What does the automated checker actually find?

Missing alt text, images that may be decorative, skipped and empty heading levels, a document with no headings at all, tables with no header row, merged cells that break reading order, low-contrast text, bare URLs used as link text, floating objects that break the reading order, missing language, and weak captions.

Is passing the checker the same as being accessible?

No, and nothing will tell you otherwise here. A checker finds structure that is missing; it cannot tell you that your alt text is wrong, that your heading says “Section 2” instead of what section 2 is about, or that your table is really a layout. Those are judgements.

Why do headings matter more than everything else?

Because they are how someone navigates without sight. A screen reader user jumps heading to heading the way a sighted reader skims — if the headings are bold paragraphs, there is nothing to jump to and the only way through a forty-page report is to read all of it.

Does any of this survive being exported to PDF?

It survives if you tick Tagged PDF, which writes a structure tree naming headings, lists and tables. It is also the reason the checker is worth running first: a tagged PDF faithfully reproduces the structure the document has, including the absence of one.

Accessible Word Documents: a Checklist

Nine things that decide whether a .docx works with a screen reader, each with the fix — and the automated checker that finds most of them in a document you already have.