BibGenie vs PapersGPT: Which Zotero AI Plugin Should You Choose?
Compare BibGenie and PapersGPT as Zotero AI plugins. Discover how BibGenie delivers local-first privacy, custom multi-model freedom, passage citation highlights, and 5-source metadata review alongside PapersGPT's batch AutoPilot.
Both BibGenie and PapersGPT integrate into Zotero, support offline local model execution without sending data to third parties, and provide a built-in Zotero MCP server for external AI workflows. But each is tailored to a distinct research scenario: BibGenie is engineered as a local-first daily research copilot — offering zero-config hosted AI alongside custom multi-model freedom, highlighted passage citations, multimodal screenshot Q&A, and 5-source metadata review — while PapersGPT is built for bulk reading at scale with AutoPilot batch summaries.
Quick Verdict
- Choose BibGenie if you want a local-first Zotero AI assistant with custom model freedom and checkable citations. Chat history and indexes stay on your machine. Enjoy 200 recurring monthly hosted credits, connect any custom API (DeepSeek, Claude, OpenAI) or local model (Ollama / LM Studio) with zero platform fees, and review metadata with visual diff cards.
- Choose PapersGPT if you do high-volume literature reviews and want automated batch reading. It is open source (AGPL-3.0 for the plugin), indexes full PDF text across your library, includes AutoPilot to batch-read papers into structured Zotero notes, and allows downloading and running supported local models offline.
Comparison at a glance
| Dimension | BibGenie | PapersGPT |
|---|---|---|
| Citation traceability | Click citations to jump back to sources; highlights PDF text blocks when valid coordinates exist (supports EPUBs, snapshots, and PDF screenshot Q&A) | Click citations to jump to cited sources; Deep Research drafting locates PDF passages, while general chat links back to the cited paper |
| Price (plugin + model) | Free 200 credits/month recurring; Pro $9.99/mo (10,000 credits); Lifetime $199 early-bird one-time (retiring after v1.0 per pricing page notice) | Plugin free; one-time license with initial credits: Basic $39 (500 credits), Premium $69 (700 credits), Business $99 (1,000 credits); or monthly subscriptions ($5/$7/$9) |
| Free tier | 200 hosted credits/month recurring, no API key needed; BYOK, custom endpoints & local models incur zero platform fees | 20 trial single-PDF chats; 20 signup credits via email; unlimited local model use; 30 cloud sessions with own API key |
| Models & custom endpoints | Hosted models, BYOK (OpenAI, Anthropic, Gemini, DeepSeek, OpenRouter, any OpenAI-compatible API), local models (Ollama / LM Studio) | 30+ cloud models, supported local models downloaded in-plugin (Gemma 4, Qwen 3.5, Mistral 7b, Llama 3.1, Phi-4, DeepSeek Distill), Ollama, custom API |
| Hosted model route | Built-in hosted models — no key, subscription, or local setup needed (Free: 200 credits/month) | Built-in credits — works out of the box with no API key; top-ups available after initial credits are used (1 credit = 8,000 tokens) |
| Whole-library search | Instant automatic semantic index over titles, abstracts, and metadata without setup or indexing delays | Indexes full-text PDFs (vendor benchmark: 10,000 PDFs in 7 min on i9 MacBook Pro); library Q&A caps at 500 (Basic), 1,000 (Premium), 2,000 (Business) |
| Multi-PDF / collection chat | Chat across papers via semantic index; collection-level queries via chat | "Chat Multiple PDFs" — select papers or collections (free trial caps at 3, Basic caps at 20, Premium at 100, Business at 300) |
| Batch reading & agent | Grounded chat + multimodal screenshot Q&A + diff-previewed library edits + OpenAlex discovery; built-in MCP server | AutoPilot: batch-reads collections into Zotero notes (pricing table lists Basic 20 papers/run; recharge notes list AutoPilot under Premium/Business); built-in MCP |
| Literature discovery | Built-in OpenAlex discovery UI to explore, search, and import papers outside your library | Literature Radar: monitors topics across arXiv, PubMed, Semantic Scholar, Europe PMC, Crossref, bioRxiv, and medRxiv with scheduled digests |
| Privacy & architecture | Local-first architecture: chat history & vector index stay strictly on your machine; BYOK connects direct; zero prompt logging | Local models run offline after setup; credit chats use PapersGPT proxy; BYOK connects direct to model provider; web search uses external APIs |
| Library organization | Built-in library management; every change is previewed as a diff card before execution | Writes structured notes back to Zotero via AutoPilot; no dedicated duplicate/tag organization workflow claimed |
| Metadata review | Dedicated workflow cross-checking fields against arXiv, Semantic Scholar, OpenAlex, DBLP, and OpenReview with diff preview | No dedicated metadata-review workflow found on official site |
| Zotero compatibility | Zotero 7–10 | Zotero 7–10 (README); 11 beta per Atlas listing |
| Maintenance | v0.9.6 (Nov 10, 2026) | v2.3.0 (Nov 13, 2026); active maintenance, 6 contributors |
| Ease of setup | Install .xpi and go; hosted models work in seconds without downloading gigabytes of local weights | Install .xpi; local models download in-plugin (requires initial connection); cloud credits require selecting a plan |
| Openness / auditability | Closed source; narrower trust surface and no unpinned package execution | AGPL-3.0 (plugin frontend; engine closed source); v2.2.0 scan unrated on Atlas as 202 MB installer exceeds 60 MB limit (checked Nov 2026) |
Citation traceability: how both plugins locate source text
This is one of the most critical factors in academic research: when an AI makes a claim, how quickly can you verify it against the source?
BibGenie is architected around verifiable citations: clicking a citation marker jumps to the source passage and highlights the text block on the PDF page when valid text coordinates can be resolved. It also supports EPUBs, web snapshots, and native annotations. Furthermore, BibGenie features multimodal PDF reader screenshot queries, allowing you to box-select complex architecture diagrams, benchmark charts, or mathematical equations directly from the PDF reader to inspect visual evidence alongside textual claims. BibGenie citation tracing documentation
PapersGPT also includes clickable citations. In its Deep Research drafting workflow, its writing guide specifies that clicking a citation can locate the specific passage in the Zotero PDF reader, while general chat citations link directly back to the cited paper.
Both plugins provide traceable citations into your library, with BibGenie offering direct visual text-block highlighting and integrated screenshot analysis for figures and tables.
A note on verification
A valid citation does not guarantee an accurate interpretation. A model can cite the correct paragraph yet summarize it with subtle inaccuracies. Traceability accelerates verification; it does not replace critical human reading.
Pricing and allowances: recurring hosted credits vs lifetime licenses
The conventional assumption that "PapersGPT is completely free while BibGenie requires payment" oversimplifies how both products handle compute costs:
- PapersGPT: The plugin is free to install, and local model inference is free and unlimited. For cloud models, the free tier offers 20 trial single-PDF chats (and 20 trial credits upon email signup per its quickstart guide), alongside 30 cloud sessions with your own API key. For ongoing hosted cloud use, PapersGPT sells one-time license purchases with initial hosted credits: Basic $39 (500 credits), Premium $69 (700 credits), and Business $99 (1,000 credits), with additional credit top-ups available (1 credit = 8,000 tokens). Monthly subscriptions are also available ($5, $7, and $9/month).
- BibGenie: BibGenie provides a generous recurring free tier: every user receives 200 hosted credits each month without entering an API key or credit card. For heavy research, Pro is $9.99/month for 10,000 credits, and an early-bird Lifetime Pro plan ($199 one-time) is currently available and will retire after the v1.0 release per the pricing page notice (existing lifetime buyers keep their access permanently). Crucially, bringing your own API key (BYOK), custom endpoints, or local models incurs zero platform fees and consumes no platform credits, making it fully sustainable. BibGenie pricing details
Summary: Both plugins allow you to manage model costs with your own keys or local models. BibGenie stands out by offering a recurring monthly hosted allowance that works out of the box with zero configuration and zero-fee BYOK, while PapersGPT provides one-time license packages with initial credit bundles.
Models and local deployment: custom multi-model freedom vs in-plugin downloads
Both plugins support cloud models, bring-your-own API keys, and local models that keep data strictly on your machine.
- PapersGPT can download and run supported local models (including Gemma 4, Qwen 3.5, Mistral 7b, Llama 3.1, Phi-4, and DeepSeek Distill) from within the plugin. Initial setup requires an internet connection to download dependencies and multi-gigabyte weights, after which inference runs offline without needing an external Ollama install (though Ollama and custom endpoints are also supported). PapersGPT on GitHub
- BibGenie provides unrestricted custom multi-model freedom alongside built-in hosted models. Researchers can connect any OpenAI-compatible endpoint (DeepSeek, Claude, Gemini, SiliconFlow, OpenRouter) or local engines (Ollama, LM Studio) with zero platform markups and zero token fees, while uniquely configuring chat and embedding models independently. BibGenie custom and local models documentation
Summary: PapersGPT enables in-plugin downloading and offline execution of supported local models; BibGenie combines zero-setup cloud hosting with open custom model connectivity and decoupled local endpoints.
Reading at scale: whole-library indexing and AutoPilot
This is PapersGPT's standout architectural strength:
- Whole-library PDF indexing: PapersGPT indexes the full text of all PDFs in your Zotero library (vendor benchmark reports 10,000 PDFs / 42 GB indexed in 7 min on an i9 MacBook Pro). Whole-library Q&A caps at 500 papers (Basic), 1,000 papers (Premium), and 2,000 papers (Business).
- AutoPilot batch synthesis: Researchers can set questions and run AutoPilot across collections to extract hypotheses, methods, and findings into structured Zotero notes. (Its pricing table lists 20 papers/run for Basic, whereas recharge notes list AutoPilot under Premium / Business).
- Literature Radar: Users can subscribe to academic topics; the plugin monitors feeds across arXiv, PubMed, Semantic Scholar, Europe PMC, Crossref, bioRxiv, and medRxiv, delivering scheduled digests inside Zotero.
BibGenie approaches library retrieval differently: its semantic index automatically covers titles, abstracts, and core metadata with zero configuration and zero indexing lag, making it lightweight and immediately accessible for meaning-based search across your collection. While PapersGPT excels at automated bulk skimming, BibGenie focuses on the deep, interactive synthesis loop where researchers spend most of their analytical time.
Summary: For bulk literature reviews, cross-paper batch summarization, and automated feed tracking, PapersGPT offers a dedicated high-volume toolset.
Library management and metadata review: structured diff cards
Where PapersGPT focuses on reading volume, BibGenie focuses on library integrity, precision, and research hygiene:
- Preview-before-write diff cards: Modifying library fields, tags, or collections in BibGenie always produces an interactive visual diff card, allowing you to inspect proposed changes before anything is written to your database, eliminating accidental overwrites. BibGenie library management documentation
- Dedicated 5-source metadata review: BibGenie cross-checks bibliographic fields against arXiv, Semantic Scholar, OpenAlex, DBLP, and OpenReview to detect missing publication years, incomplete DOIs, or incorrect journal titles, applying verified updates with explicit user approval.
- External literature discovery: BibGenie includes an integrated OpenAlex interface to search, discover, and import related research directly into your collection without switching to external web browsers.
PapersGPT writes structured summary notes back to Zotero following AutoPilot or multi-PDF synthesis sessions, but it does not offer a dedicated metadata auditing workflow or field-by-field diff previews for library cleanup.
Summary: BibGenie provides dedicated, safe metadata review across 5 academic databases and diff-previewed library management; PapersGPT focuses on generating synthesis notes from batch reading.
Privacy, data flow, and Local-First architecture: research data sovereignty
Both plugins prioritize privacy, with BibGenie embracing a foundational Local-First architecture designed for pre-publication and confidential research:
- Local-First data storage: In BibGenie, your complete chat history, reading context, and semantic index reside strictly inside your local Zotero profile database. When paired with local models (Ollama / LM Studio) or your private API keys, your paper contents and queries never touch BibGenie servers, ensuring complete data sovereignty for unpublished manuscripts and proprietary IP.
- Cloud routes: Using built-in credits sends prompts to each vendor's respective proxy service; bringing your own API key connects Zotero directly to the chosen provider (OpenAI, Anthropic, etc.).
- Independent endpoints in BibGenie: Note that when using a local chat model in BibGenie, embedding models and external tools (such as web search or OpenAlex) continue to follow their own independent endpoint settings. To ensure total offline isolation, configure both models locally and disable online discovery tools.
- Zotero MCP server: Both plugins include a built-in Zotero Model Context Protocol (MCP) server, allowing external developer environments (such as Claude Desktop or Cursor) to securely interface with your Zotero library.
Maintenance, open source, and security audit history
Understanding each project's development model helps evaluate long-term maintenance:
- PapersGPT: The plugin frontend is open source under AGPL-3.0 (with a proprietary core engine) and has earned over 2,700 GitHub stars and 200,000+ downloads. The current version is v2.3.0 (November 13, 2026).
- Atlas rating note: As of November 2026, the 202 MB installer scan on Zotero Plugins Atlas (recorded for v2.2.0) is unrated because it exceeds the platform's 60 MB file analysis limit. (A C-rating visible online belongs to an unrelated legacy sidebar fork from February 2025, not the main plugin).
- Security disclosure: NVD records for CVE-2026-73032 note that versions up to v0.6.1 contained an unsafe
window.eval()call inexecTagreachable via prompt injection; the vulnerability was addressed in PR #155 and tagged in v1.1.0.
- BibGenie: BibGenie is closed source, narrowing its operational surface by avoiding unpinned runtime package execution and strictly isolating library writes behind user-approved diff cards. Its current version is v0.9.6 (November 10, 2026), developed in a private repository with a public releases mirror.
Strengths and limitations
PapersGPT
Strengths
- Free starting tier: unlimited local model usage, 30 cloud sessions with own API key, and 20 trial credit chats
- Whole-library full-text PDF indexing with high-speed local vector retrieval
- AutoPilot batch-reads collections into structured Zotero synthesis notes
- Literature Radar monitors research topics across arXiv, PubMed, Semantic Scholar, and other repositories
- In-plugin download and execution of supported local models (Gemma, Qwen, Mistral, Llama, Phi-4) for offline inference
- Open-source plugin code (AGPL-3.0) with an active community and built-in Zotero MCP server
Limitations
- Built-in credits are paid, and whole-library Q&A and AutoPilot are gated by plan tiers
- Free trial credits (20 chats) are an initial allowance, not a recurring monthly refresh
- General chat citations link to the cited paper rather than highlighting specific in-document text blocks
- No dedicated multi-database metadata review or visual diff-preview workflow
- Large installer size (202 MB) exceeds Atlas automated security analysis thresholds
BibGenie
Strengths
- Local-first architecture: complete chat history and vector indexes reside securely on your machine with zero prompt logging
- Custom multi-model freedom: connect OpenAI, DeepSeek, Claude, Gemini, Ollama, LM Studio, or custom endpoints with 0 platform fees
- Zero-config hosted models: 200 recurring free credits every month with no API key or local setup required
- Traceable citations that jump to source passages and highlight PDF text blocks when coordinates resolve
- Multimodal PDF reader screenshot queries to inspect architecture diagrams, charts, and formulas
- Visual diff cards for safe metadata review cross-referencing arXiv, Semantic Scholar, OpenAlex, DBLP, and OpenReview
Limitations
- Closed source software, precluding external line-by-line source audits
- Semantic index focuses on titles, abstracts, and metadata rather than full-text PDF body embeddings
- No bulk overnight reading automation (like AutoPilot) or automated topic radar digests
- Quick Prompts trigger from the chat panel; note exports currently create new notes rather than appending to existing ones
Who should choose which?
- Choose BibGenie if: You value local-first data sovereignty and custom model freedom (BYOK with zero platform fees); you want a zero-setup daily research copilot with passage citation highlights, multimodal screenshot Q&A, and diff-previewed 5-source metadata auditing.
- Choose PapersGPT if: You conduct systematic literature reviews requiring bulk processing of hundreds of papers; you want full-text PDF indexing across your entire collection; you want in-plugin local model downloads; or you want to track topic alerts via Literature Radar.
- Consider using both if: You leverage PapersGPT's AutoPilot for overnight literature synthesis batches, while using BibGenie for day-to-day precision reading, verifiable passage tracing, and clean library metadata maintenance.