Methods and corrections

How we test camera shutter-count evidence

A useful result needs an identifiable file, a documented reader, and a precise claim. Here is how our research separates those pieces and what stronger verification would require.

By ShutterActuations · Updated September 15, 2026

What we have actually tested

Our September 2026 corpus run covered 1,870 CC0 RAW files from raw.pixls.us, totaling 58.45 GB and representing 931 unique camera-model labels across 85 manufacturers. The browser metadata engine parsed 1,832 files. Strongly named counter candidates appeared in 527 files. These are measurements from a public sample collection, not a random survey of all cameras in use.

Each model may have only one or a few samples, and several samples can come from the same body. Firmware, shooting modes, file processing history, and independent service counts are not uniformly known. Do not turn a sample detection percentage into a probability that an arbitrary used camera will work.

Our extraction procedure

1. Identify the input. Record the source catalog entry, license, filename, byte length, and SHA-256 fingerprint. Use an original rather than a rendered preview.
2. Record software. The recorded corpus used application commit ade8deab1cd63cf3201059330a3835c2d590f30c and @uswriting/exiftool 1.0.9. That number identifies the wrapper package, not the underlying ExifTool release.
3. Extract locally. The worker requests -json -G1 -a -s -n: JSON output, group names, duplicate tags, short tag names, and numeric values without print conversion. This preserves distinctions that a friendly label can obscure.
4. Separate engine and product. The research harness can read a format that the public file validator rejects. The published format table reports those two results separately.
5. Keep negative findings. A successfully parsed file without a counter is recorded alongside positive examples. Failure, absence, a possible candidate, and a strongly named counter are different outcomes.
6. Publish a limited extract. The selected public records contain provenance, camera model, counter fields, and evidence level. They omit image pixels, local paths, serials, GPS, owner, and authorship metadata.

Download the selected sanitized evidence records (JSON). The underlying originals are external downloads and may still contain sensitive information. The public extract is not a full copy of their EXIF.

Four levels of evidence

Zero corpus values currently qualify as VERIFIED_ABSOLUTE. Agreement with another ExifTool-based website may repeat the same interpretation and is not an independent service baseline. A lower-level finding is still useful when its limits are explicit.

A controlled capture experiment an owner can prepare

This is a proposed ground-truth protocol, not a claim that these experiments have already been completed. Use only your own camera or one you have permission to test. Begin with an accepted original NEF for a public-checker experiment; other formats remain separate research work.

1. Record model, firmware, shutter mode, drive mode, and any existing authoritative diagnostic. Keep the physical serial private.
2. Disable location recording and clear optional owner/comment fields before capture where your camera permits it. A RAW may still contain an internal serial.
3. Photograph a neutral test card. Avoid people, correspondence, addresses, reflections, and location clues. Keep the original file as baseline A.
4. Make exactly five ordinary single-frame exposures in the same mode. Use the fifth file as B. The planned A-to-B interval is five exposures; taking a separate extra “after” photograph would change that interval.
5. Record the exact tag values from A and B. Subtract A from B for each field individually. A delta of five supports behavior for this sequence, not the accuracy of the lifetime starting value.
6. Run a separate sequence for another supported shutter mode. Keep burst, live view, electronic-front-curtain, fully electronic, and computational modes distinct. Do not mix modes and then interpret one combined delta.
7. Repeat the experiment and retain unexpected results. A second independently owned body strengthens a model-level conclusion.

For example, a hypothetical field changing from 10,000 to 10,005 after five documented exposures is consistent with that sequence. It does not prove the camera has made exactly 10,005 mechanical actuations over its life. We would still need a reliable independent baseline and mode semantics.

Research sharing requires a separate decision

Selecting a photo in the checker does not submit it to us. There is currently no public research-upload form on this site. Start with a text-only compatibility report. If a sample is needed, agree on the purpose, recipient, sharing permission, and handling before sending it. Emailing an original is a transfer of the file, unlike the local checker.

Keep private originals intact for reproducibility. Prepare separate redacted screenshots or a small manually selected result table for discussion. Do not edit a RAW to remove metadata and then describe it as the untouched test original. Do not publish full camera serials or hashes of serial numbers.

Corrections and updates

A useful correction identifies the page, disputed sentence, sample record or exact model, source tag, and a source supporting the change. We distinguish a documentation correction from a new compatibility claim. New claims require their own evidence and product checks; one successful research sample does not automatically enable a file extension.

Published examples name the observed software and date. Future parser revisions can expose different fields. When a conclusion changes, the article should identify the affected scope and update its modification date rather than silently presenting old and new results as the same experiment.

Reading the research

Begin with the D2H worked case study, then compare the D100 negative result and Sony A1 disagreement in the sample evidence section. ExifTool’s Nikon and Sony tag references document manufacturer-specific structures; they do not certify the history of any individual camera.