An in-page anchor link is an href that starts with a #, meant to jump to a spot on the same page: a table of contents entry, a back-to-a-section link, a footnote. It works by matching the fragment after the hash to an element with that id. When no element has that id, the click does nothing: the page does not move and the reader is left where they were, with no feedback.
Why it happens
These break quietly. A heading is renamed or removed but the link to it is not, an id is changed during a redesign, or the link was written for an id that was never added. Because nothing errors, it survives every test that is not a real click. Two hrefs are not broken and should not be flagged: a bare # and #top both scroll to the top of the page without needing an id.
Why it matters
A jump link that goes nowhere is a small but real dead end in your own navigation, and a table of contents made of them reads as neglected. For a machine reading the structure of your page, an anchor that resolves is a signal of a well formed document, and one that does not is noise.
The fix
Either add the matching id to the element the link should reach, or correct the href to a fragment that exists. Each #fragment must match an element id, or a legacy <a name>, somewhere on the page. Fragments are URL encoded in the href but the id is not, so #my%20section matches id="my section". The Audit lists the anchor links that resolve to nothing. See also anchor text for making the links themselves clear.