They told me the use some custom addons to zotero to feed metadata into their database. They don’t simply crawl websites automatically, but semi-manually to get richer and better data.
I don’t know if they import pdfs first and lookup metadata from there or if they browse websites and see if the Zotero indicator shows embedded metadata.
Anyway: if I import one of our galley proof pdfs into Zotero, I get metadata from some indexer (don’t know which) to another one of our articles (it’s always the same, though).
Which means that somebody must already be crawling pubpub, but getting (and multiplying) incorrect bibliographic data. Which is… annoying. OJS uses its internal forms for manually creating and embedding metadata, but I think that workflow could be much improved.
If it were me, I’d try to take care that the embedded metadata would be correctly extracted by Zotero and then correctly imported into modern index dbs like https://vufind.org.