MX Learn · Lesson 4 of 6
Key principles
The techniques MX uses at the discovery layer, and the governance envelope it adds above them. Reading time about six minutes; the full lesson text is on this page even with scripting off.
Two layers, both required
Earlier lessons drew the line: the discovery layer is where an agent finds your page and parses what it claims, and the governance envelope is where it works out whether to act on what it just read. This lesson covers both sides of that picture in practice.
At the discovery layer, MX invents nothing. It defers to three established techniques and ensures each is present, complete, and correct: structured data, accessibility, and explicit intent. Above them, MX adds the governance envelope no existing standard covers.
Both sides are required. A page strong at the discovery layer with no governance envelope is the £203,000 cruise misread waiting to happen; the case study is in Appendix I of MX: The Protocols.
Technique 1: structured data
Core principle: Stop making machines guess. Every piece of information that matters to visitors - prices, contact details, reviews, hours, availability - should be marked up with Schema.org vocabulary so machines can parse it with certainty.
Schema.org is a collaborative vocabulary maintained by Google, Microsoft, Yahoo, and Yandex. It provides standardised ways to mark up content so machines can understand it. Open the card to see the same product twice:
Without structure
<div class="product"> <div class="name">Wireless Headphones</div> <div class="price">$129.99</div> <div class="rating">4.5 stars (230 reviews)</div> </div>
The AI agent sees three generic divs with text. It might infer meaning, or it might not.
With Schema.org
<div itemscope itemtype="https://schema.org/Product">
<span itemprop="name">Wireless Headphones</span>
<div itemprop="offers" itemscope itemtype="https://schema.org/Offer">
<span itemprop="price">129.99</span>
<meta itemprop="priceCurrency" content="USD">
</div>
<div itemprop="aggregateRating" itemscope itemtype="https://schema.org/AggregateRating">
<span itemprop="ratingValue">4.5</span> stars
(<span itemprop="reviewCount">230</span> reviews)
</div>
</div>
The AI agent sees a Product entity with an explicit name, an offer with price and currency, and an aggregate rating with value and count. No guessing required.
Open the card to continue.
Techniques 2 and 3: accessibility and explicit intent
The other two discovery-layer techniques are just as established. Open both:
Core principle: Accessible design IS agent-compatible design. The things that make websites accessible to humans with disabilities are the same things that make them readable by any machine. Assistive technologies and AI agents face identical challenges: they need explicit structure, not visual inference.
WCAG 2.1 AA compliance requires semantic HTML (agents understand nav, main, and article instantly), a proper heading hierarchy that never skips levels, alt text that describes rather than decorates, and form labels that connect inputs to purposes. Every WCAG fix you make simultaneously improves agent compatibility.
Core principle: Don't make machines infer. Declare. Humans are excellent at inferring meaning from context. Machines are not. If something is disabled, required, loading, or conditional, say so explicitly in your markup, not just visually.
A greyed-out button looks disabled to a human; an AI agent sees a fully functional button unless it carries the disabled attribute. An asterisk next to a form field means "required" to a human; the agent needs the required attribute. A spinner means "loading" to a human; the agent needs an announced busy state. State must be declared, never implied.
Open both cards to continue.
The governance envelope MX adds
Schema.org names a price. It does not say who issued the figure, when, against which booking conditions, or whether the version on the page is the version the publisher actually issued. That is the work the MX layer does, above the discovery layer. View each part of the envelope:
Provenance declares who wrote it, who maintains it (often a different person from the author), the canonical URI where the durable record lives, and a cryptographic signature the agent can verify against the publisher's public key.
Lifecycle declares when the record was created, when it expires, what triggered the last update, and whether a later record has superseded it. Locale conventions declare which decimal grouping is in force, which currency and timezone apply, and which booking or contract terms the figure is attached to.
Agent affordances declare what a machine may do with the record, under what terms, with what attribution. The instructions are carrier-neutral: the same record travels with the file when it leaves the web - a PDF brochure, a confirmation email, a partner data feed - without losing any of the above. Together, this is the signed record an agent verifies before acting.
View all three parts to continue.
Quick check
A product page has perfect Schema.org markup and passes WCAG audits, but carries no provenance record. Which layer is missing?
Answer the question to continue.
What you now know
- The discovery layer and the governance envelope are both required. A page strong at discovery with no envelope is the £203,000 cruise misread waiting to happen (Appendix I of MX: The Protocols).
- Technique 1, structured data: stop making machines guess. Mark up everything that matters with Schema.org, the collaborative vocabulary, so machines parse it with certainty.
- Techniques 2 and 3: WCAG 2.1 AA accessibility is the foundation for agent compatibility, and state must be declared explicitly in markup, never implied visually.
- Above discovery, MX adds the governance envelope: provenance, lifecycle, locale conventions, agent affordances, and the signed record an agent verifies before acting.