Because blind users rely entirely on how a screen reader interprets a page’s underlying code, their experience depends almost completely on correct semantic HTML, meaningful alt text, proper form labels, and sensible heading structure, rather than on visual design at all.
Testing specifically with a screen reader, not just running an automated scanner, is the only reliable way to understand how a blind user actually experiences a given page, since many of the barriers blind users encounter (confusing reading order, unlabeled buttons, unannounced dynamic updates) are invisible to a purely visual or code-pattern-based review.