Every duplicate finder offers two kinds of results, and mixing them up is how people lose real photos. Exact duplicates are identical files. Similar photos are different photos that happen to look close. Only one of these lists belongs anywhere near a bulk delete button.
Two photos can look the same to your eyes and still be completely different files. A burst sequence, a retake a second later, a version you brightened in an editor - all of these look almost identical but none of them are duplicates. A real duplicate is the same file copied more than once: same pixels, same size, same fingerprint. Duplicate finders know the difference because they read the file, not the picture. In exact mode they compute a hash - a short code calculated from every byte of the file - and only files with matching hashes are grouped. This page is the honest split between the two lists, because deleting by similarity is the single most dangerous action in photo cleanup, and deleting exact matches is one of the safest.
A hash is calculated from the full contents of a file. Change even one pixel and the hash changes completely. That is why hash matching is so strict: two files with the same hash are the same photo in every byte, including the hidden camera details saved inside it.
This strictness is a feature, not a limit. It means the exact match list has no opinions and no guesses in it. If the tool says these four files are one group, they are genuinely four copies of one photo, and keeping any single one of them keeps the photo.
Similarity mode works like a person glancing at thumbnails. The tool shrinks each photo to a tiny signature of its shapes and colours and groups photos whose signatures are close. That is useful for finding retakes, but it cannot tell a copy from a moment.
None of these should ever be swept away by a select all button.
Run your cleanup in two passes that never mix. Pass one: exact match only. Review the groups, delete the spare copies to the Recycle Bin or a staging folder, and finish the job completely. Pass two, only if you want it: similarity view mode. Walk through similar groups one at a time, on a big screen, and choose keepers with your own eyes like you would sorting printed photos.
And the standing rule from the core guide still applies: backup first, review before delete, and stage deletes for a week. Similarity pass deletions deserve the staging folder most of all, because those files exist nowhere else.
The full cleanup, from the first scan to the final delete, is shown live in the video.
Watch Full VideoFor the strong hashes used by photo tools, no - a matching hash means the files are identical byte for byte. That is exactly why exact match mode is the safe list for bulk cleanup, and why similarity scores, which are estimates, are not.
Because they are not identical files. They may be taken a second apart, re-saved by an app, or resized. The file contents differ, so the hashes differ. Those photos belong to the similarity list, where you choose keepers by eye, never in bulk.
Yes, as a viewing helper for retakes and burst groups when storage is tight - but delete from it one group at a time, by eye, into the Recycle Bin or a staging folder, and keep the staging for at least a week. Never use its select all or auto delete options.