How do I check if ChatGPT recommends my law firm? (The DIY method)

By Moshe Dor, FounderLast updated

Ask the engines what your clients ask them. Run 10 realistic prompts — “best car accident lawyer in Phoenix” and variants — across ChatGPT, Claude, Perplexity, and Google AI Overviews in fresh sessions, and record every firm named. If you are absent across 40 answers, you are invisible in the channel carrying 28% of legal research.

This page gives away our manual method in full — no email gate. Most firms that run it find out in an afternoon that competitors own their queries. What you do with that information is up to you.

Which prompts should I run?

These ten, filled in with your practice area, city, and firm name. They cover the full funnel: direct recommendation queries, research queries where firms get cited, and entity checks on your own name.

The 10-prompt DIY panel
#Prompt templateWhat it tests
1best [practice area] lawyer in [city]The money query — direct recommendation intent
2[practice area] lawyer near me in [city]Local variant; triggers different sources
3how do I choose a [practice area] attorney in [city]Advice query where firms get named as examples
4what is my [case type] case worth in [state]Pre-retention research; content citations
5do I need a lawyer for [common matter] in [state]Earliest-stage query in the funnel
6how much does a [practice area] lawyer cost in [city]Fee research; firms with fee content win
7top rated [practice area] law firms [city]List-style query; heavy directory influence
8[specific scenario, e.g. rear-ended by uninsured driver] — what do I do in [state]Scenario query; quotable content gets cited
9is [your firm name] a good law firmDirect entity check — what models know about you
10[competitor firm name] vs other [practice area] firms in [city]Reveals who owns your comparison surface

How do I run them fairly?

Fresh context, every time. Use a new chat session per prompt (logged out or temporary mode where possible) so earlier questions do not contaminate answers. Run all ten on each engine: ChatGPT, Claude, Perplexity, and Google (watch for the AI Overview at the top of results). Copy each full answer into a spreadsheet with columns for engine, prompt, firms named, and whether you appear. Forty answers takes two to three hours.

How do I score the results?

Count citations, not impressions. Your share-of-voice is: answers naming your firm ÷ 40. Do the same for the two or three competitors who keep appearing — they are your real rivals in this channel, whatever the ad auctions say. Under 10% while a competitor holds 25%+ means the channel is actively routing your cases to them. Re-run the identical panel monthly; the trend is the signal, because single snapshots are volatile.

What are the limits of the DIY method?

Three honest ones. Consumer sessions are personalized and rate-limited, so your results carry noise that API-based testing avoids. Ten prompts sample the surface — money queries in one metro can span dozens of variants. And the method tells you whether you are invisible, not why — the diagnosis of entity gaps, schema, crawler access, and content is where the real fix list comes from.

DIY vs the professional audit
DimensionDIY methodFirmCited audit
Prompts10, manually typed25, practice-area and metro tuned
Engines4, fresh sessions by hand4, via direct APIs — no personalization bleed
Citation countingEyeball + spreadsheetDeterministic parsing, every firm counted
Competitor share-of-voiceRough tallyComputed per engine, per query
OutputYour notes10-page prioritized P0-P2 fix list, <48h
CostFree (2-3 hours of your time)$490, credited toward any install

We run the same method — 25x deeper.

The $490 audit runs 25 tuned prompts across all 4 engines via API, computes competitor share-of-voice deterministically, and delivers a prioritized fix list in under 48 hours — credited in full toward any install.