How YAML-in-HTML Turns Any CMS Into an AI Memory Engine
YAML-in-HTML embeds structured, trust-scored knowledge into web pages that AI systems can retrieve and cite independently, turning any CMS into an AI memory engine without special plugins. It works with platforms like WordPress, providing fragment-level memory that AI can evaluate for trustworthiness.
Key Takeaways:
- YAML-in-HTML embeds structured, trust-scored knowledge directly into web pages.
- Unlike SEO formats like Schema.org, it supports fragment-level memory that AI systems can evaluate independently.
- Works with CMS platforms like WordPress—no plugins or infrastructure changes required.
- Part of the Semantic Digest Protocol, it helps publishers make content retrievable, citeable, and trustworthy for AI.
- Structuring content at the fragment level ensures it exists meaningfully in AI memory.
Why the Web Fails AI Systems
The web was made for people, not machines. Humans interpret context. AI struggles to know what's true, who said it, and where it came from.
Schema.org and JSON-LD improved SEO, not AI understanding. They operate at the page level, lack provenance, and don’t support claim-level retrieval. AI systems are left guessing.
"Since 2011, websites have used Schema.org and JSON-LD to feed SEO bots. But these formats were never designed for AI memory." — David Bynon
YAML-in-HTML: A Bridge Between CMS and AI
YAML-in-HTML changes that. It embeds machine-readable, trust-scored memory inside standard HTML using inert tags. No JavaScript. No plugins.
What It Looks Like
<template
data-visibility-fragment
data-type="text/yaml"
data-sdt-class="DataFragment"
data-entity="plan:H5521-290-0"
data-digest="2025-cms-ma-mapd-plan"
data-glossary-scope="cms_landscape"
data-fragment-scope="semantic-digest">
Inside is YAML content—easy to read, easy to export, and fully retrievable by agents.
Inside a YAML-in-HTML Fragment
1. YAML Structure
Defines what the fragment is, what it belongs to, and how to classify it.
data-sdt-class: DataFragment
entity: plan:H5521-290-0
digest: 2025-cms-ma-mapd-plan
glossary_scope: cms_landscape
fragment_scope: semantic-digest
2. ProvenanceMeta (Trust Block)
Documents the source, license, and retrieval details.
ProvenanceMeta:
ID: 2025-cms-ma-landscape
Title: CMS MA Landscape File, 2025
Creator: Centers for Medicare & Medicaid Services (CMS)
License: Public Domain
Published: 2025-06-01
Retrieved: 2025-06-28
Digest: 2025-cms-ma-mapd-plan
Entity: plan:H5521-290-0
3. Semantic Data Atoms
The smallest unit of AI-retrievable knowledge.
Fields:
defined_term: Primary Care Visit
value: "$0"
unit: usd
confidence: high
derived: false
glossary: term-in_primary
source: 2025-cms-pbp
provenance_ref: "#provenance-meta"
Fragment Classes: Memory with Structure
Different fragment types serve different AI needs.
DataFragment
Stores raw, structured facts (costs, stats, values).
DefinedTermFragment
Machine-readable glossary entries with provenance.
entity: term:zero_premium
Term:
term_id: zero_premium
name: Zero Premium Plan
definition: A Medicare Advantage plan that has no monthly premium beyond Part B.
FAQFragment
Question-and-answer pairs for retrieval without hallucination.
FAQ:
question: Are zero-premium Medicare Advantage plans available in all counties?
answer: No. Availability varies by county.
IndexFragment
Directories of entities (e.g., list of Medicare plans).
MetaFragment
High-level metadata about entire datasets.
Implementation: No Plugins Required
WordPress
Just add the block and YAML into a post or page.
- Fully inert and DOM-safe
- Invisible to humans, readable by crawlers
Static Sites (Jekyll, Hugo, Eleventy)
Embed YAML-in-HTML fragments as partials or template components.
Enterprise CMS
Define custom fields or server-side generators that output valid YAML-in-HTML. Key requirement: preserve the and YAML formatting.
Real-World Applications
Healthcare
- Plan comparisons, provider data, medical definitions
- Improves AI-generated health answers with facts
Finance
- Investment details, disclaimers, risk profiles
- Helps AI avoid outdated or incorrect financial info
Education
- Definitions, curriculum alignment, statistics
- Strengthens tutoring systems and reduces misinformation
Legal
- Statutes, citations, jurisdiction-specific rules
- Enables AI to cite exact legal text, not paraphrased guesses
Compatible with Modern AI Systems
- Model Context Protocol (MCP): Fragments can be queried by agents
- LLMs: YAML is ingestible directly or via conversion (e.g., JSON-LD)
- RAG: Fragment-level memory units with built-in trust
- Assistants/Agents: Clear citation and grounding from source fragments
"If MCP is the USB-C socket, YAML-in-HTML is the micro thumb drive. It’s small, lightweight, and universally pluggable." — David Bynon
A New Paradigm: Memory-First Publishing
Traditional Web:
- Optimized for crawlers
- Page-level markup
- Success = search rankings
Memory-First Web:
- Optimized for AI retrieval
- Fragment-level trust
- Success = accurate citation in AI answers
YAML-in-HTML brings this future within reach—no new frameworks, no vendor lock-in. It runs on the web we already have.
By adding these machine-readable fragments alongside human-readable content, publishers can serve both audiences effectively—ensuring their expertise is accurately represented in both human research and AI-assisted information retrieval. Learn more in this USA Today story.
For an in-depth overview of this methodology, see the original announcement on Medium.com (https://medium.com/@trust_publishing/the-web-just-got-a-memory-introducing-yaml-in-html-1491e5d2c8fb).
To learn more about implementing YAML-in-HTML and the Semantic Digest Protocol in your own projects, check out David Bynon's documentation at SemanticDigest.org.
