Readers find the page, not the words
Your documentation is written in your product's vocabulary. Your readers ask in their own. Semantic search understands what they mean and returns the right page — without changing a single line of your content.
One question, two search engines
A reader wants to know who covers the shipping cost. Your page on the subject is called Incoterms, and it never uses the word "freight" — because the person who wrote it knew the correct term.
The reader searches
who pays for freight
Looks for pages containing "freight". Your Incoterms page contains none of these words, so it cannot be returned at all.
No results — or a shipping-policy page that happens to say "freight" once.
Compares what the question means with what each page means. "Who pays for freight" and "delivery cost responsibility" are the same question.
Incoterms — first result.
This is the whole feature. Not better ranking of the pages keyword search already found — pages it could never return in the first place.
Measured on a real customer manual
21 questions against a published manual of 88 pages, run through both engines and scored by hand.
Everyday-worded questions answered in the top five
Questions phrased the way a reader would ask them, not in product vocabulary.
Correct page ranked first
Of 21 questions, how many put the right page at position one.
Exact product-name lookups
Searching for a part number, an error code or a product name. Unchanged — this is the point.
Keyword search
100%
Semantic search
100%
Nothing was traded away. A search engine that finds vague questions but loses part numbers is a downgrade for the people who use documentation most.
Nothing changes on your side
Search improvements usually arrive with a project attached — restructure the content, add metadata, re-tag everything. This one does not.
-
No re-authoring, no re-tagging
Sonat indexes the topics you have already published. You do not add keywords, synonyms or metadata, and you do not restructure anything.
-
No URL changes
Your pages keep their addresses. Nothing that links to your documentation today breaks.
-
Reversible, because keyword search stays
Keyword search is never removed. Choosing between the two is a setting in your organization settings — switch back at any time and readers see it immediately.
-
Check it before your readers see it
Installing the addon indexes your content; it does not switch your readers over. You run the initial indexing, try the results yourself, and change the engine when you are satisfied.
It is not only the search box
An AI answer is only as good as the pages retrieved to build it. Semantic search feeds retrieval for the search box and for AI answers, so the assistant draws on better source material from the documentation you already have — with nothing rewritten.
And when an answer still cannot be found, AI Insights records the question as a content gap. Better retrieval means the gaps that remain are real missing content, rather than pages your search could not reach.
Availability
Semantic search is included on the Business and Enterprise plans. An owner or admin installs it from the Addons page, runs the initial indexing over your published topics, and then switches the search engine over.
Frequently asked questions
What is semantic search?
Semantic search matches meaning rather than words. Keyword search can only return a page that contains the words you typed, so a reader who asks 'who pays for freight' never reaches an Incoterms page that never uses the word freight. Semantic search compares what the question means with what each page means, so the right page comes back even when the two share no vocabulary.
Will it make exact searches worse?
No. On the benchmark corpus, exact product-name lookups scored 100% before and 100% after. This is the result worth checking, because a search engine that finds vague questions but loses part numbers is a downgrade for the people who use documentation most.
Do I have to rewrite or re-tag my documentation?
No. Nothing about your content changes — no re-authoring, no re-tagging, no new metadata, and no URL changes. Sonat indexes what you have already published.
Can I switch back to keyword search?
Yes, at any time. Keyword search is never removed. Switching between the two is a setting in your organization settings, not a migration, and readers see the change immediately.
Does it improve AI answers as well as the search box?
Yes. The same retrieval feeds both. AI answers are only as good as the pages retrieved to build them, so better retrieval improves answers from the documentation you already have, without rewriting anything.
Which plans include semantic search?
Semantic search is included on the Business and Enterprise plans. It also has to be installed from the Addons page by an owner or admin — being on the plan makes it available, it does not switch it on.
What gets indexed?
Your published topics. Drafts, shortcuts and empty pages are skipped by design, so the indexed count is normally lower than your total number of topics. Installing the addon indexes topics published from that point on, so existing content needs a one-off initial indexing run — the semantic option stays unavailable until it finishes.
