Embeddings
Embeddings are numerical representations of meaning that let systems compare how similar two pieces of text (or images) are, even when wording differs.
They are the engine behind semantic search, recommendations, and clustering user feedback. They are usually invisible but shaping what content the AI can find.
What it means
Text is converted to vectors in high-dimensional space; closer vectors mean more similar meaning, enabling “find like this” search beyond keywords.
Why designers should care
When search feels magically relevant, or misses obvious synonyms, embedding quality and chunking strategy are often why; design feedback and “wrong source” reporting loops.
Example
A component library assistant matches “primary button disabled state” to your Button docs even though the doc says “default CTA inactive,” because embeddings capture intent, not exact words.
Common mistakes
- • Keyword-only search UI labeled as “AI search” without semantic behavior.
- • Chunks so large that retrieved passages are useless walls of text.
- • No way for users to flag irrelevant retrieved snippets.