Definition
Semantic search is a search method that understands the meaning and intent behind a query instead of only matching the literal words. Someone searching 'how do I get my webshop higher in ChatGPT' gets results about generative engine optimization, even if those pages never use those exact words. The search engine compares the meaning of the question with the meaning of pages.
Google has grown into it step by step since 2013: Hummingbird understood whole sentences, RankBrain (2015) interpreted unknown queries, BERT (2019) understood the context of words in a sentence, and MUM and later Gemini understand questions across languages and media. Under the hood, queries and passages are converted into vector embeddings, numerical representations of meaning, and matched by proximity in that meaning space.
AI engines go one step further. ChatGPT, Perplexity and AI Mode split one question into several sub-questions (query fan-out), search for them semantically, retrieve passages that best cover the meaning and write an answer from them. A page therefore does not need to contain the exact keyword to be found; it needs to answer the question best in substance and name the related entities and concepts.
For content, this means a shift from keyword density to topic coverage. A page about 'AI SEO' that also explains what GEO is, how citations work, which engines exist and how it differs from classic SEO covers the meaning space around the subject and scores higher semantically than a page that repeats 'AI SEO' twenty times. Synonyms, related concepts and concrete examples help the model place the page correctly.
Semantic search also rewards structure: headings that ask questions, paragraphs that explain one thing, and entities named explicitly make it easier to connect a passage to a question. Traze builds a semantic map of questions, entities and related concepts per subject and writes pages that cover that space instead of repeating one keyword.