🪄 The Search Boxes That Hand You the File Instead of the Page

Tools & Scripts
,
1

:magic_wand: A page is a rumour. A file is a fact.

Normal search hands you pages about the thing. Ten tabs later you have read four reviews, closed two pop-ups, and still hold nothing.

Every box below answers with the file itself — something.pdf, something.zip, something.mp4 — and your browser starts saving it. No client to install. No account. No countdown page. No “upgrade for faster speed”.

Start here: paste your keyword into filepirates. One bar, and it asks Internet Archive, Wikimedia Commons, Project Gutenberg, GitHub Releases and Software Heritage at the same time, then merges and dedupes the answers. Five archives per keystroke. :pirate_flag:

🔎 Boxes that answer with files

The open-folder diggers — somebody left the directory listing switched on, and nobody optimises a folder they forgot to close:

  • FilePursuit — the broadest of them. Category tabs for video, audio, ebooks, apps and archives, and it sorts by date, which almost nobody uses.
  • ODCrawler — links harvested from inside open directories worldwide. Press / and type.
  • eyedex — 3.6 PB across 102 monitored servers, 2.9 million files.
  • Open Directory Search Tool · lumpysoft — same idea with a category picker in front.
  • FindFiles.net — PDF, DOCX, XLSX, ZIP, MP4, STL, CAD and datasets, and it honours filetype:, site: and size: inside its own box.

The old machines that never stopped running — FTP servers answering since the nineties, indexed by nobody, matched on filename rather than page text:

The archives with a proper search field:

  • Internet Archive advanced search — filter by mediatype and collection; every item page lists direct PDF, EPUB, MP3, MP4 and ZIP.
  • Zenodo — datasets, software and any file a researcher uploaded, downloading straight off the record page.
  • CORE — 357 million records, direct PDFs, no login anywhere.
⌨️ When no box has it — bend an ordinary engine into one

Copy, paste, replace the words in capitals:

intitle:"index of" "parent directory" YOURTERM     → opens raw file browsers
intitle:"index of" YOURTERM (mp4|mkv|zip|pdf)      → a folder full of one format
filetype:pdf YOURTERM                              → pins the format (Google, Startpage)
ext:iso YOURTERM                                   → same idea, different keyword
inurl:uploads filetype:pdf YOURTERM                → upload and backup folders

The 2026 catch — the operators are no longer the same everywhere:

Google / Startpage   filetype:  ext:  inurl:  intitle:   ← the full set still works
Bing                 filetype:  contains:mp3             ← no inurl: since 2007
                     contains: finds PAGES LINKING to that file type — quietly powerful
Yandex               mime:pdf  host:  title:             ← mime: replaces filetype:
DuckDuckGo           filetype: office formats only       ← patchy, treat as a bonus

The full engine-by-engine table lives in the 2026 operator reference, and the Google Hacking Database is the old public collection of ready-made queries, still updated.

🛠️ When you would rather own the index than borrow one
  • Sosse — a self-hosted crawler and search engine with batch file download built in. One docker run, point it at the sites you care about, and you have a private file index nobody can take away.
  • Webspider — crawls a site, verifies which links are really files, and writes a list you feed straight to wget -i urls.txt.
  • pagodo + gdorking — pull the live query database, run it against a target, keep every hit URL in a file.
  • Batch File Downloader (Chrome) · Universal Bulk File Collector (userscript) — when a page holds forty files, one click queues all forty, including the lazy-loaded ones.

One habit worth stealing: when a search fails, change what you are asking for. Not the topic — the format. The thing you want often exists as a zip nobody named, in a folder nobody closed. :old_key: