The MX Books.
Three books. One mission: make your content readable by every machine on earth.
Choose your path
MX: The Introduction
Why machines misread the web and what you can do about it. The business case, the technical foundation, and a five-step action framework.
Start here. Read free.
April 2026MX: The Handbook
Step-by-step implementations for every platform. Patterns you can apply this week — from semantic HTML to checkout flows.
Practical. Immediate results.
July 2026MX: The Protocols
The definitive reference for teams implementing at scale. Architecture, standards, the five-stage journey, enterprise patterns.
Go deep. Build it right.
Start with The Introduction. Implement with The Handbook. Scale with The Protocols.
What is MX?
Machine Experience (MX) is the practice of adding metadata and instructions to your content so AI agents do not have to guess. When machines guess, they hallucinate. Explicit structure prevents this.
MX works for every machine — from GPT-4 to a 50M parameter model running locally, from browser agents to voice assistants. No APIs. No code changes. Just structured content that any machine can read.
The hostile internet beyond the web
- Cookie walls, CAPTCHA gates, newsletter modals, dark patterns — every one of these is a brick wall to a machine that cannot click "dismiss"
- PDFs lock structured content inside an opaque binary — machines must reverse-engineer layout, columns, and reading order from coordinates on a page
- Email newsletters arrive as nested-table HTML from 2004 — no semantic structure, no metadata, images as text, tracking pixels as content
- APIs return data but strip context — a JSON price field with no currency, no tax status, no validity date is a hallucination waiting to happen
- Social media embeds are iframes pointing to JavaScript-rendered content — invisible to any agent that cannot execute a full browser stack
- Paywalls serve a teaser paragraph then a login form — the machine sees an article that ends mid-sentence and treats it as the complete text
- CAPTCHAs are specifically designed to stop machines — every CAPTCHA is an explicit declaration that machine users are unwelcome
- Consent banners overlay the entire viewport — a machine without a click handler sees only the banner, never the page beneath it
- Dynamic pricing changes between requests — a machine that fetches a price twice gets two different answers and no way to know which is current
- Infinite scroll loads content on a JavaScript event — a server-side parser sees an empty container where hundreds of products should be
From the books
"Right now, while you're reading this, a machine is probably visiting your website. You won't see it in your analytics. Yet it might be deciding whether your company gets recommended to its user — or your competitor does."
Chapter 1 — Don't Make AI Think"The invisible users don't care about your constraints. They're visiting your site right now, reading it in ways you never intended, and making decisions that affect your business."
Chapter 1 — Don't Make AI Think"Codified content costs almost nothing for a machine to process. It doesn't require reasoning. It just requires reading."
Chapter 2 — How AI Reads"Design for the worst machine, not the best. If a constrained local model running on a £30 device can parse your page, every machine can."
Chapter 1 — Don't Make AI Think"Computational trust compounds in the same way brand trust compounds with humans, but faster and more ruthlessly. An automated agent is unlikely to give you the benefit of the doubt. It may simply route around you."
Chapter 11 — The Business Imperative"The techniques that make content AI-readable aren't tricks or hacks. They're often just good web development practices that we've been ignoring because we could get away with it."
Chapter 1 — Don't Make AI ThinkGet the books
The Introduction is free now. Join the waitlist for early access to The Handbook and The Protocols.