Auto Extraction (ResearchPilot™)

Before you begin Auto Extraction requires articles to be added to the Data Extraction stage and extraction questions to be defined before running. It also requires a PDF to be attached to each article — articles without a PDF will not be processed. For best results, ensure PDFs are text-searchable rather than scanned image files. See How to Create and Organize Your Extraction Questions.

Auto Extraction reads your full-text PDFs and automatically populates your Data Extraction form based on the questions you defined. For each question, it identifies the most relevant text in the PDF, extracts a structured answer, and highlights the source so you can verify it directly. Always review and validate results before treating them as final.

How to run Auto Extraction

1. Go to the Data extraction tab.
2. Click an article to open it. The PDF view opens alongside the extraction form.
3. Click Auto Extraction at the top right of the article view.
4. ResearchPilot™ analyzes the PDF and begins populating answers for each question.
5. Review each extracted answer — expand it to see the source text highlighted in the PDF and the surrounding context.
6. Accept, refine, or override answers as needed before marking the article as complete.

Getting the best results

• Write specific extraction questions — vague questions produce unreliable answers. "What is the study design?" yields better results than "Tell me about the methods."
• Use Specified answer types where possible — predefined answer options help the AI return structured, consistent answers rather than free text.
• Upload text-searchable PDFs — the AI cannot read scanned image-based PDFs. Attach a text-searchable version wherever possible.
• Always validate results — especially for numeric values, study design, and outcomes. Verify each extracted answer against the source PDF before exporting.