You open one Premiere project, run the cleanup, delete what looks unused, and feel good about the gigabytes you reclaimed. Two weeks later you open a different project on the same drive and half the timeline is red. You just deleted footage that was the hero shot somewhere else.
This is the quiet failure mode of every single-project cleanup workflow. And it's why the "is this file safe to delete" question is a trick question if you're only looking at one project at a time.
The One-Project Blind Spot
Most cleanup tools, including Premiere's own Project Manager, treat each .prproj file as a closed universe. The logic is simple. If a clip isn't on a timeline in this project, it's unused. Delete it.
That logic only holds if every project on the drive lives in its own walled garden, with its own dedicated footage. Almost nobody works that way. Editors share footage across projects constantly. A logo, a stock plate, a B-roll select, a sound effect, a generic establishing shot. The file lives in one place on disk and gets imported into whatever project needs it.
When you only check Project A, you miss every reference Project B, C, and D have to that same file. Single-project cleanup is structurally blind to the actual usage of any given asset.
Scenarios Where This Hurts
The cross-project problem isn't theoretical. A few situations where it bites editors regularly:
The recurring client. You shoot a corporate explainer in January. In March the same client comes back for a 30-second cutdown that pulls from the same footage library. You finish the cutdown, archive it, and a month later you "clean up" the January project. The cutdown's hero shot was a clip you only used in the cutdown, so to the January project, it looked unused. Now the cutdown is offline and the client wants a revision.
The shared B-roll bin. A doc editor keeps a folder called _BROLL_LIBRARY/ and imports from it into every project. Each project only uses a handful of the clips. From any single project's perspective, ninety percent of the library looks unused. Delete based on one project and you just gutted the library for every other project that depended on it.
The reused stock plate. That sky replacement plate you bought once and have dropped into six commercials? It only appears on one timeline per project. Run a cleanup that doesn't see across projects and you'll mark it deletable in five of them.
The orphaned project file. Sometimes a project lives in a different folder than its media. A director's cut, a festival version, a vendor-delivered edit. The media drive sees no project pointing at certain files. But there's a .prproj sitting on a thumb drive in a desk somewhere that depends on them.
In every case, the file is "unused" by exactly the definition the cleanup tool understands, and "used" by the definition that actually matters when a client calls.
Why This Maps to Real Filmmaking
Think of your media drive like a rental house. Single-project cleanup is the equivalent of walking into the rental house, looking only at the gear list for one shoot, and throwing out anything not on it. The fact that twelve other shoots also pull from that same inventory doesn't enter the picture.
A real rental house tracks gear across every active job. Nothing leaves the cage until every job that needs it is wrapped. That's the only safe model. Anything less is just guessing with expensive consequences.
What Actually Safe Cleanup Looks Like
A cleanup pass is only safe if it answers a different question. Not "is this file used in the current project," but "is this file used in any project on this drive, anywhere." The unit of analysis has to be the drive, not the .prproj.
Practically, that means walking the SubClip to MasterClip to Media reference chain in every project file on disk, building a master index of which files are referenced by which timelines, and only flagging files that show up in zero references across the whole index. Anything else is unsafe by definition.
This is also the only way to confidently estimate reclaimable space. A per-project estimate is always inflated because it counts shared assets multiple times. A drive-level analysis gives you the actual number. The calculator is useful for translating that real number into monthly cloud or drive cost, but the input has to come from a tool that sees the whole drive.
The Habit to Build
If you take one thing away from this: stop cleaning up projects. Start cleaning up drives. The .prproj is the wrong scope. The drive, or the shared volume, is the right scope, because that's where the references and the files actually live together.
When you're tempted to right-click Remove Unused in a single project and start deleting, remember that you're answering the wrong question. The right question is the one this blog keeps coming back to: what's truly unused everywhere, and what's quietly load-bearing somewhere you forgot.
A Drive-Wide View
This is the specific gap Clip Sweeper was built to close. It indexes every project on a drive at once and only marks files that no timeline anywhere depends on. Before reclaiming space the manual way, it's worth seeing what a drive-wide view actually shows.
The reclaim is real either way. The difference is whether you keep it, or spend the next two weeks relinking media and apologizing to clients. Run the cleanup at the right scope and the savings stick.