Skip to main content
A smart split uses a natural language prompt to decide how Box groups the pages of a PDF into output documents. Clear, specific prompts produce more reliable splits. Use a smart split when the boundaries between documents depend on what is on the page rather than on fixed page numbers.
Box Doc Split is available through the Box API only. To split documents, work with your development team or see Smart split in the Box Developer Documentation.
A smart split consumes 1 AI Unit for every 4 pages processed. For details, see Understanding AI Units in Box.

Before you begin

Make sure you have:
  • Box Doc Split and Box AI enabled. See Enabling Box Doc Split.
  • A source PDF already stored in Box
  • A destination folder in Box that you can write to
  • A prompt that describes how to group the pages

What you define

A smart split request includes the following:

Writing an effective prompt

Follow these guidelines when you write a prompt:
  • State that the Box file you supply is the source document.
  • Describe the splitting rule, such as a page count, a marker like a QR code, or a content criterion.
  • Prefer sequential ranges when order matters, such as contract exhibits followed by signature pages.
  • Call out the pages or sections to include, and where it helps, the ones to skip.
  • Avoid ambiguous language such as “split somehow” or “use whatever makes sense.”

Example prompts

Explicit page groups

Use this approach to split a file into 2 documents whose ranges you describe in natural language.
This produces 2 documents, one for pages 1 to 3 and one for pages 5 to 8.
If you already know the exact ranges, use a manual split instead. The results are the same every time.

QR code as a break point

Use this approach to treat each QR code page as a separator between documents. Pages before a QR code form one output document, and Box excludes the QR code page itself unless you say otherwise.
This produces sequential ranges bounded by the marker pages. In the example, markers on pages 5 and 11 produce documents for pages 1 to 4 and pages 6 to 10.

Content-based contractual pages

Use this approach to keep only the pages with substantive contractual content, such as ID proofs, signatures, and executed terms, and to skip boilerplate terms and conditions.
This produces one output document for each continuous block of content that matches the criteria. In the example, that means documents for pages 1 to 3 and pages 4 to 6.

When to use a manual split instead

Choose a manual split when you already know the page ranges. A manual split does not consume AI Units and returns the same result every time.
Last modified on September 23, 2026