What it means
A vector, or embedding, is the list of numbers a model assigns to a piece of text to capture its meaning. Each chunk of your content becomes a point in a high-dimensional space, and the engine answers a query by finding the chunks whose vectors sit closest to the query's vector. Kevin describes the mechanic plainly: In a vector database, the word "is" acts as a strong bridge connecting a subject to its definition. The embedding is the actual unit your content competes as inside an AI engine, not the words on the page.
Why it matters
Identical text can produce different embeddings across models, so citation behavior diverges across ChatGPT, Perplexity, and AI Overviews (AIOs). A page that sits close to a query in one model's vector space can sit just outside the cutoff in another. That is why a single piece of content gets cited by one engine and ignored by the next, and why you cannot treat AIO visibility as one monolithic target.
Take a SaaS comparison page that ranks well and gets pulled into ChatGPT answers for "best CRM for startups." In Perplexity, the same page's embedding lands fractionally further from the query vector, it never clears retrieval, and the brand watches a competitor own that answer despite weaker classic rankings.
How to use this knowledge
Write the definition explicitly. Use clear "X is Y" sentences so the embedding bridges your subject to its meaning, the path models follow for "what is" queries.
Test across engines, not one. Track citations in ChatGPT, Perplexity, and AIOs separately, because the same page produces different embeddings and different outcomes.
Keep each chunk on one idea. A chunk that mixes topics produces a muddy vector that sits far from any single query.
Add the entities a query implies. Naming related people, products, and concepts pulls your embedding closer to the queries you want.
Growth Memo guidance
In a vector database, the word "is" acts as a strong bridge connecting a subject to its definition. When a user asks "What is X?", the model searches for the strongest vector path, which is almost always a direct "X is Y" sentence structure.
You compare how similar two pages are by turning their text into numbers and seeing how closely those numbers point in the same direction.
Vectorization — the process that produces an embedding from text.
Embedding similarity thresholds — the cutoff distance that decides whether your embedding is close enough to be retrieved.
Cosine similarity — the math that scores how close two embeddings are.
Chunk — the text span that gets turned into a single embedding.
Entity — a concept whose presence shifts where your embedding lands.