Chunking
Chunking is how an AI or search system cuts a page into smaller pieces, so it can store them and find the right one for a question.
Chunking is the step where an AI or search system cuts a page into smaller pieces, called chunks, before storing it. Later, when a question comes in, the system searches those pieces and pulls out the ones that match best.
How a platform chunks your page is up to the platform. The shape of your page decides where its cuts are likely to fall.
How systems chunk a page
There are a few common approaches, and many systems mix them:
- By length: cut every few hundred words, sometimes with a little overlap between pieces
- By structure: cut at headings, paragraphs or list items
- By meaning: cut where the subject changes
Each platform picks its own method, and most don't publish the details.
Why your structure matters
A page with clear headings and one subject per section gives every method an easy job. The cuts land between subjects, and each piece holds one complete idea.
A page that runs on as one long block is harder. A cut by length can land in the middle of a sentence, leaving half of your answer in one chunk and half in the next.
For example, Harrow & Pine's care guide covers oiling, cleaning and repairing scratches in one long section with no headings. A system cutting it every 300 words splits the oiling advice in two. Neither piece fully answers how often to oil an oak table, so neither is likely to be used.