Glossary
What Is Chunking in AI?
Chunking is the process of breaking a large document into smaller, self-contained pieces of text so an AI system can store, search, and pull the right snippet when answering a question. Each piece, called a chunk, usually covers one idea or section.
When you give an AI chatbot your website, help docs, or PDFs, it doesn't read the whole thing every time someone asks a question. That would be slow and expensive. Instead, the content gets split into small pieces ahead of time. Those pieces are chunks, and the splitting step is chunking.
Think of a 20-page return policy. Chunking might cut it into 40 short passages, each a paragraph or two. When a customer asks "how long do I have to return a sweater?", the AI finds the one chunk about return windows and reads just that. The answer comes back faster and stays on topic because the system isn't wading through 19 pages it doesn't need.
Chunk size matters. Cut too small and a chunk loses context, like grabbing one sentence that mentions "30 days" without saying 30 days for what. Cut too big and you stuff in extra material that muddies the answer. Most setups aim for a few hundred words per chunk, often with a little overlap between neighbors so a sentence split across two chunks doesn't get lost.
Good chunking usually follows the natural shape of your content. Splitting on headings, list items, or FAQ pairs keeps each chunk about one thing. That's why clean, well-organized source pages tend to produce better answers than one giant wall of text.
For a chatbot or voice agent on your site, chunking is what lets it quote your actual policies and product details instead of guessing. The cleaner your chunks, the more accurate and grounded the replies your visitors hear.
Related terms
Frequently asked questions
Why does an AI chatbot need chunking at all?+
AI models can only read a limited amount of text at once, and reading entire documents per question is slow and costly. Chunking lets the system store small pieces and fetch only the few that match a question. This keeps answers fast, focused, and cheaper to run.
What is a good chunk size?+
There's no single right number, but a few hundred words per chunk is a common starting point. The goal is one clear idea per chunk with enough context to stand on its own. Many systems add a small overlap between chunks so sentences that fall on a boundary aren't lost.
Does chunking change my actual website content?+
No. Chunking happens behind the scenes on a copy of your text that the AI uses to search. Your real pages and documents stay exactly as they are, and you can update the source anytime to refresh the chunks.