フォルダの自動振り分け: how to decide what you need

Comparison pages for automatic folder sorting all end up looking the same. Name, extension, date, move, rename, tag. Read three of them and the lists blur together, because the lists are genuinely similar. The parts that decide whether a setup survives six months are not on those lists at all. They show up the first time a rule misfires, the first time a folder grows past a limit, and the first time macOS quietly revokes a permission after an update.

What follows is a way to compare these tools on the things that differ. Four questions, in the order they usually bite.

The four questions that separate these tools

Every sorting setup, from a built in automation to a paid rule engine, is made of the same four parts. What starts a run. What a rule is able to look at. What happens when a run does the wrong thing. And where the whole thing runs, with what permission.

Picking a tool means deciding which of the four matters most for the files at hand. Someone sorting invoices that arrive with identical filenames every month has a different weak point than someone sorting screenshots and video exports. Nobody needs the tool that scores well on all four. They need the one that does not fail on the question they cannot afford to lose.

A useful first move is to name the failure that would actually hurt. Files silently overwritten. Files moved somewhere unfindable. Rules that stop firing without telling anyone. That answer usually points at one of the four sections below, and the rest of the comparison gets much shorter.

Question 1: What starts the run

There are three ways a sorting run begins. The moment a file lands, on a schedule, or when a person asks for it.

Immediate triggers feel best and break in the most interesting way. A browser writes a partial file under a temporary name while the transfer is still running. A rule watching for new items sees that partial file, fails to match it on extension, and never sees it again, because the moment the download finishes is a rename rather than an addition. That single behaviour explains a large share of the reports where only part of a folder ever gets sorted.

In macOS 26, Shortcuts offers three folder triggers: when files are added to a specified folder, when files are modified in a specified folder, and when files are removed from a specified folder. There is a separate setting for whether subfolders are ignored. There is also a choice between running immediately and asking first, which is worth checking before concluding that an automation is broken. An automation set to ask sits behind a notification until someone approves it.

Scheduled runs trade speed for reliability. Nothing is half written an hour later, so the partial download problem disappears. A pattern that works well is two runs a day, one shortly after the usual end of work and one before it starts again, so the machine is never moving files under someone's hands.

Rolling a custom script is the third path, and the built in scheduling documentation is blunt about watching the filesystem directly: the practice is described as race prone, with no guarantee that a file is in a consistent state when the job fires. That is a strong hint to prefer a schedule over a watcher when writing the logic by hand.

Question 2: How much the rule can see

Name and extension are available everywhere. Creation and modification dates are close to universal. Three things separate the tools.

The first is provenance. A file downloaded by a browser carries a record of where it came from, stored alongside it. A rule that can read that can sort by originating site, which matters when a supplier names every invoice differently but always serves it from the same domain. Without it, the same job has to be approximated with partial name matches, and partial name matches miss.

The second is content. Reading text inside a PDF, or text recognised inside an image, turns "sort the scans" from impossible into routine. This is where paid tools tend to earn their price.

The third is how conditions combine, and specifically the order in which rules are evaluated. Some engines stop at the first match. Some run every rule against every file. A broad rule placed near the top of a stop at first match list silently disables everything under it. Confirm this behaviour before building anything, because a rule set built on the wrong assumption looks correct and behaves wrong.

With the built in Shortcuts route, the filtering is written by hand. The automation receives a summary of what changed, and the shortcut loops over the items and decides what to do with each one. Flexible, and slower to build as the conditions multiply. A paid engine such as Hazel 6 from Noodlesoft presents those conditions as a list to assemble. It is a one time purchase at $42, with a five member household licence at $65 and a $20 upgrade from an earlier version.

Question 3: What happens when it gets it wrong

A sorting rule that moves a file to the wrong place is not the problem. Not being able to find out where it went is the problem.

Three things determine whether a mistake is recoverable. Whether a record exists of what moved, from where, to where, and when. Whether that record can be read outside the app, as plain text, so it can be searched and reversed in bulk. And what the tool does when a file with the same name already exists at the destination.

That last one deserves attention because the Finder itself sets the expectation. Apple's own guidance on merging folders notes the available choices when names collide:

If the folders contain different versions of identically named files, the only options are Stop or Replace. Source: support.apple.com

Sorting tools split here. Some append a number and keep both. Some replace. Some skip the file and log it. A setup left on replace will eventually delete something, and the files most likely to collide are exactly the ones people automate: invoices, statements, screenshots, exports, all named by a machine to a fixed pattern.

Undo depth matters for the same reason. Automated moves are usually noticed days later, so a tool that can only reverse the last action is not much use. A plain text log is worth more than a fancy undo button.

There is a cheap insurance policy that works regardless of which tool wins. Send sorted files to a dated staging folder rather than straight to the final destination, and empty that folder manually once a week. Anything the rules got wrong sits in one place, still recoverable, and the weekly pass doubles as a review of whether the rules are still matching what they were written for. After a month the staging step can be dropped with some confidence, or kept permanently for the categories where a mistake costs the most.

Question 4: Where it runs and what it is allowed to touch

Automatic sorting reaches into the folders macOS protects most carefully. That access is granted per app, and it can be revoked.

Some apps and websites can access files and folders in your Desktop, Downloads, and Documents folders. You can decide which apps and websites are allowed to access files and folders in specific locations. Source: support.apple.com

Three things to check before committing. Whether the tool runs as a background process at login or only when opened. Whether it asks for folder level access or full disk access. And, for anything script based, which app is actually doing the asking, because a shell script inherits the permission of whatever launched it rather than carrying its own.

Scope limits belong here too, and the built in route has hard ones. Folder automations cannot point at the trash. They cannot point at external drives. The maximum folder size is 10,000 items, and an automation is disabled outright when the target folder exceeds it, or when the folder is moved to the trash. Anyone whose real target is an external SSD full of footage has just eliminated the free option, and that is useful to learn before spending an evening building rules.

Putting the four questions in one table

What to check	Built in folder automation	Automator folder action	Resident rule engine
Trigger	Added, modified, removed	Items added	Rule match, schedule
Rule building	Written by hand in the shortcut	Written as a script	Assembled in a list
Allowed targets	Internal disk, up to 10,000 items	Per folder	Varies by product
Record of what moved	Notification based	Rarely kept	Log in some products
Price	Included with macOS	Included with macOS	From $42

The table is not meant to be read top to bottom. Pick the row that cannot be compromised, eliminate whatever fails it, and compare what is left on the remaining rows. Searching for a tool that wins every row is how a two hour decision becomes a two week one.

A two week test that costs nothing

Most of this can be settled before buying anything. Create a fresh folder. Copy thirty representative files into it, copies rather than originals, so nothing real is at risk. Build the same single rule in each candidate and run it against that folder.

Three things are worth watching. How many of the thirty landed where expected. Whether the tool explains why the strays went elsewhere. And whether a file that did not move can be traced to a specific condition. A tool that cannot answer the third question during a test will not answer it in production either.

Salt the sample deliberately. Include one file with an uppercase extension, since cameras and phones often write them that way and a lowercase only condition skips them. Include one name containing full width spaces or parentheses. Include one large file that takes a while to copy, to see how the trigger behaves on something still being written.

Then run one rule, on one folder, for two weeks before adding a second. Counting misses and false moves on a single rule produces a clear answer. Ten rules at once produce noise.

What to change first

Count the times per week a file cannot be found within a few seconds. Below roughly five, no sorting engine is needed and splitting the download destinations will do. Above ten, start with the question from the list above that hurts most, and choose only for that. The Pricing page and the FAQ cover the cost and setup side of running this kind of tool day to day, and Atriens sits alongside the other file managers if a single window for folders, terminal and rules is the direction worth testing.