That by itself is a non-trivial task, as bibliography styles vary greatly. There is a fairly good machine learning based tool that does that https://github.com/kermitt2/grobid
It may also be possible to get the list from Crossref directly if the publisher has deposited it there with an article's main metadata (millions of papers apparently have that reference list data deposited, but not all do). Theoretically Zotero could also potentially extract that list when accessing that main metadata.
https://www.wikidata.org/wiki/Wikidata:Zotero/Cita/Documentation#How_can_I_extract_citations_from_a_PDF_attachment?
https://github.com/diegodlh/zotero-cita
You can then copy that list as you wish.
It may also be possible to get the list from Crossref directly if the publisher has deposited it there with an article's main metadata (millions of papers apparently have that reference list data deposited, but not all do). Theoretically Zotero could also potentially extract that list when accessing that main metadata.