Can AI search inside your WordPress documents?

Yes — and it’s worth understanding what "AI search" actually adds, because it solves a different problem than the full-text search you may already have. Keyword search finds documents containing the words you typed; AI search finds documents that mean what you asked, even when the wording doesn’t match. This article explains the difference in plain terms, what’s happening under the hood, and what to realistically expect.

FileDeck AI — semantic search and library Q&A are part of the optional AI tier, which sits above Free and Pro. You bring your own AI provider key, and it’s off until you enable it. See pricing

Keyword search vs. meaning search

Full-text content search (FileDeck Pro) already looks inside your PDFs and Office files — a search for "parental leave" finds every document containing that phrase. Its limit is literal wording: a visitor searching "time off after having a baby" gets nothing from a policy that only says "parental leave."

AI semantic search closes that gap by matching on meaning. Both the documents and the search query are converted to numerical representations of what they’re about (embeddings), so "time off after having a baby" and "parental leave policy" land close together and match — no shared keywords required.

When meaning search actually matters

The bigger and more formal your library, the more it matters:

• Policy manuals and HR libraries — staff ask questions in everyday language; policies are written in formal language. Keyword search punishes that mismatch daily.
• Contracts and technical documents — the same concept goes by different names across documents ("termination for convenience" vs. "cancellation rights").
• Public resource libraries — visitors don’t know your terminology, and every failed search is someone leaving.

For a 30-document library where everyone knows the file names, keyword search is honestly enough — semantic search earns its keep at scale.

From searching to asking

One layer up from semantic search is Ask this library: a chat widget where visitors ask a question and get an answer composed from your documents, grounded with references to the source documents — not from the AI’s general knowledge. Behind the scenes it uses the same semantic index to find the relevant passages, then answers only from what it found.

What to expect (and not expect)

• It’s grounded, not omniscient. Answers come from your library’s content, with inline links to the source documents behind each answer. When a question is only loosely related to your library, it says so plainly rather than guessing.
• Scanned documents need OCR first. A scanned PDF is pictures of text; scanned-PDF OCR turns it into searchable text before any search — keyword or semantic — can see inside it.
• Your data stays in your database. FileDeck stores the semantic index (the vectors) in your own WordPress database, and you connect your own AI provider key — there’s no third-party middleman holding your library.

How to turn it on

Semantic search is configured under Documents → Settings → AI search, and the chat widget under Documents → Settings → Ask AI — both covered step by step in AI semantic search and Ask this library.

FAQ

Can visitors ask questions about my PDFs instead of typing keywords?

Yes — Ask this library answers visitor questions from your documents’ actual content, with references to the source documents, using your own AI provider key.

Do I need AI search if I already have full-text search?

Not necessarily. If your visitors search using the same words your documents use, Pro’s content search covers you. AI search is for when phrasing and terminology diverge — common in large, formal, or public-facing libraries.

Does my document content get sent to an AI company?

Indexing and answering use the AI provider you connect with your own key, and the resulting index lives in your own database. You choose the provider; nothing is on unless you enable it.

Will it work on scanned documents?

Only after OCR — see scanned-PDF OCR. Once the text layer exists, both keyword and semantic search can use it.