What should an agency put in a client contract about accessibility responsibility?

Put four things in the contract before work starts: who is responsible for accessibility, what is in scope, who maintains it after launch, and what evidence counts as proof of work. Agencies that leave these vague discover their contract is the problem the moment a demand letter arrives.

State who owns legal responsibility

The site owner, the brand, carries the legal risk under ADA Title III. The contract should say this plainly. An agency that implies it will make the store "ADA compliant" without defining what that means has sold a promise it cannot keep. Better language: the agency will perform defined accessibility work to a defined standard, and the client retains responsibility for the store's legal standing. This protects both sides. The client knows the agency is a vendor, not an insurer. The agency is not on the hook for a lawsuit over a site it does not own.

Define the scope in writing, not in a call

Verbal promises about accessibility scope are worth nothing after a complaint. The contract should name the storefronts, templates, languages, and customer journeys covered: homepage, product pages, cart, checkout, account pages, search, and any content in languages other than English. It should also say what is excluded. Checkout on Shopify, for example, is largely controlled by Shopify itself. Third-party apps render their own markup. State these limits explicitly so the client understands the agency is not warranting the entire internet, just the work described.

Name the standard, and the testing method

A contract that says "accessible" without naming a standard is a disagreement waiting to happen. Reference WCAG 2.2 and a conformance level, usually AA, as the target for the work described. Then say how testing happens: manual keyboard testing, screen reader passes, contrast checks, and which pages get tested. Automated scans alone are not enough, and the contract should say so. If the agency will produce a report, describe what it contains: issue list, severity, WCAG mapping, and the test date. The report is the deliverable the client can show a lawyer, so define it like one.

Assign maintenance after launch

Most accessibility regressions happen after launch: theme updates, app installs, new content, promotional popups. If the contract ends at launch, the store decays the day after. Offer a maintenance term: who re-tests, on what cadence, and what triggers a re-test. An app install by the client's marketing team is the most common regression source, and the contract should say whether testing after client-side changes is included or billed separately. Without this clause, the agency either does unpaid re-testing or watches its launch work rot. Neither is good.

Keep a dated record of everything

Contracts win arguments, but records win demand letters. When a plaintiff's lawyer asks what testing was done, "we tested it in May" is weak and "here is the dated report with the test pages, methods, and findings" is strong. The contract should require the agency to deliver dated test records with each milestone and store them where the client can find them later. Note the app versions installed at test time, since app updates are a known regression vector. A year-old audit with no follow-up testing still shows diligence, but an audit followed by scheduled monitoring shows a program. Clients pay for the program once they see the difference.

Review the clause with a lawyer before you need it

This is guidance for the business side of accessibility work, not legal advice. Before adding these clauses, have an attorney review them against the jurisdictions your clients sell into. The cost of one review is a fraction of the cost of one dispute over who promised what. Agencies that treat the contract as part of the accessibility deliverable, not boilerplate around it, end up with fewer surprises and better client relationships.