What OKF Actually Is
The Open Knowledge Format (OKF) is remarkably simple. It is a standardised way to organise knowledge as a folder of markdown files with a small block of metadata at the top of each file.
Here is a working example. Imagine you run a tour company on the Great Barrier Reef. Your knowledge folder might look like this:
tour-company-knowledge/
├── index.md
└── concepts/
├── booking-cancellation-policy.md
├── weather-diversion-process.md
├── client-welcome-checklist.md
└── snorkel-equipment-maintenance.md
Each file represents one concept — a single unit of knowledge. Every concept file starts with a short block of structured data called YAML frontmatter:
---
type: Playbook
title: "Weather Diversion Process"
description: "Steps to follow when weather prevents a scheduled tour departure."
tags: ["operations", "safety", "weather"]
timestamp: 2026-06-01
---
After the frontmatter comes the body — plain markdown, whatever you want to write about that concept. Files link to each other using standard markdown links, turning the folder into a small knowledge graph.
That is most of the specification. No new software. No proprietary system. No account. The bundle can sit in a GitHub repo, a shared drive, or a zip file. Any text editor can open the files. Any compatible AI assistant can read them.
Why This Matters: Moving From Searching to Knowing
Most AI assistants today work by searching through your documents every time you ask a question. This approach, called retrieval-augmented generation (RAG), works well enough for simple lookups but struggles when an assistant needs to combine knowledge from multiple sources, understand your business logic, or follow a documented process step by step.
OKF offers a different path. Instead of searching raw documents each time, you give your assistant a structured library of your most important knowledge, organised by concept. The assistant can find the right file quickly, understand its context from the metadata, and follow links to related concepts.
For a business, this means:
- An internal assistant that answers "how do we handle a cancellation?" the same way every time, because it reads your actual policy, not a best guess.
- An onboarding assistant that a new staff member can ask "what is the first thing I do when a client calls?" and get a consistently accurate answer.
- A compliance assistant that references your documented procedures when preparing for an audit.
None of these require OKF specifically. You could do all of this with any folder of markdown files. What OKF adds is portability — the same knowledge bundle can work with different AI tools without being reformatted for each one.
The Bigger Idea: Knowledge That Compounds
OKF formalises a pattern popularised by AI researcher Andrej Karpathy called the "LLM Wiki." The idea is worth understanding because it changes how you think about knowledge maintenance.
Instead of treating every interaction as a fresh search, you give your AI a persistent wiki that it helps maintain. When you add a new document — a policy update, a new process, a client example — the AI reads it, extracts the key information, and integrates it into the existing knowledge. It updates relevant pages, notes where new information contradicts old assumptions, and discovers new connections between concepts.
Over time, your knowledge base gets richer without anyone doing the filing. The AI handles the bookkeeping. Your job is to provide the knowledge and check that it remains correct.
For a small business that cannot afford a dedicated knowledge manager, this is genuinely useful. You write something useful once. It gets organised into your knowledge base. It stays connected to everything else you have documented. And when something changes, you update it in one place, and the ripple effects flow through the system.
What This Looks Like for Real Businesses
OKF is too new to have real-world case studies. But the pattern it formalises is already practical. Here is how a business might use structured knowledge today.
Tour operator. Your cancellation policy and your weather diversion process are related, but they live in separate documents, written years apart. Organised as linked concepts with a mutual reference, a staff member or assistant can follow the connection without knowing it exists. The result: fewer situations where a customer is told "that policy was changed last month."
Fitness studio. Your membership terms, class schedule, injury protocols, and instructor availability are scattered across different systems. Organised as a knowledge bundle, an assistant helping with a rescheduling request can check the 24-hour cancellation rule, find alternative class times, and confirm the result against your documented policy — all in one pass.
Professional services firm. Your fee schedule, engagement letter templates, compliance obligations, and onboarding checklists change every quarter. A maintained knowledge bundle gives everyone — staff, contractors, and any approved AI tools — the same source of truth to reference.
In each case, the value is not in the format. It is in having the knowledge written down, structured, and linked in a way that can be reliably used.
What This Is Not
It helps to be clear about what OKF is not, because the early coverage has been speculative.
- OKF is not an SEO ranking factor. Google has not said it affects search results. Do not treat it as one.
- OKF is not a product. There is no OKF dashboard, no OKF subscription, no OKF button in Google Cloud. It is a specification with reference implementations.
- OKF is not required by any AI tool. No major platform currently requires you to publish or maintain OKF.
- OKF is not a fix for unreliable AI. Access controls, data quality, retrieval design, and human oversight still matter.
The hype around "selling knowledge bundles as a product" — where a lawyer or accountant packages expertise into an OKF you can buy — is speculation from commentators, not something Google published. It may happen. It may not. Do not plan your business model around it yet.
The Practical Path Forward
OKF is in draft. The ecosystem is immature. For most small businesses, rebuilding your entire knowledge base around a v0.1 spec would be premature.
But the underlying practice is useful right now: pick one area of your business, document it clearly as linked concepts, test it with an AI assistant, and see whether the results are more consistent than what you get from unstructured documents.
The free OKF Bundle Generator from Suganthan Mohanadasan at suganthan.com/okf-generator can convert public website pages into an OKF bundle for testing. Google's reference repository has a BigQuery enrichment agent and a static HTML visualiser, both labelled as proofs of concept.
Start with one thing. Your cancellation policy. Your onboarding checklist. Your most common client question. Structure it. Test it. Then decide whether the format matters or whether the real value was just having the knowledge written down properly in the first place.
The businesses that benefit most from AI assistants will not be the ones with the most documents. They will be the ones that make their knowledge usable.
If you would like to assess where your business knowledge currently lives and whether it is structured well enough for AI tools, Automic Consulting can help. We work with Australian SMBs on practical automation that reduces manual work and makes internal systems more reliable.