AI Agent for Zotero: Paper Retrieval and Citation Checks

Zotero organizes your research library — papers, books, web pages, and other sources — with tags, collections, and notes. An AI assistant that can access your Zotero library can help you retrieve papers, summarize what is in a collection, and check whether a citation exists in your library. The assistant speeds up retrieval; you verify the content.

Tools like fenbs keep the research questions, paper links, and review tasks that come out of a Zotero session, so each session builds on the last rather than starting from the same pile of references.

How AI agents can work with a Zotero library

Zotero exposes its library through a web API that allows authorized applications to search, retrieve, and read items. An AI agent that has been set up with access to this API can search your library by keyword, author, or collection, retrieve metadata and notes, and check whether a specific paper is present.

Setting up that access requires generating a Zotero API key from your Zotero account settings and providing it to the agent tool you are using. The access you grant determines what the agent can read or modify. See the Zotero API documentation↗ (opens in a new tab) for how API keys and permissions work.

For the broader category of AI agents and how they connect to external tools, types of AI agents explains the different architectures. For agents that work with browser-based research, AI browser agents covers a related use case.

What to ask an AI agent to do with Zotero

Retrieval tasks work well: "Find all papers in my library by this author," "list the items in the Methodology collection," or "check whether this paper is already in my library before I add it." These are tasks that save time compared to searching manually in the Zotero interface.

Summarization of abstracts or notes also works, with the usual caveat: the agent summarizes what is in your Zotero notes and the stored abstract, not the full text of the paper unless the PDF is attached and the tool can read it. Always verify a summary against the actual paper for anything you plan to cite.

How to verify citations before you use them

An AI agent can tell you that a paper is in your library and reproduce its metadata. It cannot tell you whether you read it carefully, whether the finding it describes is relevant to your argument, or whether the paper has been retracted since you saved it.

For any citation you plan to use in a document, go back to the source. Check that the finding the AI summarized is in the paper where it says it is, and check the page or section number if your citation style requires it. Retractions happen; check the publisher's page for any paper where the publication date or findings seem unusual.

For agents used in finance research, where citation accuracy has different stakes, AI agents for finance covers a related context.

How to track what you have reviewed and what remains

A research project with many sources needs a reading list that tracks status: not yet read, in progress, read and noted, read and decided not to cite. Without that list, you re-read papers you have already reviewed or miss ones that would have been useful.

Keep the reading list in a project tool, not in Zotero tags alone. Tags are useful for retrieval; tasks in a project are useful for tracking progress. Combine both: use Zotero to find papers, and your project to track the review status.

How to organize your Zotero library for AI queries

A well-organized Zotero library makes agent queries more useful. Group papers into collections by topic rather than by project — organizing by project fragments the same paper across multiple collections. Use tags for reading status: "to review," "reviewed," and "cited in [project name]" give you a fast way to find what still needs attention.

Keep your notes and annotations in Zotero for the papers you cite most often. An AI agent that can read the notes you left on a paper — your summary, the key claim you found useful, a question about the methodology — produces more useful output than one working from the abstract alone.

How to do this with fenbs

Use one fenbs project per research topic. Link the Zotero collection and keep the reading tasks there.

1. Create a free account at fenbs and add a project for the research topic.
2. Add a context note with a link to the relevant Zotero collection and the research question you are trying to answer.
3. Add each paper to review as a task in To Do. Move it to In Progress when you start reading, and Done when you have taken notes.
4. When an agent retrieves a paper or summary, add a link to the source in the task note and verify the content before using it.
5. Record any citation check findings — papers that were not what the abstract suggested, retractions, or version differences — as Lessons in the project.

fenbs does not connect to Zotero directly. It holds the research questions, links, and review tasks that organize your work alongside the Zotero library. For building agents that access Zotero programmatically, AI agent architecture covers how those connections are structured.