Every page, as clean Markdown an AI can read.

What it is

A second, machine-readable copy of every page.

Your site is built for human eyes — layout, scripts, styling. An AI agent has to dig the meaning out of all of it. The Knowledge Layer does the translation once: every page gets a clean Markdown document (with its schema.org structured data embedded) at a direct URL. Same content, a readable version underneath — what a screen reader did for people, for the machines that now read the web.

How agents access it

Four URL patterns, one content-negotiation header.

• /KnowledgeLayer/{domainKey}/{contentHash}.mdThe canonical, content-addressed Markdown for a specific page snapshot.
• /KnowledgeLayer/{domainKey}/by-url?u=…The page-stable resolver — maps any page URL to its latest Markdown, no content hash needed. Add &redirect=1 for a 302 to the cacheable .md URL.
• /KnowledgeLayer/{domainKey}/llms.txtThe per-domain index — one entry point listing every page as a Markdown link. New to the format? See What is llms.txt.
• /KnowledgeLayer/{domainKey}/organization.mdThe bundle’s typed organization node — who is behind the site (name, website, contact, official profiles), per Google’s Open Knowledge Format (v0.1 draft).

The resolver honours Accept: text/markdown (and text/*, */*, or no preference). A request that explicitly excludes Markdown — e.g. Accept: application/json — gets 406 Not Acceptable.

# Resolve any page URL to its clean Markdown (content negotiation)
curl "https://seogence.com/KnowledgeLayer/yoursite.com/by-url?u=https://yoursite.com/product" \
  -H "Accept: text/markdown"

# Discover everything the site exposes, as one index
curl "https://seogence.com/KnowledgeLayer/yoursite.com/llms.txt"

# Prefer a cacheable, content-addressed URL? Ask for a redirect
curl -I "https://seogence.com/KnowledgeLayer/yoursite.com/by-url?u=https://yoursite.com/product&redirect=1"
# -> 302  Location: /KnowledgeLayer/yoursite.com/9f8e2c....md

Output format

Frontmatter, clean body, and your structured data — inline.

Each document opens with YAML frontmatter (title, url, description, image…) plus typed OKF keys (type, okf_version, resource, timestamp, schema_type), then the page body converted to Markdown with navigation, scripts and styling stripped, then your schema.org JSON-LD preserved verbatim in a fenced block.

---
title: "Luxury Decor Bowl"
url: "https://yoursite.com/product"
description: "Handmade in Sweden."
image: "https://yoursite.com/bowl.jpg"
type: product
okf_version: "0.1"
resource: "https://yoursite.com/product"
timestamp: 2026-06-28T09:14:03.000Z
schema_type: "Product"
---

# Luxury Decor Bowl

Handmade luxury-decor bowls, finished by hand. Food-safe, one of a kind.

## Structured data

```json
{
  "@context": "https://schema.org",
  "@type": "Product",
  "name": "Luxury Decor Bowl",
  "offers": { "@type": "Offer", "price": "24.00", "priceCurrency": "USD" }
}
```

---

*Structured content by [Seogence](https://seogence.com).*

Now an OKF bundle

Typed for what reads the web next.

Your Knowledge Layer is now an OKF bundle — conformant with Google’s Open Knowledge Format (v0.1 draft). Every page’s frontmatter carries a required type, and a typed organization.md node describes who is behind the site. To be plain about it: no public crawler consumes OKF bundles today, and Google says OKF has no effect on Search ranking — it is a format, not a platform, and it works alongside your schema.org JSON-LD and llms.txt with nothing replaced.

Read the full OKF reference →

Response headers

Built for agents — with the token savings, quantified.

x-markdown-tokens vs x-original-tokens lets a client measure the savings before it spends a single token. (Example values shown.)

Discovery

Found automatically — even without JavaScript.

The embed injects a <link rel="alternate"> on each page. For non-JS crawlers, paste one static tag into your <head> — it points at your llms.txt index and works on every page with no per-page templating.

<link rel="alternate" type="text/markdown"
      href="https://seogence.com/KnowledgeLayer/yoursite.com/llms.txt"
      title="Markdown for AI agents">

Enable it

Off by default. One snippet, any host.

The Knowledge Layer is opt-in per domain — turn it on in your Seogence admin (Domains → markdown copy). Add the Seogence snippet once and it works on WordPress, Shopify, Wix, Next.js, React, or plain HTML. It re-reads itself when your content changes, by content hash, so the AI copy never goes stale.

Questions

Common questions.