How to find duplicate photos on iPhone
iOS has had a Duplicates album since iOS 16, and I would guess you have never opened it. It would not have solved your problem anyway, because it only catches exact copies.
The album Apple already gives you
Photos → Albums → Utilities → Duplicates. If it shows up there, start with it — it costs nothing, it is already on your phone, and its merge keeps the best-quality copy along with the metadata from every version it found.
Two things I wish someone had told me sooner. It only appears once iOS has actually found duplicates, so an empty Utilities section means none were detected, not that the feature itself is missing. And on a large library it can take days to populate after an update, because the detection only runs while the phone is locked and charging.
Why it finds so little
Apple's album is deliberately conservative. It is only looking for photos that are the same file — the same image, usually from a save, an AirDrop, or an app quietly writing a second copy. That is a real problem, and also a small one.
The real problem — the one actually filling your camera roll — is near-duplicates. Ten frames of the same moment because you held the shutter down. Three attempts at the same group photo. The same sunset, eight seconds apart, again and again. None of those are the same file, so none of them show up in Duplicates, and they are the overwhelming majority of what is actually filling your phone.
Finding near-duplicates by hand
It is tedious, it costs nothing, and it actually works:
- Photos → Years, then pick a month you remember taking a lot of photos in.
- Scroll to any burst — they sit together because they share a timestamp.
- Tap Select, then drag across the ones you do not want.
- Keep the sharpest one. Zoom in on a face or an edge; the winner is usually obvious at 100% and invisible in a grid.
Sort by date taken, never by album. Bursts sit next to each other in time and get scattered everywhere else.
What automated detection actually measures
Tools that find near-duplicates compute what is called a perceptual hash — a short fingerprint of the image's structure, not its actual bytes. Two photos of the same scene end up with similar fingerprints even at different exposures or crops, so the tool can measure how far apart any two fingerprints are and group anything close together.
The threshold matters more than the technique. Set it loose and it groups two different photos of the same room and offers to delete the one you actually wanted. There is no undo worth trusting for that, so a good tool errs strict: it misses a few real duplicates rather than catching everything and occasionally being wrong.
Which one to keep
The signal that actually matters is sharpness, because it can be measured, not just eyeballed: run an edge detector over the image and look at how much variation it produces. A crisp photo produces a lot of it; a blurred one is smooth. That is why a tool can pick the best frame of a burst more reliably than you can in a thumbnail grid — you are squinting at a 120-pixel preview and it is looking at the full image.
Before you delete anything
Empty Recently Deleted afterward, or none of that space comes back for thirty days. And if iCloud Photos is on, deleting here deletes everywhere — which is usually what you want, and always worth knowing before you tap anything.
Sift — Clean up photo storage. See what it does →