It's one-indexed, counting that actual pages in the PDF. The nominal page numbers (as seen on the header/footer) are often different since you don't typically number title pages and the like, which is called "front matter". [0]
It's probably linking to the 'sequence number' rather than the page number. The cover is the first sequence number but doesn't have a page number. So if there was a few blank pages before the numbered pages started, it would probably be off by that many. I'll bet there are ways to mitigate that— PDF usually seems pretty good about dealing with the discrepancies between print and digital document structures.
