Research.
Primary research on how AI search verifies and recommends professionals - Clarion's audits, methodology notes, and the entity-graph architecture they document.
State of AI Search, Issue 110 minute read
The first quarterly briefing. A randomized experiment and a launch in France put numbers on lost clicks, half of AI citations turn out to vanish overnight, several widely repeated statistics change verdict, and Clarion publishes its own readings.
Read the issue →Who AI Names When Clients Look for a Law Firm14 minute read
200 US firms of 29 to 170 lawyers, 360 client questions, 5,400 answers from ChatGPT, Perplexity and Google, pre-registered before collection. Directories wrote most of the shortlist, a matching firm was named in about one answer in fifteen, and when the client described the problem the answers cited law firm websites. One finding runs against the work Clarion sells, and it is reported in full.
Read the report →Generative Engine Optimization Statistics, Verified14 minute read
Every number carries a verdict: the verified figures (Pew's measurements, a randomized click experiment, the ~12% citation/ranking overlap, the schema null result), the refuted claims, the hold list - and a dated estimate of the winnable share: about a third in September 2025.
Read the ledger →The AI-Search Mythbusting Register12 minute read
Widely repeated AEO/GEO statistics, tested against their original sources, each with a verdict - plus the real, more modest numbers underneath them, and one finding that looks like a myth but is actually a lesson in correlation versus causation.
Read the register →How AI Verifies You're Real20 minute read
The AEO architecture walkthrough: the seven-layer entity graph Clarion's 130-firm audit specified, taught from first principles with paired examples across two professions, plus a live Schema Inspector that scores your own site - and tells you what's missing, never how to fix it.
Read the walkthrough →The AI Visibility Gap25 minute read
A 130-firm audit of how independent RIAs appear in AI search. Most are visible - but through directories they don't control. Schema coverage averages 1.93 of 5; zero firms link their regulatory record.
Read the full report →