Screen readers are the primary assistive technology used by blind and low-vision users to interact with the web. They navigate the page by semantic structure (headings, landmarks, lists, links) and announce content as the user moves through it. The most-used screen readers in 2024: NVDA + JAWS on Windows, VoiceOver on macOS/iOS, TalkBack on Android, Orca on Linux.
Screen reader users do NOT read pages linearly top-to-bottom. They jump by heading (H to next heading), by landmark (D to next landmark), by link (K to next link). Pages with proper semantic HTML (one <h1>, sequential <h2>/<h3>, named <main>/<nav>/<aside>) are navigable; pages built on <div> soup are not.
Test with a real screen reader before shipping any new flow. Lighthouse and axe catch the obvious WCAG failures, but they can't tell you whether a screen reader user can actually complete a checkout.