Zotero Collections & Tags Guide

Build a maintainable Zotero library organization system using collections, tags, and saved searches, with BibGenie's Library Organization feature to safely batch-organize your library.

Zotero's collections and tags both help you organize references, but they solve different problems. A maintainable Zotero library rarely relies on just collections or just tags alone. Instead, use collections as project or topic entry points, tags as cross-collection search dimensions, and saved searches for dynamic workflows.

If you use BibGenie, this guide is more than a "how to organize" methodology — it becomes an actionable workflow. BibGenie's Library Organization feature can inventory your existing collections, tags, and items, propose an organization plan based on your research system, then (after your confirmation) batch-create collections, add or remove tags, move items, and rename or merge messy tags.

One-sentence rule

Collections answer "Which project, course, review, or topic entry point do these references belong to?" Tags answer "What characteristics, methods, statuses, objects, or use cases does this reference have?"

BibGenie's role

Zotero provides the basic organizing capabilities — collections and tags. BibGenie's Library Organization feature acts as your library organization assistant: it first understands how you already organize, then proposes a reviewable plan, and finally executes after your approval.

For a complete introduction to BibGenie's library organization capabilities and example prompts, see the Library Management documentation.

Quick Reference

If you're just starting to organize your Zotero, follow these rules:

Scenario	Recommended	How BibGenie helps
Paper, course, project, thesis, dissertation	Collection	Generate a project collection structure from selected items and add references to the matching collection
Research topic, method, data type, region, theoretical framework	Tag	Read titles, abstracts, and metadata, and suggest consistent topic/method/status tags
`to-read`, `read`, `important`, `reviewed`	Colored tag or plain tag	Batch add or remove status tags for your current reading workflow
Recently added (7 days), unread, missing DOI, tag combinations	Saved search	Help design search criteria and follow-up organization rules
Book review and the reviewed book, different versions of the same work	Related items	Identify potentially related entries, though establishing relationships is best left for manual confirmation
Too many nested subcollections	Avoid	Identify overly deep, fragmented, or duplicate collections and suggest merging

What BibGenie's Library Organization Can Do

When manually organizing Zotero collections and tags, the time-consuming part isn't clicking buttons — it's deciding: which tags are duplicates, which items belong in which collection, which metadata is missing, and which operations affect too many items. That's where BibGenie's Library Organization feature shines.

Task	What BibGenie can do
Audit existing structure	List collection hierarchy, tag vocabulary, items under a specific collection or tag
Find items to organize	Locate unfiled items, recently added references, entries missing DOI/abstract, items by tag or collection
Design an organization scheme	Propose a classification system based on titles, abstracts, metadata, existing tags and collections
Batch-organize items	Add or remove tags, add to or remove from collections for specified items
Create collections	Create new collections under the right parent, optionally placing candidate items inside
Clean up tags	Rename tags, merge old tags into existing ones, delete library-wide unused tags
Organize collections	Rename collections, move collections, delete empty or unused collections (deleting a collection does not delete its items)
Check metadata	Read full metadata, identify missing fields, and help correct them when reliable sources are available

BibGenie never bypasses confirmation

Modifying metadata, creating collections, moving items, batch-editing tags, renaming or deleting collections/tags are all write operations. BibGenie will propose actions and wait for your approval. It should never modify your Zotero library without explicit confirmation.

A good BibGenie organization request should include scope, goal, and constraints:

Please organize the currently selected items. First read the existing collections and tags, then propose an organization plan based on titles, abstracts, and metadata.
Prefer reusing existing collections and tags; only suggest creating new ones if there is genuinely no suitable place.
Do not modify Zotero yet. First output which items will be affected, which tags will be added, and which collections will be used.

The principle behind this prompt is the core workflow of BibGenie's Library Organization feature: first discover the current state, then identify items to organize, then propose a plan, and finally execute.

What Are Collections

Collections are the folders and subfolders in Zotero's left sidebar. Zotero's official documentation recommends thinking of collections as "playlists" rather than file system folders: the same item can appear in multiple collections without duplicating the reference. The true complete library remains My Library or the corresponding group library.

Collections work best as entry points. For example:

PhD Thesis
2026 Systematic Review
Course - Research Methods
Project - AI Literature Review
Topic - Retrieval-Augmented Generation

The advantage is clarity, visibility, and project-readiness. Open a collection and you see the core references needed for that work right now.

When using BibGenie, ask it to draft a collection structure aligned with your project goals before moving items:

Please list my existing collections first, then decide whether the selected items can fit into existing collections.
If the existing structure isn't sufficient, design a one-to-two-level collection draft for my "literature review writing project."
Output each collection's purpose and the proposed items it would contain. Do not modify Zotero.

Deleting a collection does not delete items

Deleting a regular collection only removes this organizational entry point, not the items inside it. To truly delete items, you need to move them to the Trash. This distinction is important for new users.

This is also why BibGenie explicitly warns you when deleting a collection: deleting a collection only removes the organizational entry; items remain in the Zotero library. If a collection has subcollections, BibGenie will refuse to delete the parent directly and will ask you to decide how to handle the subcollections first.

What Are Tags

Tags are text labels attached to items, attachments, and notes. They have no hierarchy but can describe the same paper from multiple angles. Zotero's official documentation notes that tags can be used for topic, method, status, rating, or personal workflow.

Tags work best as dimensions. For example:

Topic: llm-agent, scientific-discovery, citation-network
Method: survey, experiment, meta-analysis, case-study
Object: clinical-trial, graduate-students, zotero-users
Status: to-read, read, skimmed, cited
Use: intro-section, related-work, method-baseline

The power of tags lies in combinatorial search. You can filter within a collection for items tagged with both llm-agent and survey, or search your entire library for everything tagged to-read and citation-network.

BibGenie's strength is batch-suggesting tags based on item content. Start by asking it to analyze only:

Please list my existing tags first, then review the titles, abstracts, and existing tags of items in this collection.
Prefer reusing existing tags; only suggest new tags when no suitable one exists.
Finally, present recommended tags to keep, merge, add, and delete in a table. Do not modify Zotero.

Collections vs Tags

Dimension	Collections	Tags
Organization	Hierarchical with subcollections	Flat, no native subtags
Best for expressing	Projects, courses, papers, topic entry points	Topic, method, status, quality, use case attributes
Can an item appear in multiple?	Yes	Yes
Preserved when copying across libraries	Usually not	Usually yes with the item
Browsing	Strong	Moderate
Cross-filtering	Moderate	Strong
Maintenance risk	Collection bloat	Tag synonym and spelling drift
Recommended granularity	Few, stable	Can be finer, but naming must be consistent

Don't make them fully redundant

If you already have a Topic - LLM Agents collection, you don't necessarily need to tag every item in it with llm-agents, unless you want to search for that topic across other collections.

When using BibGenie to organize, add this principle to your prompt:

Please avoid making tags and collections fully redundant. Don't mechanically add the same tag to every item when the collection already expresses the project membership. Only add topic tags when cross-collection search is needed.

Recommended Organization Model

For most researchers, a solid Zotero structure has three layers.

1. A Few Top-Level Collections for Big Directions

Don't create too many top-level collections. Design them as entry points you won't frequently rename:

My Library
  Inbox
  Thesis
  Papers
  Projects
  Topics
  Teaching
  Archive

Or more closely aligned with your research domain:

My Library
  Domain - AI for Science
  Domain - Bibliometrics
  Domain - Human-Computer Interaction
  Project - BibGenie SEO
  Project - Zotero Workflow Study

2. One to Two Levels of Subcollections for Projects

Don't drill down indefinitely. Two levels is usually enough:

Project - Systematic Review
  01 Search Results
  02 Screening
  03 Included Studies
  04 Background Reading
  05 Cited in Manuscript

This structure maps to the research workflow, making it more practical for actual writing than a dozen topic-based subdivisions.

3. Tags as Composable Cross-Cutting Dimensions

Tags shouldn't replace all collections. They're better for recording "what else is searchable about this paper":

method:survey
method:experiment
topic:semantic-search
topic:zotero
status:to-read
status:read
use:background
use:key-citation

Zotero has no native hierarchical tags, but you can simulate namespaces with prefixes like method:, topic:, status:. This reduces ambiguity and makes scanning in the tag selector easier.

Tag Naming Conventions

A good tag system isn't about having more tags; it's about being consistent.

Rule	Recommended	Not Recommended
Use lowercase	`semantic-search`	`Semantic Search`, `semantic Search`
Use hyphens for compound words	`citation-network`	`citation network`, `citation_network`
Prefix categories	`method:survey`	Mixed `survey`, `surveys`
Keep status tags consistent	`status:to-read`	Mixed `unread`, `to read`, `todo`
Avoid overly broad tags	`topic:llm-agent`	`ai`, `paper`, `research`
Avoid overly temporary tags polluting the library	`use:intro-section`	`maybe-use-in-chapter-2-later`

If you already have duplicate tags, rename the old tag to the target name in the tag selector to merge them. For example, rename to read to status:to-read.

If tags are already messy, ask BibGenie to do a tag audit first:

Please list the current library's tags and identify potentially duplicate or synonymous tags — for example, different casing, spaces vs hyphens, singular vs plural.
Give merge suggestions and note how many items each tag affects.
Do not execute any renaming or deletion yet.

Then confirm before executing:

I confirm the merge plan. Please rename the tag `to read` to `status:to-read`.
If `status:to-read` already exists, merge the two tags. Before executing, confirm how many items will be affected.

BibGenie's tag renaming and merging doesn't loop through items one by one; it uses Zotero's library-wide tag operations. When you rename an old tag to an existing tag name, Zotero merges them. This is more reliable and better suited for cleaning up long-accumulated synonymous tags than manually adding new tags and removing old ones item by item.

How to Use Colored Tags

Zotero supports assigning colors to a small number of tags, which can be quickly added or removed with number keys. The current official documentation states a maximum of 9 colored tags.

Colored tags are best for "high-frequency statuses," not for all topics:

Colored tag	Purpose
`status:to-read`	To read
`status:reading`	Currently reading
`status:read`	Read
`use:key-citation`	Key citation
`use:write-next`	For upcoming writing

Avoid assigning colored tag status to ordinary topic tags like machine-learning, history, or statistics. Topics multiply over time, but colored tag slots are scarce.

BibGenie can't decide which tags should permanently occupy your colored tag slots, but it can help identify your most frequently used status-like tags and suggest whether to consolidate them into fewer high-frequency tags.

Should You Keep Automatic Tags

When Zotero saves items from databases, websites, or library catalogs, it sometimes automatically imports subject headings or keywords. These automatic tags can sometimes be useful, but they often create noise: inconsistent casing, messy granularity, overly broad terms, and tag bloat.

Recommended strategies:

New libraries or heavy organizers: Turn off automatic tags and maintain core tags manually.
Medical, legal, library science users relying on controlled vocabularies: Keep automatic tags initially, then clean up periodically.
Users with a large backlog of automatic tags: Hide automatic tags first, then batch-delete after confirming they're useless.

To disable, go to Zotero Settings / Preferences > General and uncheck Automatically tag items with keywords and subject headings.

If you're unsure which automatic tags are valuable, ask BibGenie to group them by frequency and semantics:

Please check the current library's tags and list those that may be from automatic import, overly broad, or inconsistently cased. Do not delete anything, just give cleanup suggestions.

Saved Searches Are an Underrated Organization Tool

Saved searches are like "smart collections." They store search criteria, not fixed results, so they automatically update as your library changes.

Good use cases for saved searches:

Unread: tag is not status:read
Recent Additions: items added in the last 7 days
Missing DOI: items where DOI is empty
Need Review: items tagged status:to-read in a specific collection
Key RAG Papers: items with both topic:rag and use:key-citation

If you find yourself repeatedly using the same filter criteria, consider creating a saved search rather than a new static collection.

BibGenie is better suited to helping you design saved search rules than converting every dynamic task into a collection. For example:

I want to create a dynamic view of "unread RAG papers."
Please review my existing tags and collections first, then suggest Zotero saved search criteria.
If existing tags aren't suitable, explain which tags need to be standardized first.

BibGenie can also use a similar saved-search approach to find items to organize, such as "items added in the last 30 days that haven't been placed in any collection," "journal articles missing DOIs," or "items with old tags but no new tags." These tasks are best handled by first outputting a candidate list, then deciding whether to make batch changes.

Related items shouldn't replace tags or collections. As discussed in Zotero forums, the problem with Related is that it only expresses "two items are related" without clarifying "why they are related." It's therefore best for explicit one-to-one or small-scale relationships:

A book review and the reviewed book
A conference paper and its extended journal version
A dataset paper and the benchmark paper that uses it
A preprint, published version, and corrigendum of the same study
Core source papers repeatedly discussed in a review article

If you simply want to express "these items all belong to the same topic," use tags or collections instead.

Three Common Workflows

Workflow A: Writing a Paper or Review

Create a collection for your writing project, e.g., Project - LLM Agent Review.

Place candidate papers in 01 Search Results or Inbox first.

Use tags to mark topics, methods, and statuses, e.g., topic:tool-use, method:benchmark, status:to-read.

After screening, move or add core papers to 03 Included Studies.

Use use:key-citation, use:intro-section, use:method-section to mark writing purposes.

Suitable BibGenie prompt:

Please examine the `Project - LLM Agent Review` collection.
First list the existing subcollections and tags, then classify the papers into five categories following a review writing workflow: search results, screening, included, background, and cited.
Prefer using existing subcollections; justify any new ones needed.
First output a table, do not move any items yet.

After confirmation:

I confirm this classification table. Please add these items to the corresponding collections per the table, and add the `use:key-citation` tag to core papers.
After completion, summarize how many items were modified, which collections were used, and which tags were added.

Workflow B: Building a Long-Term Research Domain

Create a long-term collection, e.g., Domain - AI for Science.

Create a limited number of subcollections only when truly needed, e.g., Foundation Papers, Benchmarks, Applications.

Use fine-grained tags to record topics, methods, datasets, and experimental subjects.

Periodically check Unfiled Items and add orphaned items to appropriate collections or delete them.

Suitable BibGenie prompt:

Please analyze the items in `Domain - AI for Science`.
First read the existing tags, identify 5 to 8 stable research topics, and suggest corresponding `topic:` tags.
Avoid creating too many subcollections; prefer tags unless a topic will be maintained separately long-term.

Workflow C: Lightweight Users

If you don't want to maintain a complex system, follow these minimal rules:

Collections: Create by project or course.
Tags: Use only 5 to 15 high-frequency statuses or topics.
Colored tags: Only mark to-read, read, important.
Monthly cleanup: Automatic tags, duplicates, and unfiled items.

A lightweight system that's maintained is more reliable than a complex one that isn't.

Lightweight users can still use BibGenie, but the focus should be on reducing maintenance overhead, not generating more tags:

Please classify the currently selected items into only three statuses: `status:to-read`, `status:reading`, `status:read`. If you can't determine the status, put it in `status:to-read` and show a preview first.

Monthly Maintenance Checklist

Perform Zotero library housekeeping monthly:

Open Unfiled Items and process items not in any collection.
Open Duplicate Items and merge duplicates rather than deleting one.
Check the tag selector for synonymous tags, e.g., methods and research-methods.
Hide or delete useless automatic tags.
Check recently added items for clean title, author, year, DOI, and abstract.
Delete temporary subcollections that have lost their purpose to avoid collection bloat.

Use BibGenie for assisted organizing

Ask BibGenie to list candidate items and proposed changes before executing batch tag additions, collection creations, item moves, or metadata checks. For any batch operation, preview first and scope it, e.g., "only process the currently selected items" or "only process this collection."

A solid BibGenie monthly organization prompt:

Please help me with Zotero library housekeeping:
1. First list existing collections and tags.
2. Check for unfiled items, duplicate or synonymous tags, overly deep collections, and recently added unorganized items.
3. When checking items missing DOI or abstract, consider the item type to determine whether a DOI is actually expected.
4. Only output the problem list, impact scope, and suggested actions. Do not modify Zotero.

If you want to place unfiled items in appropriate locations, ask BibGenie to execute a more specific task:

Please find the most recent 50 unfiled items and determine which collections they best fit into based on existing collections.
Do not create new collections unless the existing structure clearly has no suitable place.
First output each item, the suggested collection, the reason, and your confidence level.

Common Mistakes

Mistake 1: Using Collections as Your Only Organization System

If you create a subcollection for every minor topic, you'll end up with dozens or hundreds of collections after a few years. Items get buried in deep trees, making them harder to find.

When using BibGenie, don't ask it to create collections for every small topic. A better instruction: "Only create collections for long-term projects, courses, papers, or stable topics; use tags for fine-grained concepts."

Mistake 2: Tags Without Naming Rules

If to read, ToRead, unread, todo all coexist, search breaks down. Tag systems fear synonym drift most of all.

BibGenie can detect this drift and rename old tags to unified ones. Renaming to an existing tag effectively merges them.

Related is better for explicit relationships, not for linking a batch of topically similar items. Topic membership should use tags or collections.

Mistake 4: Forgetting Subcollection Display Settings

In Zotero, subcollection items don't necessarily show in the parent collection by default. Change this via View -> Show Items from Subcollections. This setting affects how you understand parent-child collection relationships, so it's best to standardize in team workflows.

Mistake 5: Ignoring Automatic Tags for Too Long

Auto-imported tags clutter the tag selector. Unless you explicitly rely on these keywords, periodically hide, delete, or disable automatic tags.

Mistake 6: Treating "Organize My Entire Library" as a Vague Command

"Please organize my entire library" sounds convenient but is usually too broad. BibGenie's Library Organization feature is better suited to starting with specific pain points, such as:

Recently added but unorganized items.
Items in a collection that need further subdivision.
Duplicate or synonymous tags.
Entries missing DOI, abstract, or publication title.
Core papers to mark with use:key-citation for an upcoming writing project.

A better prompt:

Please don't organize the entire library yet. Help me diagnose the 3 most pressing categories of issues: unfiled items, duplicate tags, and missing metadata. Only output the problem scope and suggested next steps.

Mistake 7: Having AI Fill Unverifiable Metadata

BibGenie can check metadata and help correct it when reliable sources are available, but you should never ask AI to write its own DOI, abstract, journal name, author, or page number. For metadata writes, reliable sources should come from Zotero's existing metadata, user-provided information, already-read item content, or BibGenie's verifiable external reference lookups.

A good prompt:

Please check the metadata completeness of the most recent 20 items. For entries missing a DOI or abstract, only suggest corrections when verifiable through reliable sources. Mark unverifiable items as "unconfirmed" and do not generate or guess content.

Recommended Template

You can start directly from this template:

Collections
  Inbox
  Project - [Project Name]
    01 Search Results
    02 To Screen
    03 Included
    04 Background
    05 Cited
  Domain - [Long-term Topic]
    Foundation Papers
    Methods
    Applications
  Archive

Tags
  status:to-read
  status:reading
  status:read
  use:key-citation
  use:intro-section
  use:method-section
  method:survey
  method:experiment
  method:case-study
  topic:[Topic Name]

Safety Workflow for Using BibGenie

BibGenie can manage tags, collections, item membership, and some metadata, but library organization involves modifying Zotero data. Always follow this four-step process:

Ask BibGenie to read the existing structure: collections, tags, and the scope of items to organize.

Require BibGenie to prefer reusing existing collections and tags; only create new ones if genuinely no suitable place exists.

Ask BibGenie to output the organization plan, including tags to add, tags to remove, collections to move items to, which items are affected, and which operations require approval.

After manual confirmation, explicitly authorize BibGenie to execute. After execution, ask for a summary of modifications, skipped items, and incomplete tasks.

Recommended prompt template:

Please organize the currently selected items. The goal is to establish clear Zotero collections and tags:
1. First list existing collections and tags.
2. Prefer reusing existing collections and tags; don't create similar names unnecessarily.
3. Determine which items should go into the same collection.
4. Then suggest consistent `topic:`, `method:`, `status:` tags.
5. Output which tags and collections will be added or removed for each item.
6. Do not modify Zotero yet; wait for my confirmation.