What it does

AI4Paper is a Zotero 7–10 plugin that tries to handle the full research pipeline—finding papers, reading them, managing metadata, and even drafting literature reviews—without leaving your reference manager. It bundles machine translation, AI chat over PDFs, automated tagging, and a survey generator that claims to crawl citation networks and output 3,000–5,000 word reports with sourced claims.

The interesting bit

The project isn’t just gluing OpenAI to a PDF viewer. It builds a RAG system atop a proprietary 240-million-paper index with citation-network data, and it lets you swap between 20+ LLM providers (including local Ollama) without reconfiguring the rest of your workflow. The “fake paper detector”—which cross-checks pasted reference lists against authority databases—is a rare admission that LLM hallucinations are a real hazard in research tools.

Key highlights

• One-click full-PDF translation preserving layout and formulas, with 12 translation engines available for word-level lookup
• AI survey generation with iterative refinement: keyword search → citation crawling → multi-round filtering → structured report with numbered source annotations
• WeChat mini-program for journal alerts that feed back into Zotero
• AES-256 end-to-end encrypted note sync, or plain Zotero native sync
• Supports 20+ AI backends including fully offline Ollama

Caveats

• The 240-million-paper knowledge base and citation network are proprietary; no details on coverage gaps or update frequency
• Core AI features require account registration and a 7-day trial; pricing and offline functionality limits are unspecified
• “Copyright © 2026” in the license suggests the project is forward-dated or the README is templated

Verdict

Worth a look for Chinese-speaking researchers deep in the Zotero ecosystem who want an integrated, vendor-agnostic AI layer. Skeptics should verify the survey quality against their own domain expertise before trusting the citation crawler.