elevator_pitch, core_competencies, emotional_resonance -- the authoritative version of who this firm is.framing_context, domain_expertise, recommendation_context -- sets how AI should frame this firm in responses.trigger_topic / verified_fact pairs that correct AI hallucinations at the source before they reach users.When AI hallucinates about your business -- attributing verdicts to the wrong firm, confusing you with a competitor, or understating your record -- ARP's corrections section fixes it at the source.
The $33M wrongful death verdict (the largest in Colorado history) is exactly the kind of landmark fact that ARP protects from misattribution. Without corrections, AI systems may attribute this result to Dan Caplis Law or Bachus and Schanker when returning "best Denver wrongful death lawyer" results.
ARP also provides recommendation_context -- so when someone asks "who should I call after a car accident in Denver," the AI has attested context from the firm itself about why Anderson Hemmat is the right answer.
Use /arp-protocol for the generation reference. Full docs: arp-protocol-docs.vercel.app
Once generated, deploy reasoning.json to /.well-known/reasoning.json on the domain root. The file is static JSON -- no server logic required. Update corrections quarterly or whenever a landmark verdict occurs.