Semantic document search for when you do not know the exact words
Semantic document search finds passages that mean what you are asking about, even when they use different words. Search+ uses it inside every answer: when you ask a question about the PDFs in a workspace, it retrieves the passages closest in meaning, writes a reply from them, and cites each one. That helps most when you remember the idea but not the wording, or when different authors describe the same thing in different terms. Upload your PDFs, ask the question in your own words, and read the excerpts it cites.
Create an account without entering payment details.
Semantic search in practice
The value of searching by meaning shows up when the words in your question and the words in the document do not line up. Some examples of that gap:
A keyword search for layoffs finds nothing. A question about layoffs in Search+ retrieves the passages on workforce reduction and restructuring and cites them, so you can confirm they are what you meant.
The answer brings back the clause on reimbursement and quotes it through the citation, with a section reference if the PDF has headings.
Describe the finding loosely, such as the study that found remote teams met less often, and Search+ can retrieve a passage that discusses it even if the paper never uses the word remote.
Ask about the idea once. The answer gathers the passages from each PDF that cover it and cites each source separately, so you can see the two vocabularies side by side.
What is semantic document search?
Semantic search represents both the question and every passage of a document as a description of meaning, then compares those descriptions to find the closest matches. Keyword search, by contrast, compares the letters of words, so it only finds text that repeats your exact terms or a close variant. In a document workspace, semantic search is usually the first step of answering a question: the system retrieves the most relevant passages, and a language model writes the reply from what was retrieved. In Search+ that retrieval is limited to the workspace you chose, and every passage used in the reply is cited, so you can see exactly which text the match came from.
How semantic search runs when you use Search+
- Your PDFs are processed into passages
When you upload a PDF, Search+ reads its text in the background and prepares it for retrieval. Each file shows its status, and processing time depends on the size and structure of the document.
- You ask in your own words
Write the question as you think of it. There is no need to guess the vocabulary of the author.
- Passages are matched by meaning
Search+ retrieves the passages closest in meaning to your question from the documents in scope, either the whole workspace or one file.
- The answer cites what was matched
The reply is written from those passages, and each statement links to its excerpt so you can judge whether the match was right.
What semantic search adds to your documents
Passages are retrieved by meaning, not only by exact words, which catches wording you would not have thought to type.
Retrieval stays inside the workspace you choose, or inside one document when you narrow the question.
Each statement in the answer cites the source document and the supporting excerpt, so a weak match is easy to spot.
Citations carry page or section references only when the PDF provides them; Search+ does not invent a location.
A single question can retrieve passages from several PDFs at once and answer from all of them.
Workspace Context lets you write terminology or priorities once, and it applies to every question in that workspace.
Limits of semantic search in Search+
Semantic retrieval is the only search method. If you need every literal occurrence of a term, your PDF reader's find function is still the right tool.
You do not see a ranked list of matched passages or similarity scores. You see the passages the answer cites.
A passage can sound related without answering the question. AI answers can be incomplete or inaccurate, so read the cited excerpts before relying on a result.
Semantic search compared with keyword search
Both methods have a place. This is how they differ on the jobs people bring to a document collection.
| Situation | Keyword search | Semantic search in Search+ |
|---|---|---|
| You know the exact term | Finds every occurrence | Finds the relevant passages, not every occurrence |
| You only know the idea | Often finds nothing | Retrieves passages that mean the same thing |
| Authors use different vocabulary | Needs one search per term | One question covers the variants |
| You need to check the result | Shows the matched word | Cites the excerpt behind each statement |