How it works
Updated August 12, 2026 · min read
Exact Duplicates vs Similar Photos: What Can Be Deleted Safely?
Understand why deterministic exact matches and visual similarity scores require different cleanup rules.
A burst sequence, edited portrait, messaging-app copy, and original camera file can all look nearly the same in a thumbnail grid. Their storage and archival value can be very different, and visual similarity does not establish redundancy.
Run two clearly separated workflows. Use deterministic fingerprints for exact copies, then—only if needed—use a visual tool with a stated threshold for similar images. Never blend the outputs into one bulk-removal recommendation.
A practical checklist
- 1Group exact images by size and fingerprint.
- 2Label visual-similarity results separately.
- 3Compare resolution, metadata, quality, and edit history.
- 4Check albums, links, and shared audiences.
- 5Approve exact and similar-image decisions under different policies.
Understand the situation
A burst sequence, edited portrait, messaging-app copy, and original camera file can all look nearly the same in a thumbnail grid. Their storage and archival value can be very different, and visual similarity does not establish redundancy.
Keep detection evidence separate from retention judgment. Technical signals can establish that bytes match, while path, permissions, history, and ownership determine whether one file item is operationally redundant.
Use a controlled workflow
Run two clearly separated workflows. Use deterministic fingerprints for exact copies, then—only if needed—use a visual tool with a stated threshold for similar images. Never blend the outputs into one bulk-removal recommendation.
A robust pipeline handles pagination, missing values, throttling, and changes during the scan. Incomplete data should lower confidence or remain unresolved instead of being converted into a clean-looking but unsafe result.
- Label visual-similarity results separately.
- Compare resolution, metadata, quality, and edit history.
- Check albums, links, and shared audiences.
Make the final decision
For exact groups, choose the keeper from album and sharing context. For similar sets, compare resolution, focus, expression, edit intent, capture metadata, and backup policy one set at a time. A good visual result can keep several images.
ClearCopies has a deliberately narrow promise: scan supported cloud metadata, confirm exact candidates by byte size plus a matching provider fingerprint, let the user review, and export the plan without downloading original content.
Limits and risks to check
- — A perceptual score is a threshold, not proof of equality.
- — Thumbnails hide quality and metadata differences.
- — Originals and edited exports often serve different purposes.
- — ClearCopies does not currently perform visual similarity analysis.
Official references
Frequently asked questions
Are burst photos duplicates?
Usually not exact duplicates; each frame can capture a different moment even when thumbnails look alike.
Can metadata-only scanning find cropped images?
Not as exact copies because cropping changes the bytes and fingerprint.
Which method is safer for bulk review?
Exact fingerprint matching is deterministic. Similarity groups should receive slower, visual review.
Scan first. Decide with evidence.
ClearCopies reads supported cloud-drive metadata and groups exact copies by byte size plus a provider-supplied content fingerprint. Original file bodies are not downloaded for the scan. You review the result and export a plan before any separate write step.