Five-Layer Brand Visibility System
Truth → Proof → Recommendation → Action. How a brand gets recommended by AI agents, productized in 8 steps.
Ruled 7/12The web grew an invisible layer. Machines now evaluate a brand’s entire digital evidence network before recommending it. The website stays home base. The rest of the web is the trust network. WebMCP becomes the hands.
The law underneath everything below: machine readability helps a business QUALIFY. Web consensus helps it WIN. Agent actions help it CONVERT. Internal name: Truth → Proof → Recommendation → Action.
How an agent actually decides
Ask an AI “who’s the best salon near Cedar Park for blonde highlights” and it runs five moves: understand the business, collect evidence across the web, check agreement between sources, make a recommendation, take action. Old SEO was keyword → webpage → ranking → click. The new environment is question → multi-source retrieval → fact comparison → confidence → recommendation → action. Optimize for four outcomes: understood, corroborated, recommended, actionable.
The five layers
1. Canonical Truth
One structured record: entities, facts, relationships, offers, proof, actions. Artifact is canon.json. Grows from the existing site control layer, never hand-maintained separately. One canon, one build, zero drift.
2. Machine Legibility
Makes the canon retrievable: fast SSR HTML, semantic structure, a page per entity, JSON-LD, sitemap, correct robots, real forms with name attributes, llms.txt. Answers “can the machine understand this business?”
3. Distributed Proof
What credible places OTHER than the company site say: GBP, Reddit, YouTube, LinkedIn, review platforms, local press. The layer traditional website thinking misses. Consensus wins it, not 200 more articles on the owned site.
4. Citation-Worthy Answers
Publish the clearest available answer to the questions customers actually ask. Direct answer up top, specific facts, original examples, proof, next action. Sits on top of the existing Content Engine / Answer Hub doctrine.
5. Agent Actions
Where WebMCP fits, once the agent trusts the business. Read-only tools first (find a provider, check availability), bounded writes later (submit inquiry, book). First tool: Bex findBeautyProfessional.
The productized 8 steps (repeatable for any brand)
Build the Brand Canon
Capture entities, facts, relationships, offers, proof, and actions once into canon.json. The keystone every other step reads from.
Create the Entity Website
A dedicated, indexable page per recommendation-worthy entity. Salon: /locations/cedar-park/, /professionals/alaura/, /services/blonding/. Contractor: /services/kitchen-remodeling/, /projects/cedar-park-kitchen-remodel/.
Open the Correct Crawler Doors
Distinguish search crawlers, agent crawlers, and training crawlers. Cloudflare’s granular Search/Agent/Training controls (shipped July 2026) apply revised defaults to new domains September 15, 2026. This is an explicit launch gate: allow search, allow agent/user-requested retrieval, decide training separately, verify live robots.txt post-launch.
Build the Answer Hub
20-50 high-value questions: recommendation, comparison, cost, “best for,” problem diagnosis, local, firsthand-experience. One canonical page per topic cluster.
Run the Distributed Proof Engine
Platform mix by type. Local consumer: GBP, reviews, local Reddit, YouTube, local press. B2B/expert: LinkedIn, YouTube, podcasts, communities, case studies. Product/software: YouTube, Reddit, G2/Capterra, comparisons. Own the canonical explanation, EARN the independent confirmation. Authentic participation only, no fake posts, no manufactured threads, no mass AI spam.
Add the Action Layer
Read-only tools first, bounded writes later. Sequence rule, never skip ahead to writes.
Measure Recommendation Visibility
Build a Target Question Registry per brand. Run it regularly across ChatGPT, Google AI Mode/AI Overviews, Gemini, Perplexity, Claude. Track: mentioned, recommended, cited, competitors named, facts accurate, action available, change since last run.
Improve Based on Missing Evidence
Every failed target question becomes a work item. Closed loop: question → test → evidence gap → content or distribution action → retest.
Priority order (highest return, least waste)
- Canon + entity consistency
- Website machine clarity
- Target Question Registry
- Answer Hub
- Distributed proof
- Agent actions
- Continuous measurement
The correction to typical AEO advice
Most “AI SEO” advice stops at schema, llms.txt, and FAQs. Those are legibility tools. They qualify a brand. They can’t manufacture authority or public confidence. A brand becomes THE recommendation when the internet presents a coherent case: the website defines it, structured data explains it, real people discuss it, credible sources confirm it, reviews support it, content answers the question, and the agent can complete the task. Sell the whole system, never just the schema layer.
Where this stands today
| Layer | Existing DH system | Status |
|---|---|---|
| 1 — Canonical Truth | canon.json (gated on the site control layer) |
Designed, build pending |
| 2 — Machine Legibility | AR gates, robots W1 gate, Bex template | In the deploy pipeline plan |
| 3 — Distributed Proof | Nothing yet | The new work this ruling creates |
| 4 — Citation-Worthy Answers | Content Engine Master Doc (Answer Hub, citation tracking) | Doctrine live, blog schema gaps named |
| 5 — Agent Actions | tools.ts three doors, WebMCP origin trial, findBeautyProfessional |
Roadmapped |
See also: Agentic Readiness for the technical floor this system builds on.