Help Center
How search works
DECA KB uses three related mechanisms: keyword search (match the words you type), semantic search (find articles that mean something similar even with different wording), and Ask answers (AI writes an answer using articles it finds). Which one you see depends on the screen or integration.
Three mechanisms compared
| Mechanism | What it does | Where you see it | What you get |
|---|---|---|---|
| Keyword search | Matches text in titles, body, and keywords (and similar fields). | Top search in the admin, search inside a folder, public FAQ search box | A list of folders, knowledge bases, articles, and documents (ranked by relevance) |
| Semantic search | Finds content that is close in meaning, even without the same words. | When Ask or AI features pick articles; often combined with keyword search to rank results | A list of articles and documents |
| Ask answers | After finding relevant articles, AI writes an answer to the question—not a plain copy of one article. | Ask in the admin or FAQ, chatbot integrations | Answer text plus links to source articles |
Some Ask setups return only a list of articles without generating answer text. That is different from “Ask answers” in the table above.
Search in the admin and search on a public FAQ treat private knowledge bases and articles differently. Users without access never see private content.
What keyword search looks at
In the admin (within your workspace), you can search across:
- Folders — name
- Knowledge bases — name and description
- Articles — title, body, keywords, and custom data (if configured)
- Documents — title and other fields shown in lists
Short, specific terms usually work best. Pasting a very long paragraph may not match the way you expect.
Narrowing with is: prefixes
At the start of the search box query, you can limit the type (the prefix is removed before matching):
| Example | Meaning |
|---|---|
is:article returns | Search articles only for “returns” |
is:document policy | Documents only |
is:folder sales | Folders only |
is:kb support | Knowledge bases only |
Folders and workspaces
- When you search within a folder, results focus on that folder and what sits under it.
- Search stays in your current workspace; other workspaces are not mixed in.
Semantic search (similar meaning)
When articles and documents are published, the product updates data used for meaning-based matching from titles, visible body text, and keywords.
When Ask or other AI features look for articles, they often use both keyword matching and meaning to rank results—so a question that does not use the exact words in an article can still find it.
- Word match — title, body, keywords (for Ask, AI Search Context can count here too)
- Meaning — title, body, keywords on screen (AI Search Context is not used for meaning-only matching)
Ask answers
Ask selects relevant articles or documents, then shows an AI-written answer. You may also see links to the articles used as sources.
If articles are too short or off-topic, the answer may be empty or look like nothing was found. Step-by-step body text usually works better than one-line bullets.
In the article editor, AI Search Context is extra text for search that is not shown on the FAQ. Ask may use it for word matching when finding articles, but it is not used for meaning-only matching and is not quoted in the generated answer. Normal admin search and FAQ keyword search do not use it. Put internal names, codes, or aliases here so Ask can find the right article.
Search on public FAQ sites
Visitors usually search with a keyword search box. Depending on your FAQ or chat setup, Ask (AI answers) or meaning-based article finding may also be available.
Only published, public content is included. AI Search Context never appears on FAQ pages and is not part of FAQ keyword search.
Draft, private, and published
- Draft articles are not searchable on public FAQ. In the admin, they may appear if you have permission.
- Private knowledge bases and articles are hidden from users without access and from public FAQ.
- After you publish, results may take a short moment to update. Articles from CSV import are usually searchable right away. Brand-new knowledge bases or folders may need a few seconds to a minute.
Tips for better findability
- Put words users might type into the title.
- Add aliases and product codes in keywords (CSV column 3 or the article keywords field).
- Put terms you do not want on the FAQ in AI Search Context (not in the body). Put terms you want meaning search to use in the body or keywords.
- Use short, concrete terms in the admin search box; paraphrases and vague questions often work better with Ask and semantic matching.
Related
- Import articles from CSV — search and Ask after import