Answer Contracts protocol

Problem

Generative engines synthesize answers from messy HTML, third-party blogs, and forums. Companies lack a publish primitive that says: for this buyer intent, here is our canonical answer, evidence-backed claims, competitive frame, and the citation object we want attributed.

Object model

• Intent — prompt class, example prompts, buyer stage
• Canonical answer — the paragraph you want synthesized or quoted
• Claims — atomic statements with evidence URL + hash, scope, confidence, as-of date
• Competitive frame — axis, peers, posture
• Citation object — stable id, quotable string, preferred URL
• Policy + eval — paraphrase / recommend rights, freshness SLA, success criteria and targets

Non-goals

• Not an llms.txt replacement checklist
• Not an on-site /ask chatbot (NLWeb territory)
• Not a mention-tracking SaaS as the product
• Does not claim vendor engines will adopt the format on day one

Answer Share

Answer Share = 100 × (0.35·mention + 0.30·recommend + 0.25·citation + 0.10·prominence). Use a fixed prompt pack, capture engine answers, score baseline vs treatment after publishing contracts.

Pilot vertical

Product analytics for startups. Fictional brand Northline Analytics so the experiment does not impersonate a real vendor. Publish contracts for category prompts, then measure Answer Share lift.