v1.5.0 — Citations column

Citations column

New 7th tool, and the second one that needs no API key. Right-click a column header in the item list → tick Citations. Every item shows how often it has been cited, and sorting the column ranks a collection by impact.

Cell	Meaning
1,725	citation count
0	indexed, but not cited yet
-	no record found, or no confident match
...	lookup in progress
blank	nothing to look up (no DOI and no usable title)

Hover a cell to see the source, the citations per year since publication, and the title that was matched.

• Items are matched by DOI first, then PMID, then an exact title match. A title match is accepted only when the normalised title is identical and the publication year is within one, so a near-miss shows - rather than someone else's number
• Counts come from OpenAlex (cited_by_count), falling back to Crossref (is-referenced-by-count) for DOIs OpenAlex does not have — both free, no API key
• Lookups are batched 50 DOIs per request, run lazily as rows scroll into view, and are cached for 30 days in your Zotero data directory
• New settings: online lookup toggle, refresh and clear-cache buttons

Why not Google Scholar? Scholar has the fullest counts, but it has no API, its terms forbid automated queries, and it serves a CAPTCHA after a handful of them — a column that fires a lookup per row would lock you out of Scholar in your own browser within a single library scan. Expect OpenAlex to read roughly 30–50% lower than Scholar, which also counts theses, preprints and other non-indexed sources: Deep learning (LeCun et al. 2015) is 84,193 on OpenAlex, 75,945 on Crossref and 119,062 on Scholar. The ranking is very nearly the same, and the ranking is what the column is for.

Also fixed

Both the Citations and Journal rank columns previously inherited Zotero's default HTTP behaviour of retrying 5xx errors for up to an hour, which could stall the lookup queue after a single upstream hiccup. They now fail fast and retry on their own one-hour backoff.

Install: Zotero will auto-update, or grab the .xpi below → Tools → Plugins → ⚙ → Install Plugin From File…

v1.4.0 — Journal rank column (FT50 checker)

Journal rank column — FT50 checker

New 6th tool, and the first one that needs no API key. Right-click a column header in the item list → tick Journal rank. Every article gets a colored badge for the journal it appeared in, and sorting the column groups your library by venue tier:

• 🟢 FT50 — one of the Financial Times 50 journals (2016 revision)
• 🟡 A — rated A* or A in the ABDC Journal Quality List 2022 (852 journals), or on your own extra list
• 🟠 IF > 10 — impact factor above the threshold (default 10), from OpenAlex's two-year mean citedness
• 🔴 Other — a journal on none of the lists
• ⚪ ? — impact factor not looked up yet (offline, or lookups disabled)

Items without a journal (books, theses, reports) stay blank. Hover a badge to see why it was graded that way.

• FT50 and ABDC are matched by ISSN first, then normalised journal name, from lists bundled with the plugin (scripts/build-journal-lists.py regenerates them from the ABDC CSV)
• Impact factors are looked up lazily on api.openalex.org (only the journal name / ISSN is sent), cached for 90 days in your Zotero data directory, and can be switched off
• New settings: online lookup toggle, impact-factor threshold, extra A journals (one name or ISSN per line), re-grade and clear-cache buttons

Install: Zotero will auto-update, or grab the .xpi below → Tools → Plugins → ⚙ → Install Plugin From File…

v1.3.0 — Ask a question (chat with a paper)

Ask a question — NotebookLM-style chat with a paper

New 5th tool. Select a paper → right-click Ask a question (AI) (or press Cmd/Ctrl+Shift+A) → a chat panel opens, scoped to that paper. Ask follow-up questions in a multi-turn conversation; answers are grounded in the paper's extracted full text (with abstract/metadata fallback), using your existing Claude or GPT key.

• Grounded, multi-turn Q&A about a single paper
• Save to note — store the conversation as a child note (tagged ai-chat)
• New chatModel setting (default Claude Sonnet 4.6) + optional custom system prompt
• Shortcut: Cmd/Ctrl+Shift+A

Also includes a robustness fix so a failed or empty turn can't desync the conversation history.

Install: Zotero will auto-update, or grab the .xpi below → Tools → Plugins → ⚙ → Install Plugin From File…

v1.2.1

Updated sorter prompt.

v1.2.0

First release of Zotero AI Toolkit — the fusion of Zotero Expand and the AI Paper Summarizer into one plugin, plus new features.

Features

• Summarize papers — structured, literature-review-ready child notes.
• Categorize highlights — recolor highlights into a fixed, consistent taxonomy.
• Find further reading — read a paper + its bibliography, web-search, and save recommended sources.
• Suggest folder — read a paper, look at your folders, and recommend (in a popup) where to file it: an existing folder or a new one to create, then file/move it for you.

Models

• Each tool picks its own model from one combined, up-to-date Claude + GPT list; the provider and API key are inferred from the chosen model.
• Anthropic: claude-opus-4-8, claude-sonnet-4-6, claude-haiku-4-5
• OpenAI: gpt-5.5, gpt-5.4, gpt-5.4-mini, gpt-5.4-nano
• Store one or both API keys; sensible per-tool defaults work out of the box with a single Anthropic key.

Install

Download zotero-ai-toolkit.xpi below, then in Zotero: Tools → Plugins → ⚙︎ → Install Plugin From File…

Compatible with Zotero 7–9.