You remember that a supplementary table had "raw_v2" in its file name, but not which folder it is in. Windows search spins for a while; Everything shows results almost as you type. It is not magic. It does one thing, and it takes a shortcut.

Status of this article: the explanation is based on voidtools' official documentation. Everything has not yet been tested on Windows for this article, so there are no search timings, screenshots or comparison results. Section 5 is a test plan you can repeat on your own computer.

1. Short answer

If you remember part of a file name, Everything is a good fit. If you need to find what is written inside a file, it is not the specialised tool, and you should consider Windows content indexing or dedicated full-text search software.

2. Why it is fast

According to the official FAQ, Everything indexes only file and folder names, not contents. Its reference figures are about 1 second to index a fresh Windows 10 install of about 120,000 files and about 1 minute for 1,000,000 files, using about 14 MB and 75 MB of memory respectively. These are vendor figures, and your computer may differ.

The speed has two layers:

First, the scope is small. There is far less text in file names than in file contents, and no files need to be opened or parsed.

Second, it does not walk every folder. An NTFS volume has a Master File Table (MFT) recording all files and folders. Reading that table is much more direct than descending through the directory tree.

Once the index exists, files keep changing. The official documentation says Everything keeps NTFS indexes current through the NTFS USN Journal. The journal records changes on the volume and is maintained by the system, so according to the FAQ changes made while Everything is not running are not missed and are picked up at the next start.

Think of Windows content indexing as a full-text index of every book, and Everything as reading only the library catalogue while subscribing to its "new and withdrawn titles" notices.

3. Name search and content search

File name search content: search
Looks at Names and paths Text inside files
Indexed in advance Yes No (per official FAQ)
Expected speed Fast Slow (per official FAQ)
Advice Use directly Narrow by name or extension first

To find PDFs containing "calibration", restrict the extension and then add the content condition:

ext:pdf content:calibration

This has to read the files, and whether text inside a PDF can be recognised also depends on the file format and system, so test it yourself. Scanned PDFs have no text layer and need OCR first.

4. Useful search patterns

Goal Syntax Note
Both words raw v2 A space means AND
Either word xlsx|csv A vertical bar means OR
Exclude report !draft An exclamation mark excludes
File type ext:tif;tiff Separate extensions with semicolons
Limit path D:\data\ raw Path plus keyword
By size size:>100mb Find large files
By date modified dm:today Files changed today

Check the help page of your installed version (Help > Search Syntax) for the details.

5. A repeatable test plan

This Python script uses only the standard library. It creates a test folder with a known number of files and writes a manifest, with unique markers in the names for exact and partial matching. The generator has been run on Linux and has not yet been run on Windows.

import csv, random
from pathlib import Path

ROOT = Path("everything_test")   # place on the disk under test
N_DIRS, N_PER_DIR = 50, 200      # 10000 files in total
EXTS = [".txt", ".csv", ".pdf", ".tif", ".xlsx"]
random.seed(2026)
rows = []
for d in range(N_DIRS):
    folder = ROOT / f"project_{d:03d}"
    folder.mkdir(parents=True, exist_ok=True)
    for i in range(N_PER_DIR):
        name = f"zlt_{d:03d}_{i:04d}_{random.randint(0, 999999):06d}{random.choice(EXTS)}"
        (folder / name).write_bytes(b"demo")
        rows.append([str(folder / name)])
with open("everything_manifest.csv", "w", newline="", encoding="utf-8") as f:
    csv.writer(f).writerows([["path"]] + rows)
print(len(rows), "files")

Record the test conditions (each must be filled in truthfully):

  • Windows version and build
  • Everything version
  • Disk type and file system (NVMe / SATA / HDD, NTFS etc.)
  • Total files on the volume
  • Windows search index scope
  • Running as a service?

Steps (repeat each 5 times and record the median; for manual timing, state the method and precision):

  1. Exact name: search for 10 complete names picked at random from the manifest, and confirm each result is unique.
  2. Partial name: search zlt_017_ and expect 200 results.
  3. Type filter: the count for zlt_ ext:csv should match the manifest.
  4. Create: create zlt_new_001.txt and record the time until it appears.
  5. Rename: confirm the old name disappears and the new one appears.
  6. Delete: confirm it disappears from the results.
  7. Quit and restart: quit Everything, add a file, restart and confirm it can be found.
  8. Comparison: run the same queries in Windows search, recording whether the folder is in the index scope.

For a fair comparison, record indexed and non-indexed locations separately, because Windows search may behave very differently between them.

6. Permissions, scope and other limits

  • Cost and licence: the official FAQ calls it freeware and accepts donations. Free does not mean open source, and this article does not treat it as open source.
  • Platform: Windows only.
  • Permissions: the official FAQ says NTFS indexing requires the Everything service or running as administrator. If the results are empty or show only drives, check this first.
  • Non-NTFS and network locations: FAT, FAT32 and exFAT volumes and network drives can be added as folder indexes under Tools > Options > Folders, but they do not use the fast NTFS route.
  • Duplicate results: the FAQ warns that also adding an NTFS volume as a folder index produces duplicates.
  • Shared computers: check first whether indexed results could show file names to other accounts.

7. Who this is for

Suitable for: Windows users with many files who remember part of a name, such as people organising experimental data, micrographs and manuscript versions.

Not suitable for: people who mainly need full-text or semantic search, or who work mainly on macOS or Linux.

If your question is "does this PDF contain that word?", see ZIRULAB's article on PDF tools.