---
language: en
title: "Chunking"
description: "Chunking divides documents into sections for processing and retrieval. The size, boundaries, and overlap affect how much context each section preserves."
canonical: "https://tropk.ai/en/resources/glossary/chunking"
last_modified: "2026-10-12T16:13:13.659Z"
---

# Chunking

Reviewed on 2026-10-06.

## What it means

Chunking is the process of dividing a document into smaller units, called chunks or sections, for processing, [indexing](/en/resources/glossary/indexing), or [retrieval](/en/resources/glossary/information-retrieval). The division can be based on size, sentences, [headings](/en/resources/glossary/headings-html-section-titles), or relationships in meaning. In AI applications, it makes it possible to work within models’ input limits and can represent parts of a document more specifically than a single block.

## Why it matters for marketing

Anyone hiring a content assistant needs to evaluate the sections that feed its search. Poorly chosen cuts can separate a condition from the offer it applies to. Sections that are too broad can combine different topics. The chunking decision involves continuity, accuracy, and processing cost, and should take into account expected questions as well as document formats.

## Example in practice

Imagine an agency indexing a sponsorship guide with tiers and benefits. Splitting by character count separates the Gold tier name from its list of benefits. On inspection, the section appears without identifying the tier. The agency decides to test section-based boundaries and keep the heading with the benefits before increasing the size of all chunks.

## How to apply or check

Ask for a sample of the chunks produced before approving indexing. Check for incomplete sentences, missing headings, split tables, and isolated exceptions. Compare two strategies using the same questions and the same embedding model. Record the size, unit of measurement, and overlap—that is, the text repeated between neighboring sections—so you can explain any differences observed in retrieval.

## Important distinctions

Chunking does not summarize or correct the document: it organizes processing units. Tokens, characters, and words are different measures, so a stated size needs to specify its unit. Overlap can preserve continuity, but it also repeats information. There is no ideal length for every collection, and a library configuration does not determine how external search engines divide public pages.

## Sources and references

- [Chunk large documents for RAG and vector search in Azure AI Search](<https://learn.microsoft.com/en-us/azure/search/vector-search-how-to-chunk-documents>) — Microsoft

## Continue a leitura

- [Semantic chunking](</en/resources/glossary/semantic-chunking>)
- [Embeddings](</en/resources/glossary/embeddings>)
- [Passage retrieval](</en/resources/glossary/passage-retrieval>)

## About the author

[Vinicius Pacheco](</en/team/vinicius-pacheco>)

Responsible for Tropk’s technical content, focusing on crawling, indexing, structured data, and agent access to content.

---

[All pages in Markdown](<https://tropk.ai/sitemap.md>)