Windows 11 semantic search without a Copilot+ PC
Updated July 2026
Windows 11 can now search your files by meaning. Describe a document or a photo in plain language and Windows works out what you meant, instead of matching the exact characters in a filename.
There is a catch, and it is a big one: it only runs on Copilot+ PCs. If yours is not one, the feature simply is not there, and no setting will turn it on.
What “Copilot+ PC” actually means
Copilot+ is a hardware badge, not a Windows version or an update you can install. To carry it, a machine needs a dedicated NPU (neural processing unit) that meets Microsoft’s performance bar. In practice that means a recent Snapdragon X, Intel Core Ultra 200V, or AMD Ryzen AI 300 chip, nearly all of them in laptops bought since around 2024.
Here is the part that catches people out: raw power does not substitute for an NPU. A desktop with a sixteen-core processor and an expensive gaming graphics card is not a Copilot+ PC, and does not get semantic search. The gate is the presence of the NPU and its firmware support, not how fast your machine is overall. Plenty of people with genuinely powerful PCs go looking for this feature and find it missing.
How to check whether your PC qualifies
- Open Task Manager (Ctrl+Shift+Esc) and click Performance. If there is an NPU entry in the left-hand list, alongside CPU, Memory and GPU, you have one.
- Or check Settings → System → About and compare the processor name against the chip families above.
- No NPU listed means no Windows semantic search, whatever your processor or graphics card. That is the whole answer, and it is worth knowing before you spend an afternoon hunting through settings.
Option 1: Enhanced indexing (free, already on your PC)
Every Windows 11 machine can widen what the normal search box covers. Go to Settings → Privacy & security → Searching Windows and switch from “Classic” to “Enhanced”. Classic indexes only Documents, Pictures, Music and your Desktop. Enhanced indexes the whole drive.
What this does and does not fix:
- Does fix: files outside your user folders becoming findable at all.
- Does not fix: the matching itself. It is still literal keyword search. Ask for “salary” and a payslip that says “compensation” will not come up. Ask in one language and documents written in another stay invisible.
- Does not fix: scanned documents and screenshots, where the text is really a picture of text. There is nothing there for a keyword index to read.
- The first pass can take hours and is noticeable on battery.
Worth doing regardless, it costs nothing. Just do not expect it to be the feature you came looking for. More detail in our guide on searching inside files on Windows.
Option 2: Search by meaning on the hardware you already have
Here is the thing worth understanding: the NPU is not a capability requirement. It is a power-efficiency choice.
The models that make meaning-based search work are small by modern standards. They run on an ordinary processor, and considerably faster on an ordinary graphics card. Microsoft’s constraint is that they want this running continuously in the background on battery-powered laptops without hurting battery life, and an NPU is extremely good at that particular tradeoff. That is a sound engineering decision on their part. It is not a statement that your PC is incapable of the work.
BeaconFind does this kind of search on any 64-bit Windows 10 or 11 machine, with no NPU required. It runs on macOS with Apple Silicon too. You press Ctrl+K, describe what you are after, and it searches the contents of your files. Ask for “the apartment contract with the deposit amount” and it finds the document even when the file is called scan0047.pdf and never uses the word “deposit”.
- Reads 70+ file types, including PDF, Word, Excel, PowerPoint, email, code and plain text.
- Reads text out of screenshots and scanned PDFs using OCR, so documents that keyword search cannot see become findable.
- Understands 50+ languages, including across languages, so a query in one language can surface a document written in another.
- Finds photos by description, which is the other half of what Windows semantic search offers Copilot+ owners.
On the privacy question, which matters more here than usual because this is your entire disk: everything runs on your own machine. Your files, their contents, your searches and the index never leave your PC. There is no cloud processing and no telemetry, and the index is encrypted on disk. The app reaches the internet only to verify your licence and to check for updates, and you can block both at your firewall without stopping search from working.
No NPU, no problem. Free for 14 days, no card needed. Windows 10 and 11, 64-bit, and macOS 12 or later on Apple Silicon.
Download BeaconFindThe honest tradeoff
Running this on hardware that was not designed for it is not free, and you should know the cost before you download anything.
- First run downloads the AI models once, roughly 0.8 GB on the lightest tier and up to about 6 GB on the highest. You pick the tier during setup.
- Indexing takes longer without a graphics card. An NVIDIA GPU is used automatically if you have one and makes a large difference. On the processor alone it still works, it just takes longer to get through a big library the first time.
- Background indexing backs off while you are using the PC and can be paused entirely. That is the problem an NPU solves in hardware, solved here by scheduling instead.
So which should you use?
- You have a Copilot+ PC: use what Microsoft built. It is included and it is well integrated.
- You do not, and you mostly remember exact filenames: turn on Enhanced indexing and stop there.
- You do not, and you remember what a document was about rather than what it was called: that gap is exactly what meaning-based search closes, and it does not need an NPU to do it.