The business case for web accessibility

Web accessibility is the practice of building websites so that people can perceive, navigate, and use them regardless of how they see, hear, move, or process information. In engineering terms, the interface works regardless of input method (keyboard, voice, switch device) or output method (screen reader, magnification). The measuring stick is the Web Content Accessibility Guidelines (WCAG), a W3C standard written as testable success criteria — contrast ratios, keyboard operation, labels, structure — rather than design opinions.

The population is larger than most owners assume. The World Health Organization counts 1.3 billion people — 16% of the global population, one person in six — living with a significant disability. In the United States, CDC survey data puts it at more than 1 in 4 adults, over 70 million people, with cognitive disabilities (13.9% of adults) and mobility disabilities (12.2%) the most common. Most of these people do not use a screen reader. They read with low vision, tab through forms because a mouse is hard to steady, or rely on clear headings to keep their place.

Even those figures understate the audience, because ability is contextual. Microsoft's Inclusive Design toolkit maps each permanent disability to temporary and situational versions of the same mismatch: a person with one arm, a person with a wrist injury, and a parent holding an infant all operate a phone one-handed. Low-contrast text defeats a user with low vision and a contractor reading a screen in full sun. Accessibility is tolerance engineering: the discipline of making an interface hold up outside ideal conditions.

Enough to show up in revenue. The Return on Disability Group's 2024 annual report estimates that people with disabilities, together with their friends and family, control $13 trillion in annual disposable income worldwide; U.S. consumers with disabilities alone control $1.3 trillion. No business writes off a market that size on purpose. An unlabeled checkout form does it by accident.

The loss is hard to see because these customers leave in silence. The UK's Click-Away Pound survey (2019, the most recent edition) found that 69% of disabled online shoppers click away from websites they find difficult to use, abandoning purchases worth £17.1 billion a year in the UK alone. Only 8% ever tell the site owner why, so the loss never surfaces in support tickets.

Company-level results point the same direction. Accenture's 2023 study The Disability Inclusion Imperative, run with Disability:IN and the American Association of People with Disabilities across 346 companies, found that firms leading on disability inclusion earned 1.6x more revenue, 2.6x more net income, and 2x more economic profit than their peers, and were 25% more likely to outperform on productivity. That is correlation across committed organizations, not proof that alt text raises net income; the direction still matches the abandonment data above.

Plaintiffs filed more than 5,000 digital accessibility lawsuits in the United States in 2025, per UsableNet's year-end report — over 3,100 in federal court plus nearly 2,000 in New York and California state courts, up roughly 20% from 2024's ~4,000 filings. U.S. courts routinely treat business websites as places of public accommodation under Title III of the Americans with Disabilities Act; no federal statute yet names a technical standard for private-sector sites, so courts and settlements use WCAG as the de facto benchmark.

Two patterns matter for planning. First, e-commerce drew roughly 70% of all suits — screens where money changes hands attract complaints. Second, defendants repeat: 1,427 of 2025's suits targeted companies that had been sued over accessibility before, and 46% of federal cases involved repeat defendants. Settling a case and shipping the same templates invites the next one.

The widget shortcut fails in the same data. Accessibility overlays — third-party JavaScript that adds contrast toggles and text resizing in the browser — did not reduce risk. In 2024, 25% of all digital accessibility lawsuits (1,023 cases) cited the overlay itself as a barrier rather than a fix, and through 2025 plaintiffs filed 95 to 155 suits per month against companies already running one. An overlay repaints presentation in the browser; assistive technology, courts, and crawlers evaluate the underlying markup, which the overlay leaves untouched.

Exposure also scales with the sensitivity of the transaction. A broken retail checkout costs a sale; an inaccessible patient portal or legal intake form blocks a task the user cannot complete another way. That is why accessibility review sits early in serious website work for healthcare practices and law firms, whose professional obligations make the failure harder to defend.

The U.S. Department of Justice has ended the ambiguity for the public sector. Its final rule of April 24, 2024 adopted WCAG 2.1 Level AA as the binding technical standard for state and local government web content under ADA Title II, and a 2026 extension set the compliance deadlines: April 26, 2027 for entities serving populations of 50,000 or more, April 26, 2028 for smaller ones. For private businesses the rule is a signal: the federal government has named the specific, testable standard it considers adequate — the same one courts already reference.

The European Accessibility Act reaches further. Enforceable since June 28, 2025, the EAA covers e-commerce, banking, and digital services sold to EU consumers regardless of where the seller is headquartered; conformance maps to WCAG through the EN 301 549 standard. Member states set their own penalties: fines exceed €100,000 per violation in Germany and reach €250,000 in France. A California retailer with EU customers is in scope; being a U.S. company is not a defense.

In the same six places. The WebAIM Million, an annual automated audit of the world's top 1,000,000 home pages, found detectable WCAG 2 conformance failures on 94.8% of them in 2025 — an average of 51 errors per page. Automated tooling can detect only a subset of WCAG criteria, so the true failure rate is higher. The most common single failure is low-contrast text, present on 79.1% of home pages at an average of 29.6 instances each.

Six error types account for 96% of the 50.96 million errors WebAIM detected. Each is an ordinary engineering defect with a known fix, and each fix pays a second dividend:

One more WebAIM finding cuts against instinct. Home pages using ARIA — attributes designed to add accessibility information to markup — averaged 57 detected errors, more than twice the 27 on pages without ARIA. The likelier read: ARIA usage correlates with markup that was already in trouble — teams reach for it to patch structure that was never semantic, and the patch does not erase the underlying defects. Native HTML elements — button, nav, label — carry their roles built in. Accessibility failures are code-quality failures, and the fix list reads like a code-review checklist.

The inputs are shared. Alt text, heading hierarchy, semantic HTML, descriptive link text, and transcripts serve a screen reader and a search crawler through the same channel: machine-readable structure. The W3C's business case for digital accessibility documents the cleanest example — when the radio show This American Life added transcripts to its archive, search traffic rose 6.86%, and 7.23% of all visitors used the transcripts.

The overlap extends to performance. WCAG requires content that reflows at 200% zoom, interfaces that behave predictably, and layouts that hold still — pressures that push toward the same lean, stable pages that page-experience signals favor. We laid out that side of the argument in the business case for website speed. A slow page and an inaccessible page tend to share a root cause: markup and scripts nobody audited.

AI adds a third consumer of the same structure. LLM crawlers ingest alt text and heading hierarchy as plain text, and AI browsing agents operate pages through the accessibility tree — the same programmatic representation of roles and names that a screen reader uses. A button with no accessible name is a button an AI agent cannot press; a form without labels is a form it cannot fill. Work done for WCAG conformance doubles as machine readability across search and AI surfaces.

WCAG defines three conformance levels. Level A covers the most severe barriers; Level AAA is aspirational for most sites; Level AA — the middle tier — is the standard referenced by the laws above, including the DOJ's Title II rule and the EAA's EN 301 549. AA is a finite list of testable criteria, not a design philosophy. The core of it:

• Contrast: normal-size text at a ratio of at least 4.5:1 against its background, large text at 3:1 (SC 1.4.3) — a computable ratio that tools verify.
• Keyboard: all functionality operable by keyboard alone; a control you can click but cannot reach with Tab and activate with Enter fails (SC 2.1.1).
• Focus: a visible indicator showing keyboard users where they are on the page.
• Alternatives and labels: text alternatives for images, and a programmatically attached label on every form field.
• Structure: a logical heading hierarchy and a declared document language.
• Reflow: content that remains usable at 200% zoom without loss of function.

Cost depends on when you pay it. Specified at build time, AA conformance is mostly the absence of shortcuts — semantic elements instead of styled divs, labels written with the form, contrast checked when the palette is chosen. Retrofitting an existing site costs more, because by then the defects are load-bearing — audit, re-templating, and regression testing each bill separately. In both cases, automated scanners catch the six failure types in the table above but only part of the full standard; a real conformance pass adds manual keyboard and screen-reader testing.

Sequence it like any other quality gate: fix the six table rows first — they are cheap, automatable, and cover 96% of detected errors — then schedule the manual pass. This is how we build at Machina: WCAG AA lives inside web development as a property of clean markup, not a separate line item added at the end.