A web font we could not read
The page applies a web font to text whose glyphs this engine could not check, so the extracted text may differ from what a reader sees.
How the text is hidden
A @font-face decides which shape each character paints, and the page can load it from anywhere: a URL, the reader's own installed fonts (local()), or a data: URL in a compressed format. Any of those can carry a rewritten character map, and none of them can be read here — this engine never fetches a resource, never has the reader's machine, and reads TrueType tables but not WOFF, WOFF2 or CFF outlines.
Why a model still reads it
The text extracts exactly as written — every extractor reads what the markup says. The claim is the inverse of an unread image's: not that content went unread, but that the text this engine did report may not be the text on the page.
What we do about it
Every @font-face in the page's own <style> blocks is read, and font-family is tracked through the cascade so the faces actually applied to text are known. A face whose first src — the one a browser would take — is a URL, a local() name, or a data: source that does not read as TrueType is disclosed: one informational finding naming each face, where it comes from, and how many characters are set in it; chars is that total. The verdict is untouched. A face applied to no text is not reported.
How often it fires
Not measured. We publish a rate only where one has been measured against a named corpus of the right format, and no figure exists for this technique yet. Absence means unknown — never low.
Check your own file
Three commands: a key, credit, a verdict.
Start with the API