How to use AI to clean up your photos (for real)
Tags, faces, places and duplicates: a review of what AI sorting can actually do — and what still needs your hands.
The photo library is the classic place where “AI organizing” gets sold as magic. The reality is more nuanced: some things AI handles today, others you have to do yourself — and knowing which is far faster than trying both.
What AI can actually do
- Duplicate finder. Identical and near-identical photos are found reliably, and that is the biggest space saving.
- Face tagging. Recognizes people across years, once you have named them a few times.
- Coarse categorization. “Nature”, “documents”, “screens”, “people” — most of it gets classified correctly.
- Places. GPS tagging already exists in the metadata; AI guesses the location of photos without geodata, with middling success.
What AI can’t do
Context: “this cover photo matters for project X”, “this is the only photo of my grandmother’s sewing machine”. Those judgments need you — but going through an already sorted collection is far faster.
The process that worked
- Run the duplicate pass first and remove what it finds.
- Face tagging, and name the most important people.
- Sort by category, and make your own folders for projects and events.
- Never delete automatically — just move things to a “maybe” folder.
The result is not a finished archive, but a library manageable enough that you can find what you are looking for. That is the whole point.
