Document Search Software: Two Different Problems, Two Different Tools

"Document search software" is really two different products wearing one label. The first kind finds a document — a contract, an invoice, a drawing, a report — when you know something about its name, project code, or date, even when it sits five folders deep on a shared drive nobody has reorganized since 2019. The second kind searches inside documents for a phrase you remember reading. The mechanics and the best tools for the two jobs are completely different, and most document-search frustration at work comes from using a tool built for one job to do the other.

Two problems hiding under one search box

The searches that lead people here start with a concrete pain: a contract is up for renewal and nobody remembers which subfolder of the department share holds it. An auditor wants invoice 4471 from March. An engineer needs revision C of a drawing that could be on the file server or in a colleague's project folder. Each of these is one of two distinct tasks:

The first job is computationally cheap. The names and paths of even ten million files add up to a modest dataset that fits comfortably in memory, so a well-built index answers name queries in milliseconds. The second job is expensive by construction: every file must be opened, its format parsed, its text extracted and either indexed or scanned on the spot. That cost difference explains which tools exist and where each one falls over.

How the mechanics actually differ

Name search: index the metadata, not the files

On a local NTFS drive, a name-search tool never needs to crawl folders one by one. NTFS keeps a Master File Table (MFT) — a single structure listing every file and folder on the volume — and reading it directly yields a complete name index in seconds. From then on, the NTFS USN change journal reports every create, rename, move, and delete, so the index stays current without rescanning. This is the mechanism behind the free Everything from voidtools, the reference tool for instant local filename search, and the same mechanism Malvicsearch uses locally.

Network shares are where most name-search tools stop being useful. SMB — the protocol behind UNC paths and mapped drives — does not expose the server's MFT to clients, so an index of a shared drive has to be built by enumerating directories over the network. Everything, per its own documentation, only handles network folders through periodic full rescans, which on a large share means results that are minutes or hours stale. Doing this well takes engineering most tools skip: parallel enumeration to build the index quickly, SMB change notifications to catch updates as they happen, and periodic re-sweeps as a safety net. More on the server-side picture in fast file search on a file server.

Content search: extract text, then index it — or scan on demand

Searching inside documents adds a text-extraction step for every file format. Windows Search does this locally through iFilters — format-specific plugins that pull text out of Word files, PDFs, and other document types so the words can be indexed. It works reasonably well for local folders, but the client-side Windows Search index does not cover network locations, which is exactly where most work documents live.

Two tool families fill that gap. Indexed full-text tools such as AnyTXT Searcher build and maintain their own content index, paying the extraction cost up front so later queries are fast. On-demand tools such as Agent Ransack (and its paid sibling FileLocator) keep no persistent index at all: each search opens and reads the files right then, including inside Office documents and PDFs, which makes them slow across a whole drive but dependable when scoped to one folder — and they work against network paths, because they are just reading files like any other program.

Note what the on-demand approach implies for shared drives: every search pulls the full content of every candidate file across the network. Scoped to a project folder with a few hundred documents, that is fine. Pointed at a 2 TB departmental share, it is a coffee break.

The scanned-document trap

One caveat catches nearly everyone: a scanned or photographed PDF is a picture of text, not text. It has no text layer, so no content-search tool — indexed or on-demand — will find anything inside it until it has been run through OCR (optical character recognition). If your invoices and signed contracts come in as scans, budget for an OCR step before content search can work at all. We go deeper on this in how to search PDF files.

Which tool for which situation

The honest answer to "what is the best document search software" is: it depends on what you are starting with. This table is how we would decide.

What you knowRight mechanismTools that do it well
Part of the file name, a project code, customer, or date — file could be anywhere on your PC or the sharesPersistent name/path index with real-time updatesMalvicsearch (local + network shares); Everything (local NTFS only, free)
A phrase inside documents stored on your local PCContent index with per-format text extractionWindows Search with iFilters; AnyTXT Searcher
A phrase inside documents on a network shareOn-demand content scan, scoped to a folderAgent Ransack / FileLocator
Text inside scanned or image-only PDFsOCR to add a text layer, then content searchOCR tooling first; then any content tool
Retention rules, versioning, approval workflows, permissionsDocument management system (DMS)A different product category — search alone will not get you there

In practice the two mechanisms combine well: use a name index to cut the search space — from millions of files down to the forty that match a project code — then run an on-demand content search against just that folder. Each tool does only the part it is fast at.

Malvicsearch: the find-the-document half, with numbers

Malvicsearch is a Windows 10/11 (64-bit) desktop app built for exactly one of the two jobs above: finding documents by name and path, instantly, across local drives and shared drives alike. It is deterministic substring search — you type, it filters, the same query always returns the same results. There is no AI, no semantic layer, and no natural-language interpretation involved.

The mechanics follow the design described earlier. Locally, the index is built from the NTFS MFT and kept current by the USN change journal, with no Administrator rights needed. For network locations — UNC paths and mapped drives on any SMB server, including NAS boxes, Samba, and Windows Server — a parallel scanner runs up to 64 concurrent directory enumerations in 256 KB batches to build the index, then keeps it fresh three ways: SMB change notifications (a median of about 2 seconds from change to visible result in our tests), a periodic re-sweep every 15 minutes by default, and a manual refresh menu. Nothing is installed on the server, which matters when the server is a locked-down NAS or someone else's responsibility.

The numbers below are measurements from our own workstation, not projections:

The index stays on your machine. Nothing is uploaded anywhere, and search works offline against the last-built index. The trial is free for 5 days with full features and no credit card; after that a licence is $29 one-time for 1 PC — no subscription — with a 14-day money-back guarantee.

What Malvicsearch does not do: it does not search inside file contents. No PDF text search, no Office-document text search, no OCR, and no AI or semantic matching — and it runs on Windows only. If your problem is "find the phrase," pair it with Agent Ransack or AnyTXT as described above; if your problem is "find the document," that is the entire product.

Frequently asked questions

Is Windows Search enough for documents on a shared drive?

Usually not. The Windows Search index on a client PC does not cover network locations, so searches against a share fall back to slow, live folder-walking. Windows Server can index its own shares for clients, but most NAS devices and Samba servers offer nothing equivalent. That gap is the main reason dedicated tools exist; see Windows Search alternatives for the fuller comparison.

What is the difference between document search software and a document management system?

A DMS (Document Locator, Docsvault, and similar) is a different category: documents are checked into a managed repository that adds versioning, permissions, approval workflows, and retention rules, and search is one feature among many. Desktop search tools work over the folders and shares you already have, with no migration. If the actual problem is "I cannot find files," a search tool over your existing structure is the far smaller change.

Can Malvicsearch find text inside my documents?

No, and it does not claim to. Malvicsearch searches file and folder names and paths only. For text inside documents, use a content tool: Agent Ransack for on-demand scans of a specific folder (including network paths), or AnyTXT for an indexed full-text search of local drives. The fastest workflow is usually both: narrow by name first, then content-search the handful of survivors.

Why does nothing find the text in my scanned PDFs?

Because a scan is an image. There is no machine-readable text in the file for any search tool to extract, indexed or otherwise. Running OCR adds an invisible text layer to the PDF, after which content-search tools can see it. Until then, the only thing software can match is the file's name — one practical reason to give scans meaningful names.

How fast is name search across millions of files, really?

Fast enough that the search box stops feeling like a query and starts feeling like a filter. On our workstation, Malvicsearch answered a query over an index of 9,363,175 files in 12 ms, and returned a 5.4-million-match result set across local and network drives in 56 ms. The mechanism is the reason: matching a substring against an in-memory list of names is simply not much work, provided the index exists and stays current.

Does indexing a shared drive require installing anything on the server?

Not with Malvicsearch — it builds its index from the client side over standard SMB, using parallel enumeration plus SMB change notifications, so it works with any NAS, Samba, or Windows Server share you can already open in Explorer. Details on how that holds up on large shares are in fast file search on a file server.

Stop hunting through folder trees. If your documents are scattered across a PC and a couple of shared drives, Malvicsearch will find them by name as fast as you can type. Download the free 5-day trial — full features, no credit card — or see pricing: $29 one-time, lifetime licence, 14-day money-back guarantee.