As of July 2026. AI answers are probabilistic — they shift with session, location, device, and model version — so nothing here is a guaranteed placement or ranking. The audit measures what actually happens; it does not promise an outcome.
To see how your law firm shows up in ChatGPT, ask each engine the questions your clients actually ask — “best family lawyer in [city],” “how much does a DUI cost in [state]” — then log whether your firm is Named, Cited, or Invisible, and who wins the slot instead. If your firm scores Invisible on many prompts, the next step is a measured program, not a guess. JustLegal Marketing runs this audit with tracking data, not anecdotes.
If a prospective client opens ChatGPT tonight and types “best divorce attorney near me,” does your firm come back in the answer? If you haven’t checked, it’s worth checking — the result surprises a lot of firm owners. A first-page Google ranking no longer means an AI will recommend you, because generative engines do not read the ten blue links top to bottom. They synthesize an answer from a web index, a maps and entity corpus, third-party directories, and user reviews all at once. The result is a new, measurable question every firm should be asking: not “where do I rank,” but “am I in the answer at all?”
This guide gives you the exact 15-minute audit we use to answer that question — the prompt set, the scoring model, and the levers that move the score. It assumes you already know your way around SEO fundamentals; where an AI-specific term shows up, we define it on first use.
The Prompt Set: 10 Questions That Reveal Your Firm’s AI Footprint
Your firm’s AI footprint is only measurable if you test the queries clients actually use — not your firm name. Run ten prompt archetypes (superlative, cost, situational, reputation, comparison, proximity, process, niche, outcome, and aggregated-recommendation queries) across ChatGPT, Gemini, Google AI Overviews, and Perplexity, and log every result.
Typing your own firm name into ChatGPT tells you almost nothing — of course it can describe a firm you named for it. The audit only works if you simulate how a stranger with a legal problem searches. These ten archetypes cover the discovery paths that matter for a local law firm; swap in your practice area, city, and state:
- Category / city superlative — “Who is the best [practice area] lawyer in [city]?”
- Cost / transactional — “How much does a [matter, e.g., DUI] cost in [state]?”
- Situational / advisory — “Do I need a lawyer for [situation]?”
- Direct reputation check — “[Your firm name] reviews and reputation.”
- Entity comparison — “[Firm A] vs. [Firm B] for [practice area].”
- Proximity / local — “Best [practice area] lawyers near me.”
- Process / guidance — “What are the steps to file for [matter] in [county]?”
- Specialty niche — “Lawyers who handle [specific sub-niche] in [city].”
- Outcome question — “What happens in a [matter] case in [state]?” (Keep your own answering content informational — never let a page imply your firm can predict or guarantee a case result.)
- Aggregated recommendation — “What law firm do people on Reddit recommend for [practice area] in [city]?”
How to run it without polluting the data
Small procedural mistakes quietly ruin the results. Follow the same four controls every time:
- Neutralize session bias. Run every prompt in a Guest window or a clean browser profile, so your own history and prior chats don’t feed the model your firm’s name.
- Run all four engines. ChatGPT (with search on), Gemini, Google Search (to trigger an AI Overview or AI Mode), and Perplexity each source differently — a firm can be Named in one and Invisible in another.
- Wait for the full answer, then log the sources. Record the exact URLs the engine used to justify its response, not just the prose.
- Repeat, don’t one-shot. A single run is an anecdote. Google itself documents that AI Overviews appear selectively — its systems decide when an AI summary would genuinely help, so the same query can show an Overview on one device or day and not another (Google Search Central, “AI Features and Your Website,” PRIMARY). Repeating the set weekly averages out that volatility into a real baseline.
Fifteen minutes gets you a first honest snapshot across all four engines. The weekly repetition over about 30 days is what converts that snapshot into signal.
Reading the Results: Named vs. Cited vs. Invisible (and Why Each Happens)
Every audited prompt resolves to exactly one of three outcomes for your firm: NAMED (the AI’s text recommends you by name), CITED (your URL appears as a linked source even if the prose doesn’t name you), or INVISIBLE (a competitor or a directory won the slot instead). Scoring each result this way turns a fuzzy impression into a trackable number.
In classic search, visibility was a rank position. In generative search it collapses to three states:
- NAMED — the generated text explicitly recommends your firm (e.g., “[Firm] is well regarded for family law in [city]”). This is the strongest outcome.
- CITED — your website is used as a linked source or footnote supporting the answer, even when your name isn’t in the paragraph. You’re feeding the answer; you’re just not the headline.
- INVISIBLE — neither named nor cited. The engine chose a competitor, a “best-of” list, or a directory over you.
This Named / Cited / Invisible model is a measurement convention, not a vendor metric — which is exactly why it’s dependable to score against week over week.
Understanding why a firm lands in each bucket means looking at how each engine sources answers. When a query has local intent, Google’s Gemini can invoke Grounding with Google Maps — an official API tool (an API, or Application Programming Interface, is simply the rulebook that lets one program pull data from another) that connects the model to more than 250 million real-world places and pulls addresses, hours, ratings, and user reviews to build a grounded answer (Google, blog.google / ai.google.dev, PRIMARY; GA October 17, 2025). Two consequences follow directly. First, grounding uses latitude and longitude, so the same query downtown versus in the suburbs can surface different firms — AI answers are localized and non-static. Second, if your Google Business Profile is thin or your reviews are sparse, the model has nothing to ground on, and you fall to Invisible.
ChatGPT and Perplexity behave differently again. Industry vendor studies indicate ChatGPT leans heavily on third-party directories, aggregators, review platforms, and high-authority “best-of” lists to confirm a business exists and is reputable before recommending it — that mechanism is supported only by vendor-self-published research, so treat it as directionally sound rather than documented fact. Perplexity is better described as citing real-time crawlable web pages, directories, and community sources, and it shows those sources as visible footnotes — which makes Cited versus Invisible easy to read straight off the screen. (Where vendors put hard percentages on ChatGPT’s third-party sourcing, those figures are vendor-reported marketing, not platform documentation — directional, not load-bearing.)
The single most useful thing the audit reveals is the gap between your organic rank and your AI presence. You can sit at position one on Google and still be Invisible in the Overview above it — because, as the PRIMARY mechanics above show, the AI is retrieving from map corpora, directories, and citable answer content, not walking the organic top ten. Vendor trackers have tried to size that disconnect and disagree with each other on the exact figure, so don’t anchor on a number; the direction is the point, and the PRIMARY retrieval model supports the takeaway: AI citation is not the same as organic ranking.
What Actually Moves the Score: Entities, Citable Pages, and Reviews
Three levers move a firm from Invisible toward Cited and Named: a clean, consistent entity (accurate NAP, a complete Google Business Profile, and schema so machines parse you confidently); citable “answer-engine” pages built from statistics, quotations, and cited sources; and a steady stream of recent, detailed reviews. The strongest evidence backs citable content.
Lever 1: A consistent entity
An AI has to be able to identify your firm as one discrete, real-world organization before it will recommend you. That means accurate Name, Address, and Phone (NAP) data everywhere it appears, a complete and active Google Business Profile, and structured schema markup — JSON-LD code that acts like a machine-readable name tag, telling an engine “this is a LegalService,” “this is an Attorney,” “this is a Review.” Google’s own Business Profile documentation ties profile completeness and consistency to local relevance and prominence (PRIMARY). Schema is widely recommended and mechanistically sound for helping engines parse your entity — though no platform documents it as a direct AI-answer ranking input, so treat it as “helps machines confidently read you,” not a switch that inserts you into an answer.
Lever 2: Citable answer-engine content
This is the lever with the best independent evidence behind it. In the peer-reviewed 2024 study GEO: Generative Engine Optimization (Aggarwal et al., Princeton and IIT-Delhi, presented at ACM KDD 2024 — INDEPENDENT academic research, not a vendor report), researchers tested content strategies across roughly 10,000 queries and found that classic keyword-stuffing SEO largely fails in generative engines, while content-level tactics measurably win. The lifts were per-tactic, not a single headline number: adding statistics increased visibility by roughly 41%, adding direct quotations by about 28%, and citing authoritative sources produced even larger gains — up to roughly 115% for lower-ranked content. These are study results under specific conditions, not guarantees — but they point clearly at what a citable page should contain.
For a law firm, that translates into a repeatable page anatomy:
- A question-form H2 phrased as the exact query a client would type (“How much does a DUI lawyer cost in [city]?”).
- A direct 40–60-word answer immediately under it — the extractable snippet an engine can lift.
- Hard, verifiable specifics in place of vague claims (a real statute section, a real deadline, a concrete range) rather than “we win a lot of cases.”
- An attributed expert quote from a named, credentialed attorney, so the engine has a human authority to cite.
Lever 3: Reviews — volume, recency, detail, and responses
Reviews feed AI visibility through two documented paths. Google’s Business Profile Help states that “Google review count and review score factor into local search ranking. More reviews and positive ratings can improve your business’s local ranking” (PRIMARY) — that’s reviews feeding local prominence. And Grounding with Google Maps explicitly retrieves user-review text to answer subjective questions like “best” or “most aggressive” (PRIMARY) — that’s reviews feeding a Gemini answer. The honest limit: neither Google nor OpenAI publishes reviews as a weighted input to AI Overviews or ChatGPT specifically, so the right verb is that reviews correlate with AI inclusion, not that they rank you inside it. A practical best-practice target is a steady handful of detailed reviews a month, each specific enough (50-plus words, naming the actual service) to give the model real semantic context — never bought, incentivized, or gated (see below).
Every one of these levers raises the likelihood of being retrieved and cited. None controls it. We don’t promise guaranteed rankings or a spot inside any AI answer — these are evidence-based practices that increase the mathematical likelihood of your firm being named or cited.
Do It Right: The FTC and Google Rules That Govern the Review Lever
Because reviews move the score, firms are tempted to manufacture them — and that is where a visibility play becomes a legal problem. The FTC’s Consumer Reviews rule (16 CFR Part 465, effective October 21, 2024) makes fake and sentiment-conditioned reviews illegal; Google’s separate policy bans review gating and can delete your reviews or suspend your profile. The only durable review engine is a compliant one.
The FTC’s trade regulation rule on consumer reviews took effect October 21, 2024 and carries civil penalties of up to $51,744 per violation (an inflation-adjusted figure that can change — Federal Register / ftc.gov, PRIMARY). It prohibits fake or false reviews — including reviews from reviewers who do not exist or never used the service, and fake reviews generated by AI — buying reviews conditioned on a positive rating, undisclosed insider reviews from staff or their relatives, and suppressing negative reviews through threats. Separately, Google’s Maps policy bans review gating — pre-screening customers by sentiment so happy clients get routed to Google and unhappy ones to a private form — with penalties up to mass deletion of your reviews or suspension of the Business Profile, which can erase your local and AI visibility overnight.
The compliant engine is straightforward: ask every client after the matter closes, use neutral language, and encourage specificity rather than a star count. “Please mention the specific service we handled for you in [city]” is allowed and actually helps — it adds the semantic depth AI grounding rewards. “Leave us five stars for a discount” is not. Depth, not sentiment-shopping, is what feeds the answer engines and keeps you inside the rules.
From Zero to Cited: What a 90-Day GEO Sprint Looks Like
A 90-day Generative Engine Optimization sprint operationalizes the audit into a five-phase loop: baseline the ten prompts, clean up the entity, deploy citable answer content, accelerate compliant reviews, then re-measure the delta. It’s a repeatable measurement program, not a one-time fix — and the outcome is observed change, never a guaranteed placement.
This is the program JustLegal Marketing runs for firms that want the whole loop handled with tracking data. The five phases run in this order over the quarter (the day ranges below are an example cadence, not a fixed contract):
- Audit and baseline (roughly the first two weeks). Run the ten-prompt set across ChatGPT, Gemini, AI Overviews, and Perplexity in a clean Guest profile, logging Named / Cited / Invisible per engine into one tracking sheet.
- Entity cleanup (next). Standardize NAP across directories, complete and optimize the Google Business Profile so Maps grounding can parse the firm, and deploy
LegalService,Attorney,FAQPage, andReviewschema. - Citable content deployment (the middle stretch). Rebuild the highest-value pages using the citable-page anatomy above — question-form H2s, direct short answers, hard specifics, attributed attorney quotes.
- Review acceleration. Stand up a compliant, ask-everyone review process to build a steady stream of recent, detailed reviews on Google and reputable legal directories.
- Re-measure (the final phase). Re-run the exact prompt set and read the delta against the baseline. That observed change — not a promise — dictates the next iteration of the loop.
AI visibility is a moving target; models update, retrieval shifts, and a firm that adapts to the post-2025 grounding mechanics is simply harder to leave out of the answer. What the sprint delivers is a measured before-and-after, run in the open, with no hollow guarantees attached.
Frequently Asked Questions
How does my law firm show up in ChatGPT right now? Open ChatGPT in a Guest browser window, turn on its search function, and ask the ten prompt archetypes above with your practice area, city, and state filled in. Log whether your firm is Named in the text, Cited as a linked source, or Invisible. Run it across Gemini, Google AI Overviews, and Perplexity too — a firm is frequently visible in one engine and absent in another.
Why does a #1 Google ranking not guarantee AI visibility? Because an AI answer is synthesized, not ranked. Generative engines retrieve from a web index, a maps/entity corpus, directories, and reviews, then compose one answer — they do not read the organic top ten in order. Vendor trackers disagree on the exact overlap figure, but the PRIMARY retrieval mechanics make the direction clear: AI citation is a different game than organic ranking.
Can I pay to be named inside a ChatGPT or Gemini answer? No. You cannot buy placement inside the synthesized text of an AI answer. Google may run ads around an AI Overview — that’s ad inventory, not the generated answer — and buying reviews to influence AI is separately illegal under the FTC rule. Visibility inside the answer is earned through entity clarity, citable content, and legitimate reviews.
How long until the audit shows change? Treat it as a 90-day loop: baseline, fix the entity, deploy citable content, accelerate compliant reviews, then re-measure. Because outcomes are probabilistic, you’re watching for an observed delta across repeated runs — not a fixed date when you “arrive.”
Get Your Free AI Citation Audit
If your firm scored Invisible on many of these prompts, the fix is a measured program, not a guess. JustLegal Marketing is the only marketing company for law firms owned and operated by practicing attorneys, and we run this AI visibility audit with real tracking data across ChatGPT, Gemini, AI Overviews, and Perplexity — then build the entity, content, and review foundation that moves the score. Our founder, Stephan Futeral, is both an accomplished attorney and a leading authority on legal marketing, and our work rests on the same SEO and local search fundamentals AI engines still reward.
Get your free AI Citation Audit — call 843-619-0229 to see exactly how your firm shows up in AI search, with tracking data instead of opinion.
