Generative Engine Optimization for Law Firms
More people now ask an AI assistant before they open Google. Those answers name a handful of firms and skip the rest. GEO and AEO are the work of making your firm one of the names that gets retrieved, quoted and cited — and of measuring whether it is actually happening.
What generative engine optimization actually is
A working definition you can hold someone to.
Generative engine optimization (GEO) is the practice of structuring a website so that generative AI systems — Google AI Overviews, ChatGPT, Perplexity, Gemini, Copilot — can retrieve its content, trust it enough to quote it, and attribute it to your firm.
Answer engine optimization (AEO) is the narrower discipline of formatting specific passages so they work as standalone answers: a clear question, a direct answer in the first sentence, and enough context that the passage survives being lifted out of the page.
Neither is a switch you flip. Both are the cumulative result of crawlability, entity consistency, specific facts and clean structure. If a system cannot fetch your page, parse your headings or verify who you are, no amount of content will get you cited.
Practical test: paste your service page into a plain-text reader. If the first 60 words do not answer “what is this and who is it for”, an AI system has the same problem you do.
How AI answers decide which law firm to name
Four signals decide most of it. Three are on your site.
Crawlability
If the crawler is blocked or the content only exists after JavaScript runs, the page does not exist for retrieval purposes. Server-rendered HTML is the baseline.
Passage structure
Answers are assembled from passages, not pages. A heading followed by a self-contained 40–60 word answer is retrievable; a 300-word paragraph that only makes sense in context is not.
Entity confidence
Systems need to know who you are: consistent name, service description, people, contact details, and matching profiles elsewhere. Contradictions lower confidence.
Specific, checkable facts
“We get results” is not citable. “We publish our audit method, our pricing bands and our response-time standard” is.
What we actually change on a law firm site
No new design. No new pages invented to feed a model.
Every item below is a change to existing pages or to structured data. We do not rewrite your site to sound like an AI wrote it, and we do not mass-produce pages hoping one gets quoted.
Answer-first openings
A 40–60 word definition under each pillar H1, before the sales narrative. This is the single highest-leverage change on most law firm sites.
Question-shaped headings
H2s and H3s phrased the way clients ask, each followed by a direct answer rather than a build-up.
Comparison tables
Service vs software, SEO vs ads, answering service vs AI intake. Tables are extracted cleanly and quoted often.
Dated, sourced facts
Every number gets a method and a date, or it comes out. Unsourced claims are the first thing a careful system discards.
Entity consistency
One name, one description, one set of contact details across the site, the schema and your public profiles.
Structured data that matches the page
Organization, Service, FAQPage and Article markup that reflects visible content — not decoration.
What nobody can promise you
This section exists because the alternative is a vendor who lies to you.
We cannot guarantee that any AI system will cite your firm, name it first, or keep citing it next month. Model behaviour changes without notice, retrieval sets shift, and no on-page technique overrides that.
What we can do is remove the reasons a system would skip you: blocked crawlers, unparseable pages, contradictory entity data, unsourced claims, and content that never states what it is.
If an agency guarantees ChatGPT citations, that guarantee is not real. Ask them for their measurement sheet instead.
How we measure AI visibility
A number you can track month over month, not a screenshot.
We freeze a prompt set and an engine list, then record the same things every month. Citation share is reported as a count — “cited in 14 of 100 prompt-engine pairs” — never as a promise.
| What we record | Why it matters |
|---|---|
| Mentioned: yes / no | Whether the firm appears anywhere in the answer |
| Cited: yes / no | Whether the site is linked as a source — a stronger signal than a mention |
| Position in answer | First recommendation or buried in a list |
| Competitors named | Who is winning the prompts you care about |
| Sentiment and accuracy | Whether the answer describes your firm correctly |
Baseline: 20 prompts × 5 engines = 100 prompt-engine pairs, re-run monthly.
How to write a passage an AI system can quote
The difference between a page that can be cited and one that cannot is usually structural, not editorial.
A retrievable passage has four properties. It answers one question. It opens with the answer, not the context. It contains at least one specific, checkable fact. And it makes sense when lifted out of the page entirely.
The failure mode is the build-up paragraph — three sentences of context before the point. A human reader tolerates it. A retrieval system, assembling an answer from fragments, does not.
| Weak passage | Retrievable passage |
|---|---|
| “Choosing the right marketing partner is a big decision for any law firm, and there are many factors to consider before you begin.” | “Law firm SEO is the work of earning organic visibility for practice-area and city searches so that high-intent searches produce consultations.” |
| “We have helped many firms grow over the years with our proven approach.” | “Our audit has twelve checks. The first six cover index control, because everything measured afterwards depends on them.” |
| “Results vary, but many clients see improvement.” | “We report citation share as a count — cited in 14 of 100 prompt-engine pairs — and never as a guaranteed outcome.” |
Same subject, different retrievability. The right column can be quoted; the left cannot.
The part most firms skip
Structure gets you parsed. Consistency gets you trusted.
One description of the firm
The same sentence about what you do, everywhere: site, schema, LinkedIn, directories. Variations read as different entities.
Named people
A founder and authors with real pages, not a byline that is plain text. Systems weight attributed expertise differently from anonymous copy.
Matching contact details
Email, phone and any published address identical across the site and the schema. A contradiction here is the single most common entity failure.
Evidence, not adjectives
“Fourteen years working with law firms” is checkable. “Industry-leading” is not. Every claim you can substantiate is a claim a system can repeat.
A 90-day GEO programme
Sequenced so the cheap structural work lands before the expensive content work.
Days 1–14: Audit
Crawl access, server-rendered content check, entity consistency review, schema validation, baseline prompt run across 5 engines.
Days 15–30: Structure
Answer-first blocks on every pillar page, question-shaped H2s, definition sections, comparison tables where facts already exist.
Days 31–60: Entity
One canonical description of the firm, Person and Organization markup, profile alignment, dated facts on every claim.
Days 61–90: Depth
Comparison content, original observations from your own intake data, and the second measurement run to compare against baseline.
Ongoing
Monthly prompt run, quarterly content refresh, and a written note whenever a number changes.
Questions law firms ask us.
What is the difference between GEO and AEO?
GEO is the broader goal of being retrieved and cited by generative systems. AEO is the narrower craft of formatting a passage so it works as a standalone answer. In practice you do AEO work to achieve GEO outcomes.
Can you guarantee my firm gets cited by ChatGPT?
No, and neither can anyone else. What we can do is make the site crawlable, parseable, consistent and citable, then measure whether citations actually happen.
Do I need new pages for GEO?
Usually not. Most of the gain comes from restructuring pages you already have: a definition under the H1, question-shaped headings, dated facts and matching schema. We propose a new page only when an intent genuinely has no home.
How long before anything changes?
Structural fixes can show up in retrieval within weeks. Citation share typically needs two to three measurement cycles before the trend means anything.
Does this replace SEO?
No. AI answers are largely grounded in the same index. Strong technical SEO and clear entity data are prerequisites, not alternatives.
Related services and reading
Want to know what AI systems currently say about your firm?
We run a 20-prompt baseline across five engines and show you the raw results — including the prompts where you do not appear.