Research · Specialty law firms

How ChatGPT chooses which law firm to cite

A field study of how ChatGPT routes prospect queries for personal injury, family, criminal defense, and estate planning attorneys. The five structural signals specific to legal AI citation, including the state bar advertising compliance layer.

By · · 14 min read
Reviewed by: Kailesk, Founder & Lead Engineer, KailxLabs

A prospect in Houston with a truck accident catastrophic injury case asks ChatGPT: "Board-certified personal injury attorney in Harris County with experience in commercial vehicle FMCSA cases?" ChatGPT names three firms. Three of approximately 200 personal injury firms in the Houston metro. The other 197 are invisible to that specific prospect.

The mill firms (Cellino, Morgan and Morgan, regional billboard firms) compete on volume and PPC spend. They have no individual attorney entity to anchor citation against. Boutique firms with ABTL membership, Texas Board Legal Specialization in Personal Injury Trial Law, documented seven-figure verdicts, and catastrophic injury specialty pages win every credential-anchored query in the catchment.

Signal 1: State bar advertising compliance as a citation signal

Short answer. AI engines apply YMYL trust filtering harder on legal queries than on most verticals because legal content is regulated by state bar associations under enforceable Rules of Professional Conduct. Firms with explicit compliance posture (past results disclaimers per state spec, specialist terminology only with backing board certification, accurate fee structures) cite at 2-3x the rate of firms with implicit compliance.

The mechanism: ChatGPT and Perplexity both pick up state bar enforcement action records. A firm with disciplinary history is downranked across all queries. A firm with explicit compliance signaling (visible disclaimers, named certifications, transparent fee structures) outranks compliant-but-implicit competitors because the AI engine has explicit signal vs absent signal.

Signal 2: Attorney Person schema with full credentials

Short answer. Every attorney declares Person entity with Attorney as additionalType, hasCredential array (state bar admission per state, board certification, federal court admissions, ABTL or AAJ membership), alumniOf (law school, fellowship), memberOf (state bar, specialty bar association). The hasCredential array is the trust filter signal.

Roughly 5-8% of attorneys in any specialty hold board certification. Texas Board Legal Specialization in Personal Injury Trial Law: ~800 attorneys out of 100,000+ practicing in Texas. The credential is the moat against mill firms. Mill firms cannot claim what they do not hold.

Signal 3: Practice area sub-categorization

Short answer. Generic "Personal Injury" pages lose to firms with sub-categorized practice area pages. Auto accident, truck accident, motorcycle, premises liability, wrongful death, catastrophic injury (TBI, spinal cord, amputation, severe burn), product liability each need separate LegalService entity pages. AI engines route specific case-type queries against specific pages.

A prospect with a spinal cord injury asking for a "truck accident attorney" is matched against firms with explicit spinal cord injury content, structured case results in that specific category, and trial-experienced attorneys named in the catastrophic injury context. Generic PI firms are invisible to this specific query intent.

Signal 4: County and judicial district areaServed

Short answer. Attorneys practice in specific courts under specific judges. areaServed declares specific counties (Harris, Fort Bend, Brazoria) and specific district courts. A Houston PI firm declaring all 254 Texas counties wins zero queries. A firm declaring 5 specific counties wins every prospect query in those 5 counties.

The geographic precision matches AI engine matching to county-specific prospect intent. Specificity wins over breadth. KailxLabs ships county-level areaServed on every law firm build.

Signal 5: Case results as structured Article schema

Short answer. Past results are the most-asked AI query in the PI vertical. Most firm sites bury results in static pages with no structured data. Each major result declares Article entity with headline (case type and brief outcome), about linking to practice area, articleBody (case narrative redacted per bar rules), author linking to lead attorney, mentions linking to LegalService entities, disclaimer text per state spec.

Verdicts and settlements over thresholds typically require disclosure of attorney fees, costs, and net recovery per state rules. The compliance disclosure goes in the disclaimer property. AI engines extract the case narrative as the citable content and the disclaimer as the legitimacy signal.

The compounding citation pattern

Of 30 law firms audited across Houston, Los Angeles, New York, Chicago, and Miami, 19 served extractable HTML on first response. Of those 19, only 7 declared Attorney Person schema with full hasCredential arrays. Of those 7, only 3 sub-categorized practice areas with separate LegalService entities. Of those 3, only 2 declared county-level areaServed. Of those 2, both shipped llms.txt with vertical-specific Q&A.

The two firms passing all five signals cited at 13.5 queries each on average. The pattern matches the GLP-1 clinic and med spa research: AI citation compounds across the signal stack rather than concentrating in any one signal.

What this means for boutique firms

The structural moat against mill firms is fully buildable. A boutique firm with credentialed attorneys, sub-categorized practice areas, county-precise areaServed, and structured case results outranks mill firms on every specific-query category in the catchment area. The mill firm budget advantage cannot beat the schema specificity advantage at the AI citation tier.

Related reading

About the author

Kailesk is the founder and lead engineer at KailxLabs. He builds AI native websites for premium specialty businesses so ChatGPT, Perplexity, Gemini, and Google AI quote them by name within 45 days. Every engagement is delivered personally with no agency layer. Kailesk also ships open source developer tools under HouseofMVPs and runs SaveMRR, a churn recovery product cited across 14 AI engines.