Search inside every PDF on your Mac at once
You know the sentence that was in it. You do not know which of four hundred files it was in, or which folder you put it in three years ago. That is a search problem, and the tools on the Mac only half solve it.
Loupe reads the text of every PDF in the folders you choose, runs OCR on the scanned ones that contain no text, and indexes both. Press ⌥Space, type a phrase you remember, and it lists the matching PDFs with the line that matched under each. It all happens on your Mac, and no file is modified.
Why this is harder than it should be
Preview will search the PDF you already have open, which is no help when the question is which PDF to open. Finder's search box will match filenames and whatever the system index happens to hold. Between them they cover the easy half of the problem and leave the half that actually costs you an afternoon.
There are two kinds of PDF, and they behave completely differently
PDFs made by software
An invoice exported from an accounting package, a report saved out of Word, a ticket emailed by an airline: these hold real text. You can select a sentence in Preview and copy it. The Mac's own index generally reads these, and so does everything else.
PDFs made by a scanner or a camera
A contract you scanned, a receipt you photographed, a letter someone faxed a decade ago — these hold a picture of a page and nothing else. Try to select a line in Preview and nothing highlights. There is no text in the file, so there is nothing for any index to store, and the file is findable by its name and its date and by nothing else at all.
Most people's Documents folder is a mixture, which is why searching feels unpredictable: it works, then it does not, and there is no obvious reason why. The reason is which kind of PDF you happened to be looking for.
What Loupe reads
Both kinds, and both halves of the ones that are a bit of each. Text is read from every page. Any page that carries none is read with optical character recognition on your Mac instead — up to five such pages per file, wherever in the document they are — and a page with no text even after that is described by what it shows. From the search box it is all one index, except that a result found by OCR is labelled Scanned, so you always know why something matched.
It reads Word, Excel and PowerPoint files, plain text and source code the same way, and describes images so they are searchable too. Pages, Numbers and Keynote are not read yet — we would rather say so here than let you find out after paying.
Searching for what it was about, not what it said
Keyword search has one failure mode and it is a big one: you have to guess the words the author used. You search for the quote for the bathroom; the document says Estimate — sanitary ware and fitting. Not one word in common, and a keyword index is right to return nothing.
Alongside the keyword index there is a second one built on meaning, and results from the two are merged. So the estimate comes back, labelled Meaning rather than Text, and you get the file instead of a lesson in how to phrase a query.
Nothing is uploaded and nothing is rewritten
Reading somebody's documents is a lot to ask, so all of it happens on the machine: the reading, the recognition, the ranking. There is no server on the other end, which is why searching keeps working with the network off and why it costs nothing to run. And the app only ever reads — your PDFs are not re-saved, not given a hidden text layer, not moved and not renamed.
Questions people ask about this
How do I search the text inside all my PDFs at once on a Mac?
Preview searches one open document, and Finder searches names and whatever Spotlight has indexed. To search the text of every PDF on the disk at once — including the scanned ones, which contain no text at all — you need an index that has opened and read them. That is what Loupe builds: press ⌥Space, type a phrase, and it lists the PDFs whose contents match, with the matching line under each one.
Does it find text in password-protected PDFs?
No. A PDF that will not open without a password cannot be read, so it stays findable by name only. That is the correct behaviour — an indexer that could read your encrypted documents would be a worse thing to install.
Can it search PDFs stored in iCloud Drive?
It reads what is on the disk. A file that iCloud has offloaded to save space is not on the disk, so it is indexed once it has been downloaded — the same limitation every local search tool has. Files kept locally are read normally.
What if I only remember roughly what the PDF was about?
That is the case it is built for. Alongside the keyword index there is one built on meaning, so a query and a document match because they are about the same thing rather than because they share a word. Searching for “the lease for the flat” finds a document headed “Tenancy Agreement”.
How large a PDF can it handle?
Text is read from every page, however long the document. Pages that carry no text are read with OCR instead, up to five of them per file — five pages that needed it, wherever they sit, not the first five in the file. So a report with scanned exhibits at the back gets both halves, and a page that is a photograph or a chart is described rather than skipped.
Find the PDF by the sentence you remember.
Seven days, no card, no account. If it does not turn up the thing you gave up looking for, you have lost an afternoon of background reading and nothing else.
- Apple silicon · macOS 26 or later
- No account
- Nothing leaves your Mac