GAO 27-107934: Can Agencies Prove Accessibility Was Tested Before Release?
What GAO found
The U.S. Government Accountability Office examined how federal agencies test information and communication technology (ICT) for accessibility and how they respond when people report barriers. Among the three agencies GAO examined in depth, the Department of Homeland Security generally followed leading practices for testing before deployment. The Department of Agriculture (USDA) and the Department of Housing and Urban Development (HUD) partially followed those practices. GAO cited outdated testing guidance at USDA and paused implementation efforts at HUD that affected work such as verifying test results. GAO-27-107934, November 19, 2026
In a separate test, GAO submitted feedback to 23 agencies about website accessibility issues it had identified. Fourteen agencies fixed the issues GAO reported. Nine had yet to fix those issues as of May 2026. GAO recommended that USDA and HUD address their respective gaps in pre-deployment testing. The study examined selected agencies and issues. GAO findings and recommendations
The Organizational Information Continuity challenge
Can a release team retrieve the test evidence for the version users will receive? Consider an agency that updates an online benefits form. The earlier version passed a keyboard navigation test. The update adds a new button for uploading documents. At release review, the team needs to know whether the keyboard test covered the new control and whether a user could complete the revised task. A test report kept apart from the release record leaves that question open, especially after a contractor changes or an application moves to a new platform.
An AI assistant could retrieve the earlier passing report and cite it as evidence that the current form passed. The report may be accurate for the version it covered. Without a reliable relationship between the report, the tested version, and the affected user task, the assistant’s claim about the release cannot be verified.
Can the next team tell whether a reported barrier was removed? A feedback acknowledgment establishes that the agency received a report. A deployment log can show that a change went live. The evidence needed to confirm a fix is a retest of the affected task against the corrected version. If the feedback system, issue tracker, and test records lose their connection, a team taking over the service may be unable to tell whether a ticket was closed after a verified repair or simply moved to another queue.
An AI summary could count a closed ticket or a deployed change as a resolved barrier and overstate the number of problems the agency fixed. The response should distinguish a report received, a correction implemented, and a fix verified. These are prospective risks identified by GovKM.
GovKM’s proposed approach
Section 508 of the Rehabilitation Act, 29 U.S.C. § 794d generally requires federal agencies to provide comparable access to covered ICT for people with disabilities when they develop, procure, maintain, or use it, subject to applicable statutory limits. The U.S. Access Board’s Revised 508 Standards set the applicable technical criteria and scope. OMB Memorandum M-24-08 advises comprehensive testing before deployment and calls for agencies to track, review, and address public accessibility feedback. GovKM proposes three ways to make that work traceable.
- Define what the team must test. The product owner and Section 508 program staff should record the product, planned release, affected user tasks, applicable criteria, and any exclusions or exceptions with their basis. For a changed form, include the path from opening it to submitting it, including the controls and error messages introduced by the change. Use the applicable provisions of the Revised 508 Standards to define the scope.
- Link pre-deployment evidence to the release. Have testers record the version, tested tasks, methods, assistive technology where applicable, results, and known limitations. Use manual checks alongside automated testing as described in OMB Memorandum M-24-08. Section508.gov’s test-report guidance provides a starting point for documenting test scope and results. Link the report to the release decision. Make failed and untested criteria visible to the officials responsible for determining what may proceed under applicable rules.
- Track feedback through verified remediation. Link a reported barrier to the affected task and version, the assigned owner, the corrective change, and the retest result. Give the retest an explicit status so reviewers can distinguish an acknowledgment or deployed change from a verified fix. Keep reporter information in the appropriate protected system and expose only the operational details authorized for reporting. OMB Memorandum M-24-08 addresses feedback mechanisms, monitoring, and remediation.
The release record should remain usable when systems and staff change. OrgAI’s intended role is to help maintain and retrieve the relationships among applicable criteria, release versions, test evidence, and remediation decisions. That role is in development. Agency officials remain responsible for determining conformance, applying exceptions, and authorizing releases.
Verify that the evidence follows the release
Select one recently changed service and one reported barrier. Ask a reviewer who did not perform the work to locate the release-specific tests, identify which tasks remain untested, and reconstruct whether the barrier was retested after the fix. Repeat the check after a handoff or migration.
For AI, test with an earlier passing report, a current release record, and a closed ticket without retest evidence. Verify that the answer identifies the tested version and leaves remediation unconfirmed when the supporting result is missing.
The same method can help a bank or retailer validate an online service after a release. The organization should establish its applicable legal, contractual, and policy requirements. The transferable practice is to keep accessibility results tied to the version and tasks they cover.
Contact GovKM to discuss an accessibility evidence assessment or a pilot covering one service from release testing through verified remediation.



