Help
Everything you need to know to use ArXiParse.
What ArXiParse does
ArXiParse transforms scientific papers into structured data that AI can actually understand. Instead of a PDF that AI can only skim, you get extracted claims, explained equations, and interpreted figures.
Think of it as translation: from academic paper to AI-ready format.
How to use it
1. Paste any arXiv ID (like 2403.00001) or URL
2. Wait about 2 minutes
3. Get structured output you can use with any LLM
Quick tip: On any arXiv page, just swap the domain:
arxiv.org → arxiparse.org
What you can do with the output
- → Feed it to Claude, GPT, or any LLM and ask questions
- → Compare claims across multiple papers
- → Get explanations of equations in plain English
- → See what the figures actually show (not just captions)
- → Find discrepancies between text and figures
Limitations
- AI can make mistakes. Always verify important claims against the original paper.
- Not all papers work. Very recent submissions or papers without HTML conversion aren't available yet.
- Complex figures can be misread. Multi-panel plots and specialized diagrams are harder to interpret.
- 10 papers per day on the free tier.
Common questions
Why does it take 2 minutes?
We're running multiple AI models: one to understand the text, another to read the figures. More figures = more time.
What makes this different from just uploading a PDF to ChatGPT?
ChatGPT can only skim PDFs. We extract the actual structure: what claims are made, what equations mean, what the figures show. The AI can then reason about the content, not just the surface text.
Can I trust the output?
Treat it as a smart assistant's notes, not ground truth. It's great for exploration and synthesis, but always verify specific numbers or claims against the source.
What happens to my data?
Your parsed papers are stored securely on your account. We don't share them.