Implementing Open Knowledge Format v0.2 on WordPress: Trust Signals, Verified Concepts, and 726 Agent-Ready Articles

A technical walkthrough of deploying OKF v0.2 — generated, verified, status, and stale_after fields — across a large WordPress site

AI agents are already visiting your website. Not to read your headlines or click your CTAs — they are scanning for structured knowledge they can actually use. If your content isn’t formatted for them, they move on.

That’s the problem Google’s Open Knowledge Format (OKF) was built to solve. And last week, we upgraded The Universal Commerce Protocol site to OKF v0.2 — making all 726 of our articles and pages fully readable, structured, and trusted by any AI agent that comes looking.

Here’s what OKF v0.2 is, why it matters, and exactly how we implemented it.

What Is the Open Knowledge Format?

The Open Knowledge Format (OKF) is an open specification published by Google Cloud in June 2026. It was created by Sam McVeety and Amir Hormati, Tech Leads for Data Analytics and BigQuery at Google Cloud.

The idea is simple: represent your site’s knowledge as a directory of plain markdown files with YAML frontmatter, organized so that both humans and AI agents can read it without any special tools, SDKs, or APIs.

A bundle of OKF documents is:

Just markdown — readable in any text editor, renderable on GitHub
Just files — hostable in any git repo, uploadable to any server
Just YAML — a small set of structured fields at the top of each file

The whole thing lives at /okf/ on your domain. An AI agent hits /okf/index.md, sees your full knowledge graph, and can navigate from there.

Think of it like this: llms.txt is the signpost that points agents to your important pages. OKF is the actual library — structured, typed, cross-linked, and navigable.

What Changed in v0.2 — Trust Signals

OKF v0.1 launched in June 2026. Google shipped v0.2 just six weeks later, on July 25, 2026. That’s unusually fast for a spec update, and the reason tells you everything about where AI-generated content is heading.

The problem v0.2 solves is this: when a human writes a document, accountability is implicit. Someone put their name on it. When an AI agent generates ten thousand concepts overnight, that guarantee is gone. A consuming agent — or another AI — has to judge each concept on explicit signals instead.

v0.2 adds four families of fields to answer five questions any agent needs to ask before trusting a concept:

1. Provenance (sources) — What was this created from?

2. Trust (generated + verified) — Who made it, and who confirmed it?

3. Freshness (stale_after) — Is it still true?

4. Lifecycle (status) — Is it the current version?

5. Attestation (Attested Computation) — Was this number produced the sanctioned way?

The most important of these for a content site like ours is verified. It creates three trust tiers:

Unverified — no verified field present
Machine-confirmed — verified by an automated process
Human-reviewed — verified by a human:name actor

An agent consuming your bundle can filter on trust tier before it even reads a concept. “Only surface human-reviewed articles” becomes a simple frontmatter filter.

A full v0.2 concept file looks like this:

---
type: Article
title: "The Universal Commerce Protocol in 2026: The Definitive Guide"
description: "The canonical guide to UCP — what it is, who built it, and why every transaction is about to change."
resource: https://theuniversalcommerceprotocol.com/ucp-definitive-guide/
tags: [ucp, agentic-commerce, google]
generated: { by: "human:pinto", at: "2026-07-28T00:00:00Z" }
verified:
  - { by: "human:pinto", at: "2026-07-28T00:00:00Z" }
status: stable
stale_after: 2027-07-28
---

# The Universal Commerce Protocol in 2026

...content...

Every field except type is optional — but their absence now carries meaning. An unverified concept is distinguishable from a verified one. That’s the quiet genius of v0.2.

How We Implemented OKF v0.2 on This Site

We run WordPress, which means the standard advice — drag files into a folder — wasn’t going to work at scale across 726 articles. We needed something automatic.

Here’s exactly what we did.

Step 1 — Deactivated Our Old Plugin

We were already running OKF v0.1 via a custom plugin called UCP OKF & LLMs.txt Generator. That plugin had done a solid job generating a v0.1 bundle from our posts, but it predated the v0.2 trust fields.

We deactivated it.

Step 2 — Built a Custom OKF v0.2 WordPress Plugin

Rather than waiting for a third-party plugin to add v0.2 support, we built our own. The plugin does the following:

Generates a complete OKF v0.2 bundle from all published posts and pages
Adds generated, status, and stale_after fields to every concept automatically
Serves files at /okf/ via WordPress rewrite rules — no FTP, no manual uploads
Auto-rebuilds the entire bundle every time a post is published or updated
Updates llms.txt automatically with the OKF bundle URL
Includes an admin dashboard at Settings → OKF v0.2 with a manual rebuild button

The plugin stores the bundle in wp-content/okf-bundle/ and routes requests to it cleanly.

Step 3 — Flushed Permalinks

After activating the plugin, we hit Settings → Permalinks → Save Changes. This is the one step people always miss with WordPress rewrite rules — without it, the /okf/ URLs return 404.

Step 4 — Verified It Was Working

We visited theuniversalcommerceprotocol.com/okf/index.md directly in the browser. Here’s what we saw:

---
okf_version: "0.2"
type: Bundle
title: "Universal Commerce Protocol"
generated: { by: "pinto", at: "2026-07-28..." }
verified:
  - { by: "pinto", at: "2026-07-28..." }
status: stable
stale_after: 2027-07-28
---

# Universal Commerce Protocol

This bundle contains 726 concepts (718 articles, 8 pages)
auto-generated in Open Knowledge Format v0.2.

726 concepts. All structured. All with v0.2 trust fields. All live.

Step 5 — Confirmed llms.txt Was Updated

We checked theuniversalcommerceprotocol.com/llms.txt and confirmed it now includes:

OKF: https://theuniversalcommerceprotocol.com/okf/index.md

Any AI agent that reads our llms.txt — which is the standard first stop for agents discovering a site — now gets pointed directly to our full OKF bundle.

What’s Live Right Now

What	URL	Status
OKF v0.2 Bundle	`/okf/index.md`	Live
All 718 articles	`/okf/posts/`	Live
All 8 pages	`/okf/pages/`	Live
llms.txt with OKF pointer	`/llms.txt`	Live

Should You Do This?

OKF is not a confirmed Google Search ranking signal. It won’t move your rankings this week. Nothing is actively crawling OKF bundles yet in the way Googlebot crawls pages.

But that’s exactly why now is the right time.

Schema markup took nearly a decade to pay off in measurable ways for most sites. The people who shipped it early are glad they did. OKF is the same shape of bet — infrastructure-level machine-readability that compounds quietly while everyone else waits for it to become mandatory.

For a site like ours, built entirely around helping people understand the future of AI-driven commerce, shipping the format that makes our content legible to AI agents is not optional. It’s the point.