[kiwidude]

@Elmiko - it's a bit random as to which edition it will pick, as there are multiple routes.

Firstly, if your book had no identifiers except an ISBN, then it should match by ISBN. I suspect your "confirmed it is not" check would have been for books where it already had a goodreads: id found.

If you used the "Link to Goodreads" function, then the book chosen will be as a result of their title/author search via their Goodreads API. Now because they are contractually limited as to what books they can expose by that API, its a bit of a lottery whether you get a match or not. Which is why there is the "Search Goodreads" button allowing you to bring up a web page doing their website search (which unlike the API search is not limited in its results) and you can drag/drop the edition you want onto that dialog.

The other way of getting a goodreads id is if you use the Goodreads metadata download plugin. That will be using the goodreads website search like the "Search Goodreads" button again, except automating which result gets picked. Bear in mind that the ISBN can be changed to something different due to the way the funky metadata download logic works, so linking a book using metadata download can often leave you with a different ISBN than what you started with (if you have multiple metadata download plugins enabled).

And just to add to the complication, Goodreads in their website search results have some sort of algorithm to decide which edition "fronts" in the search results representing the book.

The ony way you have guaranteed granular control over exactly which edition is put on a shelf is to use the Goodreads website to drill down to an edition. If you assign that particular goodreads id to your calibre book (either by a first time sync from the shelf, or using the link book dialog and drag/dropping, or even manually typing the goodreads number into the IDs field on Edit Metadata dialog) then from that point on you will be linked to that specific edition for any shelf operations from the plugin.