Does AI Culling Keep a Record of Every Decision It Makes?

What actually gets recorded

Two kinds of entries make up the record. First, the AI's own decision on every frame: which pile it landed in and, implicitly, the confidence behind that call. Second, your corrections: any time you reverse a decision, the reversal is logged along with the reason you supply, such as closed eyes, missed focus, or better expression. Together, the two form a running history of what the software decided and what you changed, for every frame in the project, not just a summary count at the end.

Why this matters beyond curiosity

An audit trail is not just a technical nicety. If a client asks about a specific frame months later, or you want to understand why a particular wedding's cull felt off, the record lets you reconstruct exactly what happened rather than guessing from memory. It also protects you from your own inconsistency: if you notice you reversed the same kind of call several different ways across a season, the log is how you would find that out.

A concrete example

Months after delivering a wedding gallery, a couple asks whether a specific candid of their officiant exists anywhere; they remember it being taken but do not see it in the gallery. Because the project keeps a record of every decision, you can look up that frame, see that it was originally sorted to REVIEW on a low confidence score and that you later rejected it with a reason logged, closed eyes, and make an informed call about whether to revisit it, instead of relying on memory of a cull from months ago.

How the record protects the files themselves

• Rotating backups. The last ten states of a project are backed up before any sorting operation moves files, so a bad run is recoverable.
• Cumulative organize manifest. Every sort operation is tracked in a manifest, which is what makes undo possible after the fact.
• Decision and correction log. Every keep, reject, review, and correction, with its reason, sits in the project database rather than only in the folder structure on disk.

Honest limitations

A record of decisions is not the same as a record of intent. The log tells you what happened and, when you gave one, why you corrected it, but it does not capture context you never wrote down, such as why a particular frame mattered to the couple. It is also specific to a project; it is not a searchable archive across your entire career unless you keep your project files. Treat it as a reliable record of one wedding's cull, not a universal photo database.

How Cull AI Studio handles this

Every decision Cull AI Studio makes and every correction you enter, along with the reason you give, is written to the project database as part of a full audit trail. Before any sort operation moves files, a rotating backup of the last ten project states is taken, and a cumulative manifest tracks changes so the whole operation can be undone. This is separate from what happens when you catch and fix a specific mistake, which is covered in what happens when AI makes the wrong selection; this record-keeping is the underlying feature that makes that response possible in the first place.