Glossary of AI Infrastructure Terms

SOVP (Sovereign Validation Protocol)

The Sovereign Validation Protocol is a deterministic, machine-readable infrastructure validation protocol developed by Litzki Systems LLC. It scans deterministic parameters across structured validation clusters and produces a three-state verdict of CERTIFIED, NOT_CONFORMANT, or INCOMPLETE. On CERTIFIED it issues an Ed25519-signed identity document anchored in DNS; the verdict itself is reported in the scan result and is not part of that signed document.

Context

SOVP is the technical foundation for assessing agentic commerce readiness. It measures the structural integrity of the digital infrastructure that autonomous agents use to classify and retrieve an entity. Content, traffic, and brand perception are outside its scope.

Litzki Systems has filed SOVP as an IETF Draft, and it is patent pending. Litzki Systems designates the scan engine Ψ_core Engine.

Full technical specification →

Layer 0 (Infrastructure Trust Layer)

Layer 0 is the foundational tier of a three-layer agentic architecture stack. The stack runs as follows:

• Layer 2 (Agent Identity): Federated identity and authorization (OpenID, AuthZEN)
• Layer 1 (Agent Accessibility): API discoverability and structured access protocols
• Layer 0 (Infrastructure Trust): DNS-anchored cryptographic evidence of infrastructure integrity (SOVP)

Layer 0 is where autonomous procurement agents issue their topology verdict, before any system logs an API request, before any dashboard registers traffic.

Agentic Infrastructure specification →

Signal Sovereignty

Signal Sovereignty is the ability to own, control, and guarantee the accuracy of the machine-readable infrastructure signals that agentic systems use for entity classification, procurement decisions, and knowledge graph construction.

Signal Sovereignty replaces probabilistic inference with deterministic ground truth. An entity that relies on crawlers to define its digital presence is at the mercy of what models happen to find and interpret. An entity with Signal Sovereignty provides cryptographically verifiable, DNS-anchored signals that establish identity directly.

Signal Sovereignty is the structural outcome of full SOVP compliance.

Agentic Commerce

Agentic Commerce describes autonomous B2B procurement and vendor evaluation executed by AI agents without human intermediaries. In agentic commerce, an AI agent receives a procurement task, independently discovers candidate vendors, evaluates their infrastructure signals, and either eliminates or advances them, all at machine speed, before any human review.

The decisive evaluation happens at Layer 0: the agent silently filters out infrastructure that falls short of its topology requirements. The exclusion is immediate, structural, and transparent only to the agent that executes it.

Agentic Infrastructure →

Agentic SEO

Agentic SEO is an infrastructure-first approach to visibility in the agentic economy. It is a structural discipline: the configuration of a domain's machine-readable signals to ensure correct classification, retrieval, and trust assignment by autonomous agents. It operates on a separate layer from content strategy, keyword methodology, and traditional search engine optimization.

Where traditional SEO optimizes for human search intent, Agentic SEO optimizes for machine topology verification: structured data integrity, entity canonicalization, DNS-anchored identity signals, and deterministic protocol compliance.

What it produces

The output of Agentic SEO is measurable: a SOVP scan returns CERTIFIED, NOT_CONFORMANT, or INCOMPLETE according to the current Full Validator taxonomy.

Agentic SEO specification →

Agentic Architecture

Agentic Architecture is the structural design of digital infrastructure to support autonomous agent interaction across all three layers of the agentic stack (Layer 0–2). An organization with a compliant Agentic Architecture has:

• DNS-anchored cryptographic identity (Layer 0: SOVP CERTIFIED)
• Machine-readable API discoverability with structured access protocols (Layer 1)
• Federated identity and authorization compatible with agent-to-agent communication (Layer 2)

Agentic Architecture is distinct from AI capability: a company can deploy sophisticated AI internally but fail Agentic Architecture externally, making it invisible to the agents of its prospective clients.

SRS (Sovereign Readiness Score)

The Sovereign Readiness Score is the numeric aggregate output of a SOVP Full Validator scan, expressed on a scale of 0 to 100. It reflects the weighted aggregation of all current cluster scores.

A domain is CERTIFIED only when the SRS meets the SOVP threshold requirements across all mandatory clusters. A high SRS does not guarantee CERTIFIED status if a hard-gate cluster threshold is not met.

Reference values

Current SOVP Benchmark v3.0 snapshot (5 September 2026): frozen snapshot of 29 October 2026, cut-off 5 September 2026, without one internal test entry and the sample list (three of its domains also appear in a market list and count): 1,525 domains, mean SRS 43.57, median 44, p25 36, p75 51, p90 58, mean AI readiness 11.03 (originally published: 1,525 / 43.48 / 10.97; see the SOVP Benchmark report). The 559-domain benchmark from v2.0 remains a historical, methodologically distinct baseline.

SOVP Benchmark data →

CERTIFIED / NOT_CONFORMANT / INCOMPLETE

The verdict issued by a SOVP Full Validator scan is CERTIFIED, NOT_CONFORMANT, or INCOMPLETE — no other intermediate states. Every threshold is a hard gate; every cluster either passes or it does not, provided the scan could measure it.

Three-state since when

The verdict has been three-state since 3 Sep 2026. Runs scanned before that date carry the earlier values NOT_READY and PARTIAL; in the benchmark with cut-off 5 Sep 2026 that applies to 1,473 of 1,525 domains, because they were measured in late August.

The verdicts in detail

CERTIFIED means: the scanned infrastructure meets all SOVP protocol thresholds across all structured validation clusters. The result is institutionally usable: as evidence of agentic commerce readiness, in vendor qualification processes, and as a verifiable signal for autonomous agents.

NOT_CONFORMANT means: the scan completed, but at least one cluster threshold was not met. The detailed report identifies which clusters failed and what infrastructure changes would move the domain toward CERTIFIED status on a re-scan.

When the scan reaches too little

INCOMPLETE means: the scan reached too little of the domain to settle a verdict, typically because a firewall or bot-protection system blocked it. It states something about reachability at scan time and leaves the question of infrastructure quality open. The report withholds the score and points to a re-scan as the next step. Introduced 2026-08-01. Scans before this date distinguished CERTIFIED and FAILED alone and were not reclassified retroactively.

What CERTIFIED is not

CERTIFIED is a technical infrastructure evidence, not a regulatory certification. It is analogous to a penetration test report: an independently verifiable, reproducible record of infrastructure state at a specific point in time.

SOVP QuickScan

The SOVP QuickScan is a free validation tool that checks the mandatory infrastructure parameters across 4 weighted clusters: AI-Readiness (40%), Infrastructure (25%), Meta (20%), Structure (15%). Two corrections outside the clusters are added: up to 5 bonus points for an llms.txt and a content-noise correction from −5 to +5 points. No login is required. The scan runs on a Cloudflare server. A usage counter keeps score and verdict for 90 days. The result is visible only to the user.

QuickScan verdicts:

• CRITICAL (0–59): Fundamental Layer-0 signals missing, invisible to agentic systems.
• DEVELOPING (60–89): Partial compliance, detectable but not trustworthy to autonomous agents.
• ADVANCED (90–99): Strong compliance, approaches the threshold for Full Validator certification.
• UNVERIFIED (100): Maximum QuickScan score, but cryptographic DNS-anchoring not yet validated, requires Full Validator.

Renamed tier

Renamed 2026-08-01: this tier was previously labeled INCOMPLETE. Renamed to avoid collision with the unrelated Full Validator INCOMPLETE verdict (scan could not reach the domain) — the two are unrelated concepts that happened to share a name. Older blog posts or screenshots may still show "INCOMPLETE" for this QuickScan tier.

QuickScan and Full Validator

The QuickScan is a directional signal, not a substitute for the Full Validator. Only the Full Validator produces a CERTIFIED verdict and, with it, an Ed25519-signed identity document anchored in DNS. The verdict is reported in the scan result and is not part of the signed identity document.

Launch QuickScan →

SOVP Full Validator

The SOVP Full Validator is the complete SOVP scan. Until October 2026 it was a product of Litzki Systems LLC in its own right; today the engine runs inside the SEO portal and is used only in individual engagements. It runs deterministic parameters across structured validation clusters using the Ψ_core Engine, and produces a CERTIFIED, NOT_CONFORMANT, or INCOMPLETE verdict. On CERTIFIED it issues an Ed25519-signed identity document anchored in DNS; the verdict is not part of that signed document.

The path to CERTIFIED

The Full Validator is the only path to a SOVP CERTIFIED status. Its output is audit-grade: the result is DNS-anchored, machine-verifiable, and directly usable as evidence in vendor qualification processes.

SEO reporting service →

SOVP Validator Portal

The SOVP Validator Portal was the monitoring product of Litzki Systems LLC until October 2026 (plans with rescans, dashboard, Vault record, and score badge). It is closed to clients and used only internally and for demonstrations.

Ψ_core Engine

The Ψ_core Engine (pronounced "psi-core") is the deterministic scan engine that powers the SOVP Full Validator. It processes deterministic infrastructure parameters against the SOVP protocol specification and produces the weighted cluster scores that aggregate into the Sovereign Readiness Score.

The Ψ_core Engine is deterministic: given the same infrastructure state, it produces the same result. It does not use probabilistic models, ML scoring, or statistical approximation. The engine runs on a dedicated Hetzner server infrastructure in Nuremberg, Germany (GDPR-compliant jurisdiction).

Ψ_core in the protocol

Distinct from this is Ψ_core in the protocol: the binary signature verification function from draft-litzki-sovp-04, Section 4, Ψ_core = Verify(K_pub, σ, JCS(M)). It returns 1 or 0, not a score (see Sovereign Validation Protocol).

The Ψ_core Engine is patent pending.

Ed25519-Signed Identity Document

On a CERTIFIED Full Validator result, SOVP issues an Ed25519-signed identity document. Ed25519 is a high-performance elliptic-curve signature scheme that produces a compact, unforgeable proof of data integrity.

The signature attests that the identity document is authentic, unmodified, and attributable to a specific domain at a specific point in time. It is DNS-anchored: the domain's DNS records store the verification key, making independent verification possible by any party without contacting Litzki Systems. The verdict (CERTIFIED, NOT_CONFORMANT, INCOMPLETE) is reported in the scan result and is not part of the signed identity document.

The difference

The signed identity document is what distinguishes SOVP from a conventional audit report: the domain's identity is machine-verifiable by autonomous agents without human intermediaries.

Entity Map

An Entity Map is a machine-readable declaration of an organization's identity, relationships, products, and structural properties. The domain serves it at /.well-known/entitymap.json, where AI agents, LLM crawlers, and knowledge graph spiders consume it during entity classification.

An Entity Map provides ground truth: deterministic, structured, canonically linked identity data that overrides probabilistic inference. Agents relying on HTML content and probabilistic pattern matching are susceptible to misclassification, incomplete extraction, and hallucination. The Entity Map eliminates that surface.

LLM-Readiness

LLM-Readiness measures the degree to which large language model crawlers, RAG pipelines, and retrieval systems can correctly process a domain's infrastructure. It is not a measure of content quality. It measures the structural properties that determine whether automated systems can reliably ingest, de-duplicate, and canonically resolve a domain's entity data.

Signals

Key LLM-Readiness signals include: structured data completeness (JSON-LD), canonical URL consistency, machine-readable entity declarations (Entity Map, entitymap.json), AI crawl policy files (ai.txt, llms.txt), and absence of entropy-generating structures that cause RAG pipeline collapse.

The SOVP Full Validator AI-Readiness cluster assesses LLM-Readiness. Benchmark mean AI readiness: 11.03 across 1,525 domains (SOVP v3.0, September 2026, SOVP Benchmark).

Knowledge Graph Readiness

Knowledge Graph Readiness describes the structural readiness of a domain's entity data for ingestion into knowledge graphs: Google's Knowledge Graph, Wikidata, Crunchbase, and industry-specific graphs used by procurement agents.

A domain with high Knowledge Graph Readiness has: a correctly structured Organization schema with unique identifiers (LEI, EIN, Wikidata QID), external cross-references that allow de-duplication, consistent canonical naming across all channels, and machine-readable entity declarations linked from the domain root.

Domains that fail Knowledge Graph Readiness are either absent from knowledge graphs or misclassified, both outcomes make them invisible to agentic procurement workflows that rely on graph-based entity resolution.

Machine-Readable Infrastructure

Machine-Readable Infrastructure is the configuration of a domain's technical properties: DNS records, HTTP headers, structured data, protocol endpoints, and identity signals, such that autonomous agents, LLM crawlers, and knowledge graph spiders can process them without human intermediaries or natural language interpretation.

Machine-Readable Infrastructure is the precondition for Agentic Architecture. A domain can have excellent human-facing content and poor Machine-Readable Infrastructure, making it effectively invisible to the systems that drive agentic commerce decisions.

SOVP measures Machine-Readable Infrastructure deterministically.

Deterministic Validation

Deterministic Validation produces the same result for the same input, regardless of model version, crawl timing, statistical noise, or operator interpretation. This is in direct contrast to probabilistic approaches (conventional SEO audits, AI-based scoring systems, and machine learning classifiers), where the same infrastructure can yield different scores at different times.

Determinism in SOVP

SOVP is deterministic. Given a specific infrastructure state, the Ψ_core Engine produces the same cluster scores and the same verdict every time. This property is essential for institutional use: a CERTIFIED verdict that could change without any infrastructure change is not useful as evidence.

Deterministic Validation is the architectural property that makes SOVP results audit-grade and machine-verifiable by third parties without re-running the scan.