AI for Business Owners · Independent decision intelligenceSource-backed reporting · No paid editorial rankings
Owner AI Fieldbook

A practical, source-backed fieldbook for owners deciding where AI belongs in customer service, marketing, finance, operations, people, knowledge, and risk—with tests that fit a smaller team.

Owner briefings

An AI-built website still needs a manual accessibility check

The Justice Department says automated checkers can help but cannot establish that a website is accessible. An owner needs to test the customer path, preserve a way to report barriers, and fix what the builder regenerates.

Answer capsule

The Justice Department says automated checkers can help but cannot establish that a website is accessible. An owner needs to test the customer path, preserve a way to report barriers, and fix what the builder regenerates.

What the source establishes

  • The Justice Department published its Guidance on Web Accessibility and the ADA on March 18, 2022 for state and local governments and businesses open to the public.
  • The guidance lists barriers including poor contrast, missing text alternatives and captions, inaccessible forms, weak heading structure, and mouse-only navigation.
  • DOJ says automated accessibility checkers and overlays can help, but a clean report does not necessarily mean a website is accessible; it recommends pairing automated and manual checks.
  • The page is informal technical assistance, not a final agency action or a source of legally enforceable duties beyond applicable statutes, regulations, and binding precedent.

Map the customer path before pressing publish

The direct owner answer is that an AI website builder can draft pages, layouts, images, forms, and copy, but the business still owns the public experience. List the customer jobs that matter: finding hours and contact details, reading a menu or service description, booking, buying, applying, signing in, watching a video, submitting a form, correcting an error, and reaching a person. Test the complete path rather than judging the home page or the builder's accessibility badge.

Decide what the generated system is allowed to change and who approves publication. A new image can lose useful alternative text; a rewritten heading can break navigation; a color suggestion can weaken contrast; a generated form can omit labels or clear error instructions. Keep current business facts and accessibility requirements outside the prompt so regeneration does not quietly replace an approved customer path.

The accountable team should translate this point into a named workflow, affected population, source data, human owner, approval right, exception path, retained evidence, and review date. That translation is what separates an interesting AI development from a decision that can be governed and evaluated.

Pair automated checks with hands-on use

Run an automated scan because it can find repeatable defects quickly, but do not treat zero reported errors as proof. The DOJ guidance says a clean automated report does not necessarily mean everything is accessible and recommends pairing automated and manual checks. Use the keyboard without a mouse, zoom the page, inspect headings and link names, review image purpose, confirm captions, submit forms incorrectly and correctly, and observe whether focus and errors are understandable.

Include people who use relevant assistive technology when the business can do so responsibly, and use qualified accessibility help for consequential services. Test on the devices, browsers, payment or booking services, plug-ins, and embedded content customers actually encounter. An AI builder may produce technically neat markup while the real task remains confusing, blocked, or impossible.

The accountable team should translate this point into a named workflow, affected population, source data, human owner, approval right, exception path, retained evidence, and review date. That translation is what separates an interesting AI development from a decision that can be governed and evaluated.

Make correction part of daily operations

The DOJ recommends a way for the public to report accessibility problems. Put that route where a customer can find and use it, name who receives the report, and define how the team supplies an accessible alternative while a defect is fixed. Record the affected page, task, device or assistive technology, impact, owner, correction, retest, and date. Do not force the person who found the barrier to become the business's unpaid tester.

Recheck after the AI builder, theme, plug-in, booking service, menu, image library, form, payment flow, or content changes. Assign recurring pages and components an owner, and keep a simple release checklist that includes accessibility beside factual accuracy, price, privacy, security, and customer service. The goal is not a one-time score; it is a customer path the business can maintain and repair.

The accountable team should translate this point into a named workflow, affected population, source data, human owner, approval right, exception path, retained evidence, and review date. That translation is what separates an interesting AI development from a decision that can be governed and evaluated.

Keep the guidance inside its stated boundary

The DOJ page addresses ADA web accessibility for state and local governments and businesses open to the public. It says businesses have flexibility in how they meet general nondiscrimination and effective-communication requirements and points to technical resources, while also noting that its informal guidance is not final agency action and does not create obligations beyond applicable law.

Business type, location, customer, contract, funding, sector, platform, and the services offered online can change the applicable requirements. An automated score, overlay, vendor promise, manual sample, or this briefing cannot determine compliance. Use the guidance to organize an owner-controlled publishing and correction routine, then seek current qualified legal and accessibility advice for the business's actual facts.

The accountable team should translate this point into a named workflow, affected population, source data, human owner, approval right, exception path, retained evidence, and review date. That translation is what separates an interesting AI development from a decision that can be governed and evaluated.

Decision test

Ask whether the source changes the decision itself, the evidence required, the implementation sequence, or only the language used to describe an existing capability. Record which claims are directly supported, which are provider statements, which require an independent test, and which remain unknown. A source-linked review should make uncertainty easier to see, not bury it inside a blended score.

Questions to take into review

  • Which customer need and evidence anchor the content?
  • Are reviews, endorsements, and images authentic and permitted?
  • Which questions have approved answers?
  • How does a customer reach a person?
  • Which price and scope records are current?
  • What changes require owner approval?
  • Which constraints and exceptions matter?
  • What can change automatically?
The publication supports research and executive decision preparation. It does not provide legal, financial, accounting, employment, clinical, cybersecurity, investment, procurement, or implementation advice.