Semantic Chunking
Semantic chunking is a way of splitting documents into pieces for AI apps by meaning, cutting where the topic shifts instead of every set number of words. In a 2024 test, Renyi Qu and colleagues found its extra computing cost wasn't backed by steady gains over fixed-size chunking, though it helped on test documents stitched together from many topics.
Source: Qu, Tu and Bao, Is Semantic Chunking Worth the Computational Cost? (arXiv, October 16, 2024), checked October 29, 2026
What Semantic Chunking means for your business
This is a developer's term. When someone builds an AI app that answers from a pile of files, the app splits those files into small pieces so it can find the right one fast. Fixed-size chunking cuts after a set length of text, whatever it says. Semantic chunking compares each sentence with the next and starts a new piece when the meaning changes, or groups sentences that talk about the same thing.
The finding goes against a common belief that splitting by meaning must work better. On real documents with normal structure, the paper found the simpler fixed-size method often did better.
For a business website, this is a setting inside someone else's app, not something you do to your pages. We haven't found a public source from OpenAI or Google that says their search tools split web pages this way. Advice to chop a website into chunks is a separate question, and the post linked below takes it on.
Whatever method a tool uses, a piece cut from a section that sticks to one topic is still about one thing. If one section of your page covers pricing, hours and a company history all at once, giving each section one clear job is where our content work would start.
Last reviewed October 29, 2026. All terms
How AI answers it today
On October 29, 2026, both gave the developer meaning. Google's AI Overview cited software company blogs, a Reddit thread and YouTube tutorials, and the results below included IBM and a research paper. ChatGPT walked through the steps with sample code and cited nothing.
Our run through ChatGPT and Google's AI Overview, October 29, 2026. A sample, not what every buyer sees.
Related terms in answers and how content gets used
Find out where your brand stands.
Tell us what you're trying to figure out. We'll get back to you within one business day with a time to talk.