TL;DR Summary:
Open Knowledge Format: Google introduced a vendor-neutral specification that represents knowledge as a directory of markdown files with YAML frontmatter to help AI agents understand concept relationships.Graph Over Flat Text: This format turns a website into a graph of linked concepts instead of isolated pages, allowing agents to learn how ideas connect rather than reading content as flat text.Machine-First Discipline: Building this bundled graph forces creators to define clear relationships between concepts, revealing gaps in website meaning and enabling a future where the graph becomes the canonical source for AI.How Can I Make My Website Readable by AI Agents?
Google just released a format that helps AI agents understand how your ideas connect instead of reading your pages one at a time as flat text. It's called the Open Knowledge Format, and it turns your website's knowledge into a graph that shows relationships between concepts.
What Google's Open Knowledge Format Actually Does
On June 12, 2026, Google's data team published the Open Knowledge Format. OKF represents knowledge as a directory of markdown files with YAML frontmatter. Each concept gets its own markdown document. A table gets a file. A metric gets a file. An API gets a file.
The YAML block at the top carries queryable fields: type, title, description, resource, tags, and timestamp. The markdown body carries the explanation. Concepts link to each other with ordinary markdown links. Google calls this structure "a graph of relationships."
There's no runtime. No SDK. No build step. Google describes it in three phrases: "just markdown," "just files," "just YAML frontmatter."
The format targets internal company knowledge. The context Google says is "locked behind whichever surface created it." It's version 0.1, which Google calls "a starting point, not a finished standard." Nothing in the announcement mentions public websites.
That gap is the interesting part.
Why Knowledge Graphs Beat Flat Page Copies for AI Agents
The agent-readable version of your website is flat. An AI model reads each page as isolated text. Serving pages as markdown strips away design and navigation, but it keeps the same page-by-page structure. You get a clean copy of every page. You lose how those pages relate to each other.
A knowledge graph keeps the relationship layer. When your concepts link to each other, an agent learns what each concept is and how they sit relative to each other. That's most of what understanding a website means.
Two pages might both mention the same concept. One page explains the framework underneath it. The other describes a narrower goal beside it. Without explicit connections, a machine never learns which is which. A graph says it outright through links the machine follows.
Google's Open Knowledge Format is an off-the-shelf way to build that graph. Markdown makes it cheap. Structure makes it carry the relations.
Testing Google's Open Knowledge Format on a Real Website
I wrote an OKF bundle for the No Hacks website. One markdown file each for the brand, the host, Machine-First Architecture, the agentic web, Agent Experience Optimization, Answer Engine Optimization, llms.txt, and WebMCP.
Each file follows Google's conventions. YAML fields on top. Plain markdown body underneath. The work was deciding which concepts mattered and how they connect, not writing the files.
The Machine-First Architecture file looks like this:
---
type: framework
title: Machine-First Architecture
description: A framework for building websites whose full meaning is available to a machine reading them, with the human experience layered on top rather than the other way around.
resource: https://machinefirstarchitecture.com
tags: [Framework, Machine-First Architecture, Agentic Web]
timestamp: 2026-06-13
---
Machine-First Architecture is Sani's framework for the agentic web. The core idea: build the content so a machine reading it gets the complete meaning, the facts, the structure, the relationships, and the human reading gets that same meaning with the design on top. This is why formats that strip a website to plain text, like markdown for agents and llms.txt, matter. Its capability side is WebMCP, and its measurement side is Agent Experience Optimization.
Those bracketed links are the graph. An agent following them learns that WebMCP sits under Machine-First Architecture. It learns that llms.txt is the same kind of bet. A flat copy of my pages never says this out loud.
Across eight files, that's the whole structure: concepts and the relationships between them.
The Maintenance Cost of Parallel Knowledge Layers
A bundle like this is a second copy of what the website already says. A second copy means keeping two things in sync. The moment the website changes, the bundle is wrong until you update it too.
That tax applies to every parallel machine-readable layer. An llms.txt file. A markdown mirror of your pages. A bundle like this one. The version an agent reads is only as accurate as your discipline in keeping it current.
The bundle I created raises an immediate question: how do I know if an agent reading this understands the relationships I encoded? This is where measurement becomes critical. ClickRank lets you test how AI agents interpret your structured content by simulating agent queries and showing which resources they surface. For an OKF bundle, that means you verify whether your graph's connections—the links between Machine-First Architecture, WebMCP, and llms.txt—guide an agent to the right concept when asked, or whether the structure you built is invisible to the models reading it.
Google built OKF for internal company knowledge. Nothing in its plan points at public websites. Hosting a bundle for a visiting agent is off-label use. The agent I made it for, one that fetches the bundle and follows the graph, might never show up.
The reason to do it has to stand without that payoff. It does. Writing the bundle forced me to state plainly what No Hacks knows and how its ideas connect. That surfaced gaps I would not have found writing another page. It's the same discipline as Machine-First Architecture: put your meaning in a form a machine can read, and you find where you were vague.
Where Website Knowledge Graphs Built With Google's Open Knowledge Format Could Lead
None of what follows is a prediction. It's a direction. It depends on agents reading website knowledge graphs. Today, none do. The shape is still worth seeing.
The identity file could grow into a knowledge graph. Today, llms.txt is a single line announcing who you are. A published bundle is the full version of that idea. A map of everything your website knows and how the parts connect. The thin identity layer and the structured knowledge layer become one thing.
Agents could query that map instead of scraping your pages. An agent that pulls your bundle and follows its links gets a cleaner, relationship-aware read than one parsing your HTML one at a time. You get more say in how your concepts are represented when an AI describes you.
Testing whether that representation holds requires simulating agent behavior at scale. ClickRank runs queries through multiple AI models to show you which of your resources they cite. It reveals whether your knowledge graph's structure influences agent responses or gets ignored in favor of flatter content.
The map could become the canonical layer. The version a machine reads stops being a copy of your website and becomes the source. The human pages become one rendering of it. That's the fully machine-first website the agentic web has been pointing at, reached through a side door Google opened for internal data.
Why Markdown Keeps Winning for Machine-Readable Content
John Gruber created Markdown in 2004, with Aaron Swartz as his beta-tester. The design goal was readability: text you read as-is, without rendering, that converts cleanly to HTML. Two decades later, it runs GitHub, Reddit, most documentation you read, and the chat boxes of AI tools themselves. It won by being legible without being rendered. That's the exact property that makes it easy for a machine to read.
Machine-facing formats keep landing on the same ground. llms.txt. Cloudflare's markdown. Now OKF. Google itself is not of one mind about it. Its Search side called llms.txt "purely speculative" for ranking. Its Chrome side added an llms.txt check to Lighthouse's agent-readiness audit. Its data team has now published OKF.
Testing Whether Your Website Has a Relationship Layer
Open your most important page. Paste it into a plain-text editor. The links collapse into plain words. Look at what's left. Find anything that states how its ideas relate to the rest of your website. Not that one page links to another. The relationship itself.
There's usually nothing. That absence is what a knowledge graph fills, whether or not you touch OKF.
OKF is this week's news. The substrate under it, plain text a machine can read, has been here since 2004. What Google added was a standard and a name.
Measuring Whether Your Knowledge Graph Works
Building structured content for agents is one step. Knowing whether agents understand it is another. You've created relationships between concepts. You've written markdown files that link to each other. You've built a graph.
ClickRank shows you whether agents use your knowledge graph. It simulates how ChatGPT, Claude, Gemini, and Perplexity query and retrieve information. You see which parts of your structured content agents surface when answering questions about your domain. Without that feedback loop, you're maintaining a knowledge graph blind to whether agents are using it. See how AI agents interpret your content at ClickRank.
