PDF to Markdown with source citations
PDFLLM is a browser-based PDF to Markdown converter built for work that must remain traceable. It extracts embedded text when available, runs local OCR for scanned pages, rebuilds headings, paragraphs, tables and reading order, then links each exported block back to its source page and PDF coordinates. Your PDF stays in the browser by default, and the result can be downloaded as Markdown, ZIP or RAG-ready JSONL.
Product information updated September 2, 2026What does the converter preserve?
Readable structure
Headings, paragraphs, lists, images and recoverable tables become stable Markdown blocks instead of one unstructured text dump.
Verifiable sources
Every block retains its page, anchor and PDF-coordinate evidence so you can compare an answer with the original document.
Portable output
Download plain Markdown, a complete ZIP document pack or section-aware JSONL chunks for downstream AI and knowledge-base tools.
How does PDF-to-Markdown conversion work?
Read the PDF locally
PDFLLM inspects each page in your browser and uses the embedded text layer whenever that evidence is reliable.
Recover difficult pages
Scans and image-only regions are rendered, preprocessed and recognized with local PP-OCRv6, with a fallback for incompatible browsers.
Review before export
The Document Trust Score and page-level risk map surface garbled text, low confidence, table recovery and reading-order uncertainty.
Why source-linked Markdown is different
A useful conversion is not merely readable. It should preserve enough evidence to verify facts and enough structure to reuse the document safely.
| Capability | Basic text extraction | PDFLLM |
|---|---|---|
| Scanned pages | Often empty | Local OCR with quality flags |
| Reading order | Flattened text | Layout-aware blocks |
| Source evidence | Page breaks at best | Page anchors and PDF coordinates |
| AI ingestion | Manual cleanup | Markdown, ZIP and RAG JSONL |
Common PDF-to-Markdown questions
Does PDFLLM upload my PDF?
PDF files, extracted text, page images and export packs stay in your browser by default. Account and payment services receive usage or billing data, not the document contents.
Can it convert tables and multi-column pages?
PDFLLM attempts table reconstruction and reading-order recovery, then marks uncertain results for review. Complex layouts are never presented as guaranteed perfect.
What can I do with the Markdown?
Use it in Claude, ChatGPT, Codex, Obsidian, a code repository or any text workflow. The document pack remains readable without a PDFLLM account.