The Open Knowledge Format has landed
Why we’re paying attention to the first real standard for feeding AI your knowledge
On 13 June 2026, Google Cloud quietly released something we think more businesses should know about: the Open Knowledge Format, or OKF. If you haven’t heard of it, you’re not behind. It’s brand new, it’s only at version 0.1, and almost nobody outside the data and AI world is talking about it yet. We think that’s about to change, and here’s why we’re all over it.
First, what actually is it?
OKF is an open specification for capturing the knowledge that AI systems need in order to do useful work. Not the public information a model already learned from the internet, but your information: what a particular metric means inside your business, how two of your systems join together, the runbook for an incident, the meaning behind a column in a database, the deprecation notice for an old process.
The clever part is how little it asks of you. An OKF “bundle” is simply a folder of plain Markdown files, one file per concept, with a small block of structured tags at the top of each (things like type, title, description, a link to the resource, and a timestamp). Files link to one another with ordinary Markdown links, which turns the folder into a connected map of your knowledge.
That’s the whole idea. No new software to install, no proprietary platform, no special account. As the Google team puts it, an OKF bundle is just Markdown, just files, and just a little structured data. You can read it in any text editor, store it in any repository, and hand it to almost any AI agent without translation.
The problem it’s trying to solve
In most organisations, the knowledge an AI needs is scattered everywhere: in data catalogues, in wikis and shared drives, buried in code comments, and, frankly, inside the heads of a few senior people. When you ask an AI assistant a question like “how do we calculate active users?” it has to stitch an answer together from all of those incompatible sources, assuming it can reach them at all.
The result is that everyone keeps solving the same problem from scratch. Every tool invents its own way of storing knowledge, and that knowledge stays locked inside whichever system created it. OKF’s answer is refreshingly simple: agree on a shared format, so knowledge written once can be understood by anyone, and any tool, anywhere.
Why a standard matters, and why this is good news
The “knowledge as a living library” idea isn’t new. Plenty of teams already keep notes for their AI tools in formats like Obsidian vaults, or the AGENTS.md and CLAUDE.md files developers have adopted over the past year. The trouble is that every version has been bespoke. They look similar, but none were designed to work together, so the effort gets repeated again and again.
What’s been missing is a common format that anyone can produce, anyone can read, and that survives being moved between systems and organisations. OKF is an attempt to be exactly that. It’s the first credible, vendor-neutral standard in this space, and that’s genuinely worth celebrating. Standards are how messy, fragmented markets eventually grow up.
A word of caution: this is very early
We want to be honest about where OKF sits today. This is version 0.1. It is, by Google’s own description, a starting point rather than a finished standard. The full specification fits on a single page, and the tools shipped alongside it (an agent that drafts knowledge files from a database, and a simple viewer) are deliberately described as proofs of concept.
So expect it to change. The format is explicitly designed to evolve as more people use it and as we all learn what AI agents really need in practice. That’s not a weakness, it’s how good standards are built, in the open and with feedback. But it does mean anyone adopting it now should treat it as an early, fast-moving foundation rather than a settled rulebook. Build with it, but build flexibly.
Where we fit in
Here’s the thing: this isn’t really SEO, but it lives right next door to it. For years our job has been making sure the information about your business is accurate, well structured, and easy for search engines to find and trust. The rise of AI assistants extends that job in a new direction. Increasingly, the question isn’t only “can Google find you?” but “can an AI understand your business correctly when someone asks it a question?”
That’s precisely the gap OKF speaks to, and it’s why we see it as a natural part of what we do at True North Digital. Getting your company’s knowledge into a clean, accurate, well-maintained format for AI is the same discipline we’ve always applied to your web presence, just pointed at a new and growing audience: the agents your customers and your own teams are starting to rely on. If the knowledge an AI draws on is messy or wrong, the answers will be too, and that reflects on your brand.
So, have you thought about your own OKF?
It’s worth asking the question now, even while the format is young. What does your business actually know that an AI would need to represent you accurately? Your products and how they differ, your pricing logic, your policies, the way you describe your services, the meaning behind your key numbers. Today most of that is scattered or undocumented. Tomorrow it could be a tidy, AI-ready knowledge bundle that you own and control.
We’d love to help you think it through, whether that’s a quick audit of what knowledge you already have, a plan for capturing it, or a first OKF bundle built for your business. If you’d like to explore what this could look like for you, get in touch.
Further reading
Google Cloud’s announcement, the specification and sample bundles are on GitHub.
