New from Google — June 2026

What is the Open Knowledge Format?

Google's new open standard that lets AI agents read your brand knowledge without any SDK — just markdown, just files, just YAML. And why every brand needs one in 2026.

The 30-second answer

Open Knowledge Format (OKF) is an open specification announced by Google Cloud on June 12, 2026. It turns your brand knowledge into a portable directory of markdown files — one file per concept — that any AI agent can read without a proprietary SDK, integration, or account.

Think of it as the structured, multi-file evolution of llms.txt. Where llms.txt is a single flat summary, OKF is a curated knowledge wiki: one file for your product overview, one for pricing, one for your methodology, one for key definitions — all interlinked, all machine-readable.

Why it matters: AI agents assembling answers about your brand currently have to guess from scattered web pages. An OKF bundle gives them a single, authoritative, structured source of truth. Brands with rich AI-readable context files earn more citations, more accurately, in Perplexity, ChatGPT browse mode, and agentic AI systems. The cost is a few markdown files. The upside is AI agents finally understanding your brand correctly.

The format in one screen

Every OKF document is a markdown file with a small block of YAML frontmatter at the top:

---
type: Product Overview
title: VibecodeAEO Platform
description: AI Brand Intelligence platform that tracks how often ChatGPT,
  Perplexity, and Gemini cite your brand and generates fixes to improve citations.
resource: https://vibecodeaeo.com/seo
tags: [aeo, ai-visibility, brand-intelligence]
timestamp: 2026-06-18T00:00:00Z
---

# VibecodeAEO Platform

VibecodeAEO measures your AI Visibility Score across five engines...

## Core Features

- [Citation tracking](product/overview.md)
- [Ranky AI strategist](product/how-it-works.md)
- [Pricing](product/pricing.md)

That is an entire OKF document. The YAML block provides structured fields the agent can query. The markdown body provides the human-readable content. Links connect documents into a knowledge graph.

The six YAML fields

type Required

What kind of concept this is — "Product Overview", "Pricing", "Reference Guide", "API", "Metric". You define the type taxonomy; OKF does not impose one.

title Recommended

Human-readable name for the concept. Keep it concise and specific — agents use this as the concept identifier.

description Recommended

One-paragraph plain-language summary. This is what AI agents read first when deciding whether this document answers their query.

resource Recommended

Canonical URL for this concept on your website. Links the OKF document back to the live page for agents that want the full context.

tags Optional

Array of topic labels. Used for faceted queries and cross-linking within the bundle. Keep tags lowercase and hyphenated.

timestamp Optional

ISO 8601 last-modified date. Helps retrieval agents prioritize fresh content. Update it whenever the file changes.

Bundle directory structure

An OKF bundle is a directory you can serve anywhere — a subdirectory of your website, a git repo, a tarball. The recommended structure:

mybrand-okf/
├── index.md              # Bundle root — overview and navigation
├── product/
│   ├── overview.md       # Product features and capabilities
│   └── pricing.md        # Pricing, plans, free trial
├── reference/
│   ├── what-is-aeo.md    # Key concept definitions
│   └── glossary.md       # Terminology index
├── concepts/
│   └── index.md          # Concept map and cross-links
└── log.md               # Change history (optional)

The index.md at the root is the entry point. List every other file here with a one-line description so agents can navigate without crawling the whole tree.

OKF vs llms.txt — use both

Dimensionllms.txtOKF bundle
FormatSingle flat markdown fileDirectory of markdown files
LocationRoot of domain (/llms.txt)Any path (/okf/index.md)
DepthLinks + one-paragraph summariesFull structured content per concept
Best forQuick AI agent orientationDeep brand context and accuracy
Agent reads whenFirst visit to a domainNeeds depth on a specific concept
Created byJeremy Howard, 2024Google Cloud, June 2026

Ship both. llms.txt is the quick orientation file; your OKF bundle is the deep knowledge base. Together they give AI agents everything they need to represent your brand accurately.

Three principles behind the design

📄
Minimally opinionated

OKF requires exactly one thing: a type field. Everything else — what types exist, what other fields to include, how to structure the body — is left to you. The spec defines the interoperability surface, not the content model.

🔌
Producer/consumer independence

Bundles hand-authored by humans can be consumed by AI agents. Bundles generated by a pipeline can be browsed in a visualizer. The format is the contract; the tooling at each end is independently swappable.

🌐
Format, not platform

OKF is not tied to Google Cloud, BigQuery, or any vendor. No proprietary account needed to read, write, or serve it. It is a vendor-neutral open standard — its value comes from how many parties speak it.

How to ship your OKF bundle in 15 minutes

  1. Identify your top 5-10 concepts — product overview, pricing, key features, methodology, definitions for terms you own.
  2. Create one markdown file per concept with the YAML frontmatter block at the top.
  3. Write an index.md that lists every file with a one-line description — this is the bundle entry point for AI agents.
  4. Cross-link your documents with normal markdown links [text](path.md) so agents can navigate the knowledge graph.
  5. Serve the directory at /okf/ on your domain so it is accessible at https://yourdomain.com/okf/index.md.
  6. Add a reference in llms.txt: - [OKF Knowledge Bundle](https://yourdomain.com/okf/index.md): Structured brand context

Or skip the manual work — use the free generator and download a complete OKF bundle for your brand in seconds.

FAQ

What is the Open Knowledge Format (OKF)?
OKF is an open specification announced by Google Cloud on June 12, 2026. It represents brand knowledge as a directory of markdown files with YAML frontmatter, making that knowledge portable and readable by any AI agent or tool without requiring a proprietary SDK or integration.
How is OKF different from llms.txt?
llms.txt is a single flat file listing your most important pages and a short brand summary. OKF is a richer multi-file bundle — each concept (product, metric, process, API) gets its own dedicated file with structured fields. OKF and llms.txt complement each other: ship both for maximum AI context coverage.
Who created OKF?
OKF was created by Google Cloud engineers Sam McVeety and Amir Hormati, announced on June 12, 2026. It formalizes the LLM-wiki pattern pioneered by Andrej Karpathy into a vendor-neutral open standard.
Do I need an SDK or account to use OKF?
No. OKF is just markdown files with YAML headers. You can create them in any text editor, host them in a git repo or on your web server, and any AI agent can read them. No SDK, no runtime, no proprietary account required.
Where do I serve my OKF bundle?
Serve your OKF bundle at a consistent path on your domain — a common convention is /okf/ (e.g. https://yourdomain.com/okf/index.md). You can also publish it as a GitHub repo or include it in your documentation site.
Will OKF help me get cited more by AI engines?
Yes. OKF gives AI agents structured, curated context about your brand that they can read without guessing from scattered web pages. Brands with rich, machine-readable context files are cited more accurately and more frequently by retrieval-based engines like Perplexity and ChatGPT browse mode.

Generate your OKF bundle automatically

VibecodeAEO builds a complete Open Knowledge Format bundle for your brand in seconds. Free, no account required.