PDF Parsing and Reading

Entry points and where records are saved

In research mode, open “Research Toolbox → Document Parsing,” or use the standalone Document Parsing page. If you need only the reference list, select “Reference Extraction.”

Parsing does not require selecting a project first. Upload records are stored in the current account's personal parsing history. Completed results can be added to project materials or used in conversations. The history is not restricted to the current project.

Upload and parse

1. Select or drag in local PDFs. You can submit several files at once; each file must be no larger than 50 MB.
2. Submit them according to the page's instructions and check the status of each file.
3. Successfully submitted jobs continue processing in the background. You can close the page and return to the parsing history later.
4. When a job is complete, open its record to view Markdown, structured data, references, or AI notes.

Full-text parsing and reference extraction are different operations. A references-only record can be upgraded to full parsing where the interface offers that action. Parsing again and generating AI notes may consume additional credits; check the notice before confirming.

Read and export

• Markdown: Read and download the extracted full text.
• Structured data: Verify recognized fields such as title, authors, and abstract.
• References: Review citation details and export formats such as BibTeX or RIS.
• AI notes: Generate reading notes as needed; existing notes are preserved.
• Original PDF: When available, open the original preview to verify figures, formulas, and page layout.

Parsing quality depends on the PDF's structure and scan quality. Scans may need OCR first, and restrictions must be removed from encrypted files. Verify complex formulas and tables against the original.

Bring results into a research conversation

Select one or more parsed full texts, click “Use in Conversation,” confirm that the attachments have been added to the draft, and then enter a question. For example:

Using the full text of these two papers, compare treatments for the interface between the solid electrolyte and the electrode.
List the material systems, experimental conditions, results, and limitations for each, and identify the supporting passages.

Compare several papers in a research conversation with explicitly selected materials. Do not assume the parsing page automatically reads every PDF in a project. Reading tools can also handle a selected paper through the question-and-answer entry offered on the current page.

Find results and handle failures

Parsing history supports search and filters. You can also open the same result from the PDF category in Personal Space or from Conversation Files. In research mode, these entry points keep you in the workspace and provide a button to return to the source.

• Still queued or parsing: Check the status and avoid submitting the same file repeatedly.
• Parsing failed or timed out: Read the error and check whether the file is damaged, encrypted, or contains only scanned images.
• Insufficient credits: Buy or redeem credits, then use the page's parse-again action. The file and existing results are preserved.
• Parse again: Creates a new parsing record while preserving the original results and notes for comparison.

Reopening an existing record only reads its results. Parsing again, upgrading, or generating notes starts a new processing operation.