Lexical, Vector and Hybrid Search
One customer types "SP-60 pot" into Leafline's site search: a product code. Another types "big planter for a tall snake plant". The first needs an exact match on a code no dictionary knows. The second needs a match on meaning, because no product is literally called that. Each needs a different kind of search, and modern systems run both.
The rule
Lexical finds the exact words, vector finds the meaning; hybrid finds both.
Lexical search looks up the query's words in an index that records which pages contain which words. Pages score higher when they contain more of the query's words, especially the rarer ones. It is precise and cheap. Its weakness: it misses a page that says the same thing in different words.
Vector search turns the query and every page, or every passage, into an embedding: a long list of numbers arranged so that texts with similar meanings sit close together. The search returns the passages nearest the query. It handles paraphrases well. Its weaknesses: exact tokens such as product codes and rare names, and returning something merely similar, like a spider plant page for a snake plant query.
| Query | Lexical | Vector |
|---|---|---|
| "SP-60 pot" | Finds the product with that exact code | May return any large pot |
| "big planter for a tall snake plant" | Misses "60 cm ceramic pot" | Finds it by meaning |
Hybrid search runs both and merges the two ranked lists. A common merging method, reciprocal rank fusion, gives each page a score of 1 ÷ (60 + its position) from each list it appears in and adds the scores up.
For SEO the lesson is practical: use the exact terms people search for, such as names, codes and specific phrases, so lexical matching finds you, and explain the idea clearly and completely, so vector matching finds you too. Many AI answer systems retrieve passages in a similar way, which is another reason to keep each passage focused on one idea.
The shortcut
To predict a hybrid result, look for the pages that appear near the top of both lists. They almost always win.
The lexical list
1: SP-60 ceramic pot. 2: Snake plant care guide. 3: Pot size chart.
Drills
Answer fast — trust the rule.
Which search is best at finding the exact product code "SP-60"?
What is an embedding?
A pure vector search for "snake plant care" returns a spider plant page. What went wrong?
Using 1 ÷ (60 + position), which page wins: page X at position 1 in the lexical list only, or page Y at position 2 in both lists?
What does hybrid search suggest about writing a page?
Remember
- Lexical finds exact words, vector finds meaning, hybrid merges both lists.
- Shortcut: pages near the top of both lists win the merge.
- The trap: dropping the exact names and codes people type, or never explaining the idea behind them.