The JSON is metadata, not another photo
A Takeout folder can contain an image such as IMG_0001.jpg and a small JSON document associated with it. The JSON is structured text that records information held by Google Photos; deleting it does not delete the image, but it can remove the best remaining evidence for edits made in the service.
Google distinguishes metadata already embedded in original media from additional information exported in a secondary JSON file. A repair should inspect both sources rather than assuming every image has been stripped.
Useful media fields
Commonly useful fields include photoTakenTime, creationTime, geoData, geoDataExif, title, description and mobile upload origin or device information. Not every export or item contains every field, and a zero coordinate is not proof of a real location.
The JSON timestamp is commonly an absolute Unix time. Converting it into EXIF wall time without a reliable timezone offset can create a subtle but visible shift.
Album JSON is a different record
Album folders may contain a metadata JSON that names or describes the album rather than a single media item. Treating every JSON document as a photo sidecar creates false filename matches and loses the distinction between media metadata and album relationships.
The same physical photo can appear in a year folder and several album folders. A library-level index can merge those relationships after confirming exact content.
Why filenames do not always line up
Long names may be truncated, copies can receive numeric suffixes, edited media can change names and supplemental files can use full or shortened suffixes. A sidecar can also be placed in a different numbered ZIP from its media.
A reliable matcher uses ordered evidence: exact local paths, recognized suffix variants, the title recorded inside JSON, content context and conservative cross-folder rules. A common basename alone is not enough when thousands of camera files reuse names.
Do not turn uncertainty into a silent match
If two media files are equally plausible for one sidecar, assigning the metadata to either can corrupt a date or location. The correct result is an ambiguous category in the report with enough source paths to investigate.
Takeout Rescue leaves unresolved candidates visible and preserves the source archive so improved rules can be applied later.
When JSON can be removed
Keep the original ZIPs permanently if practical. JSON files may be omitted from a clean output only after supported dates, coordinates, descriptions and album relationships have been written or otherwise preserved and representative output has been verified.
The output report should retain a mapping back to the original sidecar even when the portable library no longer needs thousands of Google-specific JSON files beside the media.