Zotero’s existing built-in fields work well for bibliographic metadata, but they are increasingly limiting for modern research workflows.

With LLM-assisted analysis, users can now generate structured metadata at scale across large libraries: topic classifications, method labels, dataset/code indicators, study population, theoretical framework, confidence scores, distilled claims, review status, and other project-specific annotations. These outputs are often too structured for notes, too semantically important for tags, and too variable across domains to justify adding more fixed built-in fields.

A system of custom, user-defined fields would make Zotero much more useful for large-scale literature analysis and curation. Ideally, users could:

- define their own fields and field types
- apply them across selected item types or all items
- view/search/filter/sort by them
- edit them manually or populate them programmatically via plugins/API tools
- optionally include them in exports and reports

This would give users a clean way to store structured, machine-generated or workflow-specific metadata without overloading Extra, notes, or tags.

In the LLM era, the bottleneck is no longer generating useful metadata from papers. It is storing and managing that metadata in a first-class way. Custom fields would let Zotero support these workflows while remaining flexible across disciplines and use cases.