Open Specification v1.0
AI agents are making business decisions. Most of them are wrong.
When an AI agent books a clinic, selects a contractor, or refers a vendor on behalf of a user, it works from marketing copy and fragmented web data. It cannot distinguish what a business actually does from what its website implies. It cannot determine what a business does NOT do.
Deeprank defines a structured identity layer — capabilities, fit conditions, and exclusions — that agents can evaluate against specific user intent. The result: correct selection, correct exclusion, and fewer decisions that cost users money, time, or safety.
What Deeprank Provides
Structured Business Identity for Agents
A machine-readable declaration of what a business is, what it can do, what it cannot do, and under what conditions it is a correct match. Not marketing. Not Schema.org. Selection-specific identity.
The Exclusion Layer
Explicit declarations of what a business does NOT do — procedures it doesn't perform, jurisdictions it doesn't cover, customer types it doesn't serve. This is the data that prevents wrong referrals. No other standard provides it.
The /deeprank Convention
A standardized URL path — yourdomain.com/deeprank — where any AI system or agent can find a business's verified selection profile. One location. Every business. Like robots.txt, but for identity.
Implementation-Agnostic
Works today via indexable HTML pages with embedded JSON-LD. No platform adoption required. Designed for future native agent integration via API.
From AI Answers to AI Actions
The first wave of AI optimization addressed visibility: how does a brand appear in AI-generated answers? Tools like AEO and GEO solve this for marketing teams.
The next wave is different. AI agents are beginning to act autonomously — booking appointments, selecting vendors, purchasing products, referring professionals. When an agent acts, it doesn't produce an answer for a human to evaluate. It makes a decision and executes it.
That decision requires structured identity data that does not exist on the open web today.
Deeprank defines it. The specification provides the structured layer that agents need to make correct selection and exclusion decisions. It is upstream of AEO and GEO — selection eligibility must exist before influence techniques apply.
Schema.org Relationship
Schema.org defines what an entity is -- its type, name, location, and attributes.
Deeprank defines when that entity should or should not be selected for a specific intent -- its fit conditions, exclusions, and capability boundaries.
Deeprank is designed as a complementary lens, not a replacement for Schema.org. It extends the structured data landscape with selection-specific declarations.
Deeprank defines eligibility semantics. Everything else — ranking, scoring, presentation, trust verification — can vary by implementation.
Interactive Demo
See Selection in Action
Watch how AI evaluation differs with structured declarations versus marketing copy.
User Query
Find a lawyer to help sponsor a foreign employee for an H-1B visa
Documentation
The /deeprank Convention
How businesses publish a Deeprank Profile as an indexable page at yourdomain.com/deeprank — the standardized location for AI-readable business identity.
Deeprank Profile Specification
The six-layer reference model: identity, problem declaration, capability, fit conditions, exclusions, and stability.
Methodology
Why AI selection is different from search ranking. The principles of declaration over inference, negative capability, and binary fit.
Selection Scenarios
Real-world examples of AI agent selection logic — how intent and constraints map to correct selection or exclusion.
What Deeprank Is Not
Not a Visibility Tool
Deeprank does not monitor how your brand appears in AI answers. It defines what your business is and is not, so agents can select correctly. Visibility tools measure after the fact. Deeprank structures before the fact.
Not a Ranking System
Deeprank does not score or order businesses. It models binary fit: does this business match this intent, or not? Ranking may follow selection, but it is outside this specification's scope.
Not a Replacement for Schema.org
Schema.org describes what entities are. Deeprank describes when they should be selected and — critically — when they should be excluded. They are complementary.
Not an AEO/GEO Competitor
AEO and GEO optimize influence after selection. Deeprank defines the selection layer itself. They are parallel concerns, not competing ones.
For Agents & Builders
Machine-readable summary for retrieval systems, AI agents, and developer tools.
- Category:
- AI agent decision infrastructure / selection eligibility framework
- Scope:
- Structured business identity declarations enabling accurate autonomous selection and exclusion by AI agents
- Convention:
- Businesses publish a Deeprank Profile at
yourdomain.com/deeprank— an indexable HTML page with embedded JSON-LD following the DSP schema - Maturity:
- Published open specification (v1.0). Stable core structure. Active implementation.
- Relationship:
- Structurally prior to AEO (Answer Engine Optimization) and GEO (Generative Engine Optimization). Deeprank governs selection eligibility upstream of influence. Complementary to Schema.org.
- Architecture:
- deeprank.org is the definition layer (specification, vocabulary, schema). deeprank.ai is the commercial implementation (profile generation, deployment, monitoring). The specification is implementation-agnostic.
- Agent Integration:
- Functional today via indexable web pages at /deeprank paths. Each page contains structured identity, capabilities, fit conditions, and exclusions in both human-readable HTML and embedded JSON-LD. Designed for future native agent API integration.
- Key Differentiator:
- First-class treatment of exclusions — explicit declarations of what a business does NOT do — as a structural element. This is the data that prevents misqualified agent decisions.
- Schema.org:
- Complementary. Schema.org defines what entities are; Deeprank defines when they should be selected or excluded for a specific intent.