Exact Nearest Neighbor
Also called Brute-Force Search
Actually comparing your query against every item to find the true closest — correct, but it does not scale.
Think of it like
Reading every résumé in the pile to find the best fit, instead of trusting a filter.
Example
For a few thousand FAQ entries, a startup just loops over all of them — no index needed, no recall loss.
How it actually works
Exact search computes the distance to every stored vector and keeps the smallest. It is perfectly accurate and dead simple, which makes it the right choice for small collections or as a correctness baseline. The catch is linear cost: double the data, double the work, so past a certain size everyone reaches for approximate methods.
For product teams
Great for small datasets or when you cannot tolerate any miss; a trap if your data is going to grow.
For engineers
O(n) scan with a top-k heap; used as ground truth to measure an ANN index’s recall.
Related
- Approximate Nearest Neighbor — The scalable, slightly lossy alternative.
- Vector Similarity — Compares vectors using a distance measure.
Read anything AI without the jargon
Look up any term in plain English, or save terms as you read with the free Chrome extension.
Open DecoderAdd to Chrome