The real Q is which sites SUPPLY ISBN.

Calibre tries to get that from the book when adding.
BTW Not all books HAVE ISBN. That cost $ that new self publishers may not want to shell out. It all depends on the method of sales and their requirements.
(the reason some books have TWO EAN bar codes. One is the ISBN, the other is just the publishers ID and the price . The latter does not need PLU maintenance for ever title, which allows Jobbers to stock the book/magazine racks in big chains. )
(some of) The various plugin's just scrape sites for identifiers as well as other metadata.
+1 on the extract ISBN PI. It tries to find ANY ISBN (it could be for other titles, if that data is in the first few pages

).
You need to vet every find. (

this is not a task I would do in bulk)