Every long-lived collection of files converges on the same state: the same photo existing as the original, the export, the copy someone mailed back, and the version a messaging app re-compressed. Duplicates are not a hygiene failure - they are what normal file handling produces over time. Clearing them safely is mostly about knowing what actually counts as a duplicate.
Where they breed
The Downloads folder. Re-downloading is easier than finding, so
invoice.pdf, invoice (1).pdf and invoice (2).pdf accumulate as a
matter of course.
Photo exports and edits. Every “export for web”, every edit saved as a copy, every share that round-tripped through another app.
Messaging apps. Received images saved to the camera roll, then backed up alongside the originals they duplicate - often recompressed, so not even byte-identical.
Migrations and merges. Consolidating an old laptop, an external drive and a cloud account into one place imports each collection’s copy of the same files. The cleanup pass in our migration guide exists precisely because moving is when duplication peaks.
Defensive copying. budget_final.xlsx, budget_final_v2.xlsx,
Copy of budget_final.xlsx - versions kept by hand out of fear of exactly
the overwrites that version history
handles automatically. Once versioning is on, this habit can retire, and its
back catalogue is deletable.
Why filenames cannot be trusted
Name matching fails in both directions. The same contents hide behind
different names (IMG_4021.jpg and beach.jpg), and different contents
share a name (every folder has a notes.txt). Size matching narrows things
but collides constantly at small sizes.
The honest test is a checksum: a fingerprint computed from the contents alone. Identical checksum, identical file - whatever the names say. Different checksum, different file - even if they look the same in a thumbnail. Our checksums guide covers computing them by hand; duplicate finders are essentially tools that do so at scale and group the matches.
One boundary worth knowing: a recompressed image - the WhatsApp version of your photo - is genuinely different bytes, so checksum tools correctly do not call it a duplicate. Near-duplicate detection is a separate, fuzzier problem; solve the exact duplicates first, which is most of the volume anyway.
Finding them at scale
On a local machine, the established tools do the checksum-and-group work:
fdupes or jdupes on the command line, dupeGuru if you want a window. Point
them at one tree at a time and read before deleting.
For files already in dosya, there is an opt-in duplicate scan: switch it on and it compares checksums across your folders and presents the matches as groups - each group one set of identical files, wherever they live and whatever they are called. It is off by default and scoped to folders you choose, so nothing happens to your storage until you ask.
Deleting without regrets
The finding is mechanical; the deleting deserves rules.
Keep the copy in the organised location. When the same file exists in
Photos/2023/iceland/ and Downloads/, the Downloads copy is the duplicate,
whatever their timestamps say.
Never bulk-delete both sides of a match. Automation should propose; choosing which of a pair dies is the human step, and any tool that auto-picks deserves suspicion.
Let the trash be the safety net. Deletions land in a recoverable trash with a retention window, so a wrong call inside that window is an undo, not a loss. Do the big cleanup, then let a few weeks pass before thinking about the trash at all.
Fix the source, or repeat annually. The Downloads habit, the by-hand version copies, the messaging re-saves - each has a structural fix (search instead of re-download, versioning instead of copies, app auto-save settings). Skip the fixes and the cleanup becomes a subscription.
The payoff, honestly stated
The space matters less than people expect - duplicates are typically a
single-digit percentage of a collection, worth checking against
how much storage you actually need
before expecting miracles. The real gain is trust: one copy per file means
search results you can act on, backups that are not re-protecting the same
bytes four times, and never again wondering which budget_final is the real
one. That is worth an afternoon, once, with the safety nets on.
Frequently asked questions
How do I find duplicate files reliably?
Compare checksums, not names. Identical checksums mean identical contents whatever the files are called. Tools like fdupes and dupeGuru do this locally, and dosya offers an opt-in scan that groups checksum matches across your folders.
Why do I have so many duplicate photos?
Exports, edits saved as copies, messaging apps re-saving received images, and migrations that merged several collections. It is the normal by-product of files being handled, not a mistake.
Is it safe to delete duplicate files?
Yes, with two rules: keep the copy in the organised location, and never let a tool auto-delete both sides of a match. A recoverable trash gives you an undo window for wrong calls.
