MX for Good

Someone navigating a housing crisis, a benefits lapse, or a legal emergency doesn't evaluate sources the way a researcher would. They ask an AI assistant and act on what surfaces. That assistant draws on whatever signals it can read: reviews, news coverage, third-party mentions, structured data declarations. Organizations with large web footprints generate those signals naturally. Community legal aid clinics, mutual aid networks, and public health programs — built on years of referral and relationship, not marketing — often don't. The result is an asymmetry: the organization that surfaces is the one most legible to the system, not necessarily the one with the strongest services or deepest community trust. A clinic with strong services but no machine-readable structure is, to the agent, indistinguishable from one with no services at all.

The disability rights movement — through the 504 sit-ins, the ADA, Section 508, and WCAG — established that equal access to information is a civil right. The technical patterns they won are the foundation of agent legibility — and the starting point for what MX builds on and extends.

That movement also named an asymmetry: systems designed without the full range of users in mind will fail them, and keep failing them until the design process changes. More than 95% of home pages still fail basic WCAG criteria. The gap hasn't closed.

AI-mediated access is opening the same gap. Commercial organizations are already moving to optimize for agent legibility — for reach, for bookings, for conversions. The people most at risk from a world designed around those incentives are the ones who always lose first: non-English speakers navigating English-first services, caregivers asking a voice assistant for help while managing everything else, blind users whose needs shaped the standards that underpin this work, people in a crisis whose best option never makes it into the answer.

MX for Good exists to learn from that asymmetry rather than repeat it — to make the case that designing for the full range of users, from the beginning, is what the discipline is for.

Machine Experience (MX) design is the discipline that works on this problem. It is not a checklist run at the end of a project. It is not a developer concern handed off after design is complete. It is the practice of making design decisions that hold up under both human and machine scrutiny — and documenting them in ways that can be specified, reviewed, and audited.

The cost of agent illegibility is different for different communities. For public-interest organizations — nonprofits, government agencies, advocacy groups, public health departments — the cost is narrative authority: whose services get found when someone asks for help. For commercial businesses — retailers, hospitality, travel — the cost is reach: customers who don't arrive, products that don't surface. The mechanism is the same in both cases. The stakes are what make the practice worth building.

What that gap looks like in practice — for a public-interest organization and the people depending on it — is the subject of the ICIRR case study.

The Machine Legibility Index (MLI) is the diagnostic instrument of MX design — a structured audit across four dimensions: Identity, Reachability, Structure, and Currency. It measures whether a site can be found, parsed, trusted, and acted upon by an AI agent.

The framework is open. The methodology is licensed under CC-BY-SA 4.0; the Chrome extension under MIT. Contributions — new criteria, translations of the methodology, sector-specific extensions, and case studies from any sector — are welcome.