Google dropped something significant on June 12, 2026: the Open Knowledge Format (OKF) — a new open standard for making website content readable by AI agents. Not just crawlable. Not just indexable. Actually readable — structured, linked, and navigable the way a human researcher would want it.
We implemented Google OKF and llms.txt across all 698 articles on The Universal Commerce Protocol the same week it launched. Here’s exactly what we did, why we did it, and what it means for how AI agents like ChatGPT, Perplexity, and Gemini will find and cite our content going forward.
What Is Google’s Open Knowledge Format (OKF)?
OKF is deceptively simple. It represents your site’s knowledge as a directory of Markdown files with YAML frontmatter — readable by humans and AI agents alike, with no SDK, no proprietary platform, and no lock-in required.
Each file describes one “concept” — in our case, an article, page, or content category. The file contains:
- YAML frontmatter — structured metadata: title, description, URL, tags, publish date
- Markdown body — clean readable text
- Cross-links — markdown links connecting related articles, creating a knowledge graph
Here’s what a concept file looks like for one of our articles:
---
type: Article
title: "How UCP Enables Zero-Click Commerce"
description: "An exploration of how the Universal Commerce Protocol enables AI agents to complete purchases without human interaction."
resource: https://theuniversalcommerceprotocol.com/zero-click-commerce
tags: [agentic-commerce, UCP, zero-click]
published: 2026-05-14T00:00:00+00:00
timestamp: 2026-06-18T17:53:26+00:00
---
# How UCP Enables Zero-Click Commerce
[Article content in clean readable text...]
## Related Articles
- [Agent Commerce Architecture](/posts/agent-commerce-architecture.md)
- [UCP Protocol Overview](/posts/ucp-protocol-overview.md)
Those cross-links between files are what turn a folder of documents into a knowledge graph — something AI agents can actually traverse rather than just scan.
Why We Did This
The way AI systems consume web content is changing fast. In 2026, ChatGPT, Perplexity, Gemini, and Claude are not just search engines — they are answer engines that cite sources, summarize content, and increasingly act as the primary interface between users and information.
If your content isn’t structured in a way these systems can understand, you’re invisible to them — regardless of how good your writing is or how strong your SEO is.
OKF solves a specific problem Google called the “context assembly problem”: when an AI agent needs to answer a question, it has to piece together context from dozens of fragmented, incompatible sources. OKF eliminates that by giving agents a single structured bundle they can read, navigate, and understand in one pass.
For a site like ours — 698 articles covering the Universal Commerce Protocol, agentic commerce, and AI-native trade infrastructure — this is directly relevant. These are exactly the topics AI agents are being asked about. We needed to make sure our content was the source they cited accurately.
What We Built
Rather than manually creating hundreds of Markdown files, we built a custom WordPress plugin that handles everything automatically.
The Plugin Does Four Things:
1. Bulk generates OKF for all existing content One click in the WordPress dashboard processes all 698 articles and 8 pages, creating a structured concept file for each one.
2. Auto-updates on publish Every time we publish or edit an article, the plugin automatically regenerates that article’s concept file and refreshes all indexes. No manual work required.
3. Serves the bundle at /okf/ The entire knowledge bundle is publicly accessible at structured URLs AI agents can discover and crawl:
theuniversalcommerceprotocol.com/okf/index.md— master entry pointtheuniversalcommerceprotocol.com/okf/posts/index.md— all 698 articlestheuniversalcommerceprotocol.com/okf/categories/— content by topictheuniversalcommerceprotocol.com/okf/pages/— static pages
4. Generates and maintains llms.txt We also generate a llms.txt file at our site root — a plain text signpost that tells AI crawlers what our site is about and where to find the OKF bundle. Think of it as robots.txt but for AI agents.
The Bundle Structure
After running the plugin on our 698 articles, here’s what our OKF bundle looks like:
/okf/
├── index.md ← Master entry point (AI reads this first)
├── posts/
│ ├── index.md ← All 698 articles listed with dates
│ ├── ucp-protocol-overview.md
│ ├── agentic-commerce-guide.md
│ └── ... (698 files)
├── pages/
│ ├── index.md
│ └── ... (8 pages)
└── categories/
├── index.md
├── agentic-commerce.md ← 188 articles
├── strategy-business-impact.md ← 201 articles
├── protocol-technical-architecture.md ← 149 articles
└── ... (10 categories)
The category files are particularly valuable. An AI agent reading categories/agentic-commerce.md instantly sees all 188 articles in that topic, linked and dated, without having to read all 698 files first.
OKF vs. What We Were Already Doing
We already had Yoast SEO handling JSON-LD structured data on individual pages. OKF is complementary, not a replacement.
| JSON-LD / Schema.org | OKF | |
|---|---|---|
| Purpose | Tells AI what each page is | Tells AI how content connects |
| Scope | Per-page | Whole site |
| Format | Embedded in HTML | Standalone .md files |
| Cross-links | No | Yes — knowledge graph |
| Agent navigation | No | Yes |
JSON-LD answers: “What is this article about?” OKF answers: “How does this article relate to everything else we publish?”
Both matter. We now have both.
What AI Agents See Now
When an AI agent visits our site and discovers the OKF bundle, here’s what it gets from the master index:
# Universal Commerce Protocol
The definitive resource on Google's open standard for agentic commerce —
developer guides, implementation resources, and breaking news for builders
working with UCP.
This OKF knowledge bundle contains structured knowledge extracted from
698 articles and 8 pages.
## Categories
- Strategy & Business Impact — 201 articles
- Agentic Commerce — 188 articles
- Protocol & Technical Architecture — 149 articles
- Merchant Integration — 119 articles
- Developer Guides — 94 articles
...
From there it can follow links into any category or article, traverse cross-links between related content, and build an accurate picture of our knowledge base — without guessing, without misattributing, without hallucinating details.
What Is llms.txt — And Why We Added It
Alongside Google OKF, we also implemented llms.txt — a growing standard for AI discoverability that works hand-in-hand with OKF. Think of it as robots.txt but written for AI agents instead of search engine crawlers.
We added theuniversalcommerceprotocol.com/llms.txt — a simple plain text file that functions as a first-impression brief for any AI system that visits our site. When ChatGPT, Perplexity, Claude, or Gemini crawls our domain, this is what it reads first:
# Universal Commerce Protocol
# https://theuniversalcommerceprotocol.com
> The definitive resource on Google's open standard for agentic commerce.
## Key Resources
- OKF Knowledge Bundle: /okf/index.md
- All Articles: /okf/posts/index.md
- Categories: /okf/categories/index.md
- Sitemap: /sitemap.xml
It’s a small file with a disproportionate impact. llms.txt frames how AI systems understand our site before they read a single article, and it points them directly to our Google OKF bundle so they can explore the full knowledge graph from there.
OKF + llms.txt together is the complete package: llms.txt is the door, OKF is the library inside.
Why This Matters for UCP Specifically
The Universal Commerce Protocol is about the infrastructure for AI-native commerce. Our readers are the builders, developers, and strategists working on agentic systems — the same people building the AI tools that will consume structured knowledge like OKF.
Being an early adopter of OKF isn’t just a technical improvement for us. It’s a signal about what we believe the future of the web looks like: structured, agent-readable, and built for a world where AI is the primary interface between content and people.
If you’re publishing content that AI agents should know about and cite accurately, OKF is worth implementing now. The format is simple, the spec is one page, and — as we just proved — you can go from zero to 698 structured concept files in an afternoon.
Do It Yourself
Google OKF resources:
- Full specification on GitHub: GoogleCloudPlatform/knowledge-catalog
- Builder reference, validator, and examples: openknowledgeformat.com
llms.txt resources:
- Specification and examples: llmstxt.org
If you’re on WordPress and want to implement Google OKF and llms.txt the same way we did, the approach is straightforward — the plugin we built handles bulk generation, auto-updates on publish, serving the bundle publicly at /okf/, and generating llms.txt automatically. The core logic is under 200 lines of PHP.
The AI agents are already reading. The question is whether your content is ready for them.
The Universal Commerce Protocol covers the infrastructure, standards, and implications of AI-native commerce. Our OKF bundle is live at theuniversalcommerceprotocol.com/okf/index.md.
Leave a Reply