“Accessibility-first screen reader prediction” is not a solution, folks. Parsing a DOM and building your _own_ accTree to then *emulate* how desktop screen readers _might_ announce a page as if *maybe* using read-all just ends up putting more burden on human testers. Never mind missing the bugs, features, heuristics, and more built into a DOM parser and screen reader engine. Unless your goal is satisfy a dashboard while removing humans from your #accessibility testing. #a11y #ScreenReader