---
sidebar_label: Smart Document Splitting
sidebar_position: 5
description: DocJacket's AI detects document boundaries in multi-document PDFs, lets you review them when needed, and splits selected files before extraction.
---

<!-- Canonical: https://help.docjacket.com/docs/ai-features/smart-document-splitting -->
<!-- Source: docs/ai-features/smart-document-splitting.mdx -->

# Smart Document Splitting

Sometimes you get a single PDF that contains multiple documents — a contract, addenda, disclosures, and amendments all bundled together. DocJacket's AI can detect where one document ends and another begins, pick likely extractable documents, and send the selected documents into extraction.

## How to use it

1. Start **Upload Documents** and upload one PDF
2. If the PDF has multiple pages, DocJacket opens the **Smart Split** step
3. AI identifies the documents, page ranges, document types, and likely primary contract
4. Review the selected documents, extractable-document selection, and primary contract
5. Edit boundaries, split a document, remove a document, add a document, or change the selected contract if needed
6. Click **Split & Extract** to create the split files and start extraction

If DocJacket detects one document spanning the full PDF, it skips the split step and continues to extraction. If the detected split is high-confidence, DocJacket shows a 5-second auto-extract countdown. Click **Cancel & review** during that countdown if you want to inspect or change the split before extraction starts.

Smart Split appears in the broader **Upload Documents** extraction wizard. The transaction **Documents > Files > Extract Data** flow uploads directly into an existing transaction and then opens extraction progress/review.

## What the review screen shows

The Smart Split review screen includes:

- A detected-document list with selected count
- The likely primary contract
- Document type and page range for each detected document
- Confidence labels for each detected document
- A PDF preview and page map
- **All** and **None** selection controls
- Edit tools for boundaries, split points, added documents, and removed documents

DocJacket auto-selects extractable document types, such as purchase agreements, listing agreements, amendments, addenda, counterproposals, and similar contract documents. You can select or clear documents before extracting.

## Confidence and warnings

Documents below the review threshold are marked for review. If any selected document is below 70% confidence, DocJacket warns before continuing because a bad page boundary can produce inaccurate extraction results.

Low confidence does not mean the AI failed. It means you should verify the page range and document type before sending the selected pages into extraction.

## When this is useful

- **Title company packages** — Often arrive as one big PDF with dozens of documents
- **Agent submissions** — Agents sometimes scan everything into a single file
- **Email attachments** — Documents forwarded as a single combined PDF
- **MLS downloads** — Property data packages that bundle multiple forms

## Tips

- The AI names each split document based on its content (e.g., "Purchase Agreement", "Lead Paint Disclosure")
- You can rename any document after splitting
- Each split document gets categorized automatically
- If the AI misidentifies a split point, adjust the page ranges before confirming
- If you only want some files extracted, adjust the selected documents before continuing
- Use **Skip Split** only when you want to extract from the original uploaded PDF without creating split files first
