A sentence embedding is a numeric vector that represents a sentence or phrase so that texts with similar meanings sit close together in vector space, letting software compare meaning instead of matching exact wording.
| Term | Sentence Embedding |
|---|---|
| Category | Content and Topical Relevance |
| Also known as | Text Embedding |
| Where it appears | Semantic enrichment evaluator |
What it means in RankGear
RankGear generates sentence embeddings inside the Semantic enrichment evaluator, which reads the passages on your page alongside the passages across the competitors it analyzes. Each passage becomes a vector, and passages that express the same idea land near one another even when they use different words. That is what lets the evaluator judge whether a page genuinely covers a concept, entity, or question the ranking field keeps returning to, rather than only checking for a literal string match.
How to interpret it
Read the embedding as a similarity signal, not a score. When the evaluator reports that your page sits semantically far from an idea the top results share, take it to mean that the idea is thinly covered or phrased in a way the model does not recognize as related. Respond by adding real, useful material — the missing entity, the unanswered question, the supporting concept — in natural language. Chasing a similarity number by repeating phrases does not help; embeddings reward meaning and coverage, not keyword density.
Example
Suppose you are optimizing a page about container ships, and the evaluator embeds every passage from the ranking field. Pages that rank consistently include passages on draft limits, TEU capacity, and port congestion. Your page handles freight cost well but never touches those subtopics, so its passages sit far from that cluster in vector space. The evaluator surfaces the gap, and you close it by writing a genuine section on capacity and port constraints — not by sprinkling the phrase “container ship” more often.
Important considerations
- Embeddings measure semantic similarity, not correctness or ranking. A high-similarity passage is a comparative indicator of coverage, not a Google score, and it does not by itself make a page rank.
- Embedding values are model- and provider-specific. Vectors from one embedding model are not comparable with another’s, so RankGear compares your page against competitors using the same model to keep the comparison like-for-like.
- Related wording should preserve search intent. Rephrasing to match a cluster only helps when the new text truly serves the reader; repetition, stuffing, and unsupported claims do not create topical authority.
- Embeddings capture meaning, not truth. Two false statements can be semantically close, so use the signal to guide coverage and then verify the facts yourself.
Related terms
Part of the RankGear glossary · how RankGear measures · the 870 factors.