A trust layer for the AI-readable web. One file tells machines what to believe about any entity — with proof.
KOS Protocol is an open standard for publishing machine-readable, provenance-tracked knowledge about any entity via a single JSON file called kos.json. It adds verifiable trust to web data by giving AI systems freshness decay scores, typed relationships, and confidence metrics for every data point — so they can cite facts instead of guessing.
When someone asks ChatGPT, Perplexity, or Gemini about a business, these systems scrape web pages and infer meaning from HTML. They have no way to know which facts are current, who verified them, or how much to trust them. The result: hallucinated phone numbers, wrong addresses, outdated services, invented credentials.
Machines parse HTML, extract fragments, and treat all data equally. A 3-year-old cached page has the same weight as yesterday's update. No provenance. No freshness signal. No confidence score. The AI fills gaps with plausible-sounding fabrications.
Machines fetch one structured file with every fact tagged: when it was created, when a human last confirmed it, who provided it, and how confident the data is. Stale data is mathematically flagged. The AI cites verified knowledge instead of guessing.
KOS Protocol is built on three ideas that work together to make knowledge computable.
Every data point in kos.json carries metadata about its origin: when created, when last verified by a human, who provided the information, and a confidence score from 0 to 1. Data confirmed by a human scores above 0.8. AI-generated data is capped at 0.8 — the range above that is reserved exclusively for human verification. This turns self-reported claims into auditable, weighted signals.
Information gets less reliable over time. KOS quantifies this using exponential decay: f = e(-λ × days). A business address has a half-life of 347 days. An event expires in 14 days. A core concept barely decays at all. This gives AI systems a real-time, computable trust signal — not a static label, but a number that changes every second.
Isolated facts are useful. Connected facts are powerful. KOS defines three relationship types: dependency (A requires B), reference (A cites B), and semantic (A relates to B). Cross-entity references link knowledge graphs across domains — creating a distributed web of verifiable knowledge that AI systems can traverse.
KOS does not replace Schema.org, robots.txt, or sitemap.xml. It fills the gap they leave open: computable trust.
Schema.org describes documents. KOS Protocol describes knowledge. A website can — and should — use both. kos.json can generate Schema.org markup, but Schema.org cannot express provenance, trust scores, or relationship types.
KOS Protocol is not theoretical. These three domains publish live kos.json files with cross-entity references, forming a distributed knowledge graph that AI systems can traverse right now.
The protocol itself. Specification, generator, consumer guide, structure checker, and registry — all self-described via its own kos.json.
Niseus Authority Systems. Digital infrastructure for local businesses. Implements KOS to describe its own services, methodology, and capabilities.
A concept discovery tool that maps connections between ideas. Uses KOS to describe its technology, research methodology, and content ecosystem.
On April 4, 2026, KOS Protocol was deployed at kosprotocol.dev. Within 19 hours, Perplexity and Grok independently discovered and cited the protocol — accurately reproducing the freshness decay formula, the 10 node types, and the 0.8 confidence cap for AI-generated data. No social media posts. No press release. No backlinks. The protocol needed existence, not marketing.
Define your entity. Map your knowledge to typed nodes — service, location, credential, person, content, concept, asset. Assign honest provenance to every node. The Generator provides complete, machine-executable instructions that any AI system can follow to build a valid kos.json for any entity.
Use the Structure Checker on the homepage to verify your file. It checks structural correctness, referential integrity, provenance completeness, and computes live freshness decay scores. Fix any issues before publishing.
Place the file at yourdomain.com/.well-known/kos.json and add a link element in your HTML head. AI agents that follow the KOS discovery convention will find it automatically — just like web crawlers find robots.txt.
Free. Open. Sovereign. No intermediary between your knowledge and the machines that need it.