How the platform actually works.

Discovery

The Auditor agent assembles a 200-prompt library per practice area and jurisdiction. Prompts come from three sources: aggregated AI-search query data we have licensed, your historical buyer-intake notes (you upload if you wish), and competitive scan against your 14-firm peer set.

• Prompts are clustered by intent (high / mid / low) and ranked by estimated query volume.
• Coverage is computed per-engine using a structured probe (not a free-text scrape).
• Library refreshes monthly; you can pin or veto individual prompts.

Sweep

Every 30 minutes (Counsel + Enterprise), daily (Practice), or weekly (Audit Pro) we issue probes to ChatGPT, Claude, Perplexity, Gemini, Copilot, Grok, Meta AI, and Google AI Overviews. Each probe is a structured request, not a screen-scrape — we use first-party APIs where available and respect rate limits.

• Verbatim answer captured per (engine, prompt, timestamp).
• Citations and source links are extracted; sentiment is classified by a small classifier we own.
• Drift = (current_answer - last_answer) similarity drop; alert threshold tunable per workspace.

Scoring

VIS·IDX is a composite of five factors, each weighted by the engine and practice area. The factors and weights are visible on /dashboard/visibility — no black box.

• Citation weight (38%) — proportion of answers that cite your firm.
• Sentiment (22%) — model-attributed sentiment about your firm.
• Share of voice (18%) — your firm vs. the top-3 peers in your bench.
• Query coverage (14%) — proportion of prompts where you appear.
• Recency (8%) — how fresh your appearances are across engines.

Optimize

The Bench Mapper drafts structured content the models actually read — JSON-LD schema for representative matters, partner biographies, FAQ blocks, and citation submissions for the 14 sources that move the needle.

• Schema templates live in your workspace; you edit, the linter validates.
• Citation submissions are tracked end-to-end (drafted → pitched → published → indexed).
• Drafts ship to your CMS (WordPress, Sanity, Contentful, Webflow) on approval.

Publish + trail

Once cleared, the Brief Writer assembles a quarterly Visibility Brief and pushes the cleared content to your CMS. Every step is logged immutably.

• Publish events are timestamped, hashed (SHA-256), and signed (Ed25519).
• Audit log is admissible as a business record under FRE 803(6).
• Quarterly Cognoverge Index uses anonymized aggregate metrics; opt-in for ranked publication.