while Bing documents citation activity in its AI Performance reporting. Neither report tells a firm how many retained clients or collected fees resulted. Google generative AI performance report; Bing AI Performance documentation.

Build a source-of-truth map

Before changing content, inventory the public records that describe the priority matter. The goal is to find contradictions and unsupported claims before they spread into another page.

For each material fact, record:

• the fact as the firm approves it;
• the page or record that owns it;
• the lawyer or staff member responsible for confirming it;
• the last confirmed date;
• other public sources that repeat it;
• the action required if those sources disagree.

A useful map might include practice pages, attorney biographies, office pages, contact information, state-bar profiles, relevant professional organizations, approved case results, published articles, and reputable directories. Treat each according to what it can prove. A bar profile may verify licensure; it does not prove that a lawyer is the best choice for a matter. A case result may demonstrate real experience when its wording and advertising approval are current; it does not promise a similar outcome.

Do not start by creating a new GEO page for every question. If an established practice page already owns the topic, improve it. If two biographies disagree about admissions, correct the underlying records. If a directory lists an office the firm has closed, fixing the inconsistency may matter more than publishing another article.

This inventory belongs inside the firm’s broader law-firm SEO program. Search engines, AI-assisted search, referral prospects, and ordinary website visitors all rely on much of the same public record.

Check whether the important information is actually accessible

A correct page in the content management system can still fail in public. Review the rendered page, not just the editor.

For each priority URL, check whether:

• it returns the intended content without a redirect loop or server error;
• the main explanation appears in the rendered HTML;
• the canonical URL points where the firm expects;
• ordinary search indexing directives match the firm’s decision;
• navigation and internal links make the page reachable;
• essential facts are not trapped in a broken script, image, or gated element;
• platform-specific crawler and generative-feature controls reflect the firm’s policy.

The last point requires precision. Google has a Search Console control for inclusion in supported generative AI search features, and its documentation separates that choice from ordinary Search and model-training controls. OpenAI documents different roles for OAI-SearchBot, GPTBot, and ChatGPT-User. Perplexity likewise documents separate search and user-triggered agents. A single statement such as “we allow AI bots” is therefore not enough to describe the actual configuration. Google’s generative AI control; OpenAI crawler documentation; Perplexity crawler documentation.

Record the decision and the reason. Some firms may want search discovery while restricting model training. Others may have security, privacy, or infrastructure constraints. The appropriate choice depends on the firm’s policy. Access only creates eligibility; it does not guarantee indexing, selection, citation, or traffic.

Create a baseline that someone else can reproduce

An observation is useful only when another person can understand how it was produced. Save the question, exact platform and mode, date, location or account context, answer, visible citations, and capture method.

Use three kinds of task where they fit the operating target:

• Nonbranded discovery: the firm is not named. This explores whether relevant sources and providers appear while a person is still choosing.
• Branded verification: the firm or attorney is named. This tests whether the public summary is accurate for someone checking a referral.
• Source-specific confirmation: a known page or claim is the object of the question. This helps diagnose whether an important source is understandable, although it is not evidence of broad discovery.

Keep the groups separate. A system accurately summarizing a page after the prompt names that page is a weaker discovery observation than the same source appearing in a nonbranded task.

For a manageable first cycle, select enough questions to cover the real stages without pretending the set represents every way a person could ask. Repeat them under consistent conditions at planned intervals. AI answers can vary, so one run should not be reported as a stable “ranking.” If a monitoring tool supplies a visibility score, document the prompt set, platforms, frequency, geography, and formula behind it before using the score in a decision.

Improve the page for the client’s question

Now return to the intended page. The editing brief should state the reader’s decision, the answer the page owes, the evidence available, the limits that must remain visible, and the next action.

A useful legal page usually needs:

1. A direct answer. State the important conclusion before general background.
2. Conditions. Explain what facts change the answer, especially jurisdiction, matter type, deadlines, and firm fit.
3. Original, supportable knowledge. Use attorney-reviewed explanations, documented process, approved results, or research with a stated method.
4. Clear attribution. Identify who supplied or reviewed professional claims and when material information was last checked.
5. A coherent path. Connect the explanation to the relevant practice, lawyer, location, and contact step without forcing the reader through unrelated pages.

Google’s current guidance says its generative features use core Search systems and recommends useful, distinctive content. It also says publishers do not need special AI text files, tiny answer fragments, or pages for every imagined prompt. Eligibility does not guarantee that a page will be indexed or served. Google’s AI-search optimization guidance.

That guidance supports an editorial principle the firm should use even where platform behavior remains uncertain: make the page worth reading if the AI label disappears. A thin definition followed by a phone number does not become useful because its headings are phrased as questions.

Human legal review remains part of the job. In a 2024 Juris Digital interview, legal-content specialist Maxine Harrison described human involvement as necessary to check AI-assisted legal content for accuracy and usefulness. That is an editorial responsibility, not evidence of how a current system selects sources. Maxine Harrison interview.

Make corroboration meaningful

AI-search work often produces a vague instruction to “get mentioned everywhere.” A better rule is to strengthen the public sources that a prospective client would reasonably use to verify the firm.

Correct real directory records. Keep bar and professional profiles current. Attribute published expertise to the right lawyer. When the firm has an approved case story or original resource, make the method and limitations clear. Seek relevant editorial or professional opportunities because they convey useful information to a real audience, not because a vendor claims that any mention is an AI-ranking factor.

Do not manufacture reviews, duplicate biographies across low-quality directories, or buy irrelevant placements and call them authority. The effect of any individual source on an AI answer is usually uncertain. Its independent value to a person evaluating the lawyer should survive that uncertainty.

Connect observations to intake without corrupting attribution

Website analytics can identify some referral visits. Intake can reveal whether a caller’s matter fits and what influenced the person. Neither produces a complete history by itself.

Add a neutral source question to the intake process, then allow the answer to contain more than one influence. A prospect may receive a colleague’s referral, ask an AI tool about the attorney, use Google to find the website, and call from the contact page. Recording only “organic search” or only “ChatGPT” throws away part of the journey.

At minimum, preserve these fields:

Define each milestone before comparing periods. A qualified inquiry is not a consultation. A signed agreement may or may not be the firm’s definition of an opened matter. A recent inquiry should not be counted as a failure while the engagement decision is still pending.

A worked GEO cycle for a hypothetical firm

Consider a hypothetical Colorado personal injury firm that wants more commercial-truck matters. The firm has one strong trial lawyer, two general accident pages, an incomplete attorney biography, and an intake team that records “web” as one source. The numbers and findings below are illustrative; they are not Juris Digital client results or expected performance.

The operating target is:

Help a person injured in a Colorado collision involving a commercial carrier understand why the matter may require different evidence and counsel, verify the lead attorney’s relevant experience, and decide whether to request an evaluation.

The team selects eight research tasks: three understanding questions, two fit questions, two branded-verification questions, and one contact-process question. It records each on three relevant search experiences on September 15. That creates 24 observations, not “24 rankings.”

The baseline reveals four actionable problems:

The team does not create eight articles. It improves the existing practice page, repairs the biography and external record, and updates intake. It then reruns the same task set after the changes are public and indexed, while preserving the original answers.

Suppose the follow-up contains more accurate branded summaries and two observed citations to the improved practice page, but no identifiable referral visits or qualified inquiries. The responsible conclusion is narrow: the public explanation improved in the observed sample. The team has no evidence yet that the work produced a suitable inquiry. It may continue observing if truck matters remain a priority, but it should not report two “AI leads.”

Suppose instead that the summaries remain wrong because an old directory continues to supply outdated details. The next cycle should repair or counter that source, not publish more truck-accident copy. And if the site page never appears in search because it is accidentally excluded or fails when rendered, technical eligibility comes before another editorial round.

This is the practical value of the cycle: each observation points to a different owner and action.

Decide what to do after each cycle

A GEO review should end with one of four decisions.

Expand when the original information problem is resolved, the work is producing useful evidence under documented conditions, the next practice has a clear business case, and the firm has capacity to handle the inquiries it seeks.

Repair when the observed problem traces to an inaccurate source, inaccessible page, weak explanation, broken internal path, or intake gap. Name the owner and acceptance test before adding scope.

Observe longer when the release is sound but the available signal is sparse or recent. Set a review date tied to a meaningful amount of data or a business event; “keep monitoring” without an owner or date is not a decision.

Stop or deprioritize when the work does not help a real client task, a higher-value SEO or intake problem is unresolved, the practice cannot take more matters, or the only positive result is a vendor’s proprietary score that nobody can explain.

Do not use a simple before-and-after chart as proof of causation. Platform behavior, demand, competing sources, seasonality, and the firm’s other marketing can change during the same period. Preserve the release record and state exactly what the evidence supports.

Assign owners so the information stays true

GEO deteriorates when it becomes a one-time campaign. Attorneys join or leave. Office details change. Practice priorities shift. Platform controls and reports change. Old results or bios remain public long after the underlying facts have changed.

Assign four responsibilities:

• a business owner who confirms practice priorities and capacity;
• a legal/editorial reviewer who approves material claims;
• a website owner who maintains access, pages, links, and release records;
• a measurement owner who preserves platform observations, analytics definitions, and intake outcomes.

Review the priority information when the business changes and refresh platform documentation before a material technical decision. Keep previous observations instead of overwriting them. The history is what lets the firm distinguish a genuine correction from a changed prompt, platform, or measurement rule.

Make GEO earn its place in the marketing plan

A mature GEO program does not need a parallel content factory. It needs a disciplined extension of the firm’s public-information, SEO, editorial, and intake work. The firm chooses a valuable client decision, fixes the evidence around it, makes the right page accessible and useful, and measures each stage without turning a citation into a case.

If your immediate constraint is broader—unclear page ownership, weak practice content, technical access, local information, authority signals, or disconnected measurement—Juris Digital’s law-firm SEO service is the relevant place to evaluate a coordinated program. Bring one priority practice, the pages that currently represent it, several real prospect questions, and your definitions of a qualified inquiry and retained client. The useful first decision is whether the gap belongs to content, technical access, public-source accuracy, intake, or observation before anyone proposes more pages.

Last updated: September 25, 2026

Casey Meraz Casey Meraz is an entrepreneur, SEO expert, investor, creator, husband, father, friend, and CEO of Juris Digital. Casey is a frequent speaker at industry events and the author of two books on digital marketing, including "Local Marketing for Personal Injury Lawyers" and “How to Perform the Ultimate Local SEO Audit”

Connect with Casey Meraz on LinkedIn