I'm not sure what version of Zotero you're using, but unfortunately in translation-server which is what we use, this feature seems to be broken, filed here: https://github.com/zotero/translation-server/issues/79

We currently use worldcat's open search API for this metadata and the results we get are only available in marcxml which is probably where that junk is come from and dublincore which is also pretty messy and not too structured. The results on their website come from a database we don't have access to. I could probably improve it a bit on our end but it might worth working towards fixing zotero instead because they query more databases, or merging the results from both somehow.

Change 486851 had a related patch set uploaded (by Mvolz; owner: Mvolz):
[mediawiki/services/citoid@master] [WIP] Remove xISBN and replace with zotero

https://gerrit.wikimedia.org/r/486851

In T214802#4914359, @Mvolz wrote:

...or merging the results from both somehow.

Combining multiple sources would be the best option, but maybe too internet-consuming. Possibly a good idea is also to try some local authorities (per T212585 and subtasks) first.

Change 486851 merged by jenkins-bot:
[mediawiki/services/citoid@master] Remove xISBN and replace with zotero

https://gerrit.wikimedia.org/r/486851

@Dvorapa - Merging the two results sounds like a good idea in theory, but I suspect would be a bad idea in the long run. It would cause the look-up to take longer and potentially lead to more bugs and difficulty tracking down bugs. For example if one source lists a person as an author and another source lists them as an editor (as there sometimes isn't a clear distinction between author and editor), you would end up with the same person being listed as both. I think we should decide on a single provider for ISBN data and make sure our code accommodates it well.