Whenever Microsoft starts messing with something like the search function in Windows, a lot of people are going to be holding their breath and expecting the worst. Well, get ready, because they’re redesigning the whole thing.
In July, we shared how we are improving the Windows Search Box with less clutter and more control. We introduced a calmer home screen, removed promotional content from web results, and gave you more choice over whether web and Microsoft Store suggestions appear.
Today, we’re carrying those investments forward with a new Windows Search experience built on WinUI 3. This modern foundation makes Windows Search faster and more efficient with resources, while delivering an even more streamlined and focused design.
↫ Anshul Rawat at the Windows Blogs
The company claims this new version of search is faster and uses less memory, which, if true, are very welcome improvements. It’s also just a single column of results now, and they’re adding inline previews for things like weather and files on your machine, as well as for answers to all kinds of queries you might normally go to an online search engine for. You can now also use search to perform all kinds of tasks in Windows, like turning on dark mode, managing windows, and so on.
All of this can be done using relatively natural language, and Microsoft claims they’ve implemented better detection of typing errors and synonyms, which is quite welcome. Weirdly enough, though, it’s not yet integrated into the Start menu, since it’s a preview just for testers, but that’s something they’re working on for future releases.
Judging by the videos and screenshots, I’m actually kind of positively surprised. This looks quite decent and nice, and if the promised performance improvements actually materialise, this may actually be an upgrade to search worth looking forward to. Of course, this is still Microsoft, so it’s just as likely they’ll screw up somewhere between now and when this goes live to regular users.
Still, I think it looks decent.

Tell me if I’m wrong, please, but a lot of this seems AI based. The result on the Eiffel Tower, Turning dark mode on… Isn’t this AI powered stuff? I didn’t see AI mentioned in the article (I did a search).
Honest question: is (local) search really such a big thing? I have 1000nds of documents from over 20 years of work, million lines of source code. But except a grep for filenames I literally never search anything. If something is not in the folder where it should be, then it simply does not exist. Index jobs are usually the first thing I deactivate, since it eats lots of resources achieving nothing (for me).
What are people searching for on their own computers? (At large, so it justifies that much effort into all this UI and indexing work?) I just don’t understand the use-case.
If you know the name of the file you want to open then search would be much quicker than navigating to the folder yourself.
In windows? Granted, I’m not using windows any longer, but in decades of past use, windows search sucked so badly, it was always the first thing i switched off. And no, it could not find a file by name, at least not in a reasonable time. Linux finds everything instantly. Enough said.
I use search all the time, but the built in Windows search is crap. I use DocFetcher instead.
I write legal briefs as part of my job and have been doing it for a very long time. The briefs are kept in the respective client’s folder, not just lumped together in a big “briefs” folder. If I want to search for an argument I may have last made 5 years ago, I do a keyword search on document contents. Each brief generally has 3 or 4 arguments.
Similarly, I can search from a list of witnesses in a case (or potential jurors) from within my files for matches. I don’t need to keep a separate database of everyone I’ve ever encountered.
While you personally do not use search much, a lot of people use it a lot. I remember when Apple added Sherlock to Mac OS and made it much easier to track down files.
Andreas,
Windows 95 search got you covered (peak search window bliss IMHO). It exposed most parameters that nowadays would require flags in a nice and user-friendly way: date range, text contents (could well be made a awk/grep compatible), file types, location (and I think it could search network mounts). If they would release the Windows 95 search box back and add options for metadata (good for photos), it would be incredible.
And it didn’t require an index job. I am with you. I always disable the search indexer. In Windows, it loves to enjoy the fact that you got some free CPU (just writing text or simple code), and HEY LETS THERMALLY CASTRATE YOU AND RUIN YOUR BATTERY LIFE BECAUSE THE CPU IS FREE.
I am the kind of person who searches for files in my own computer. I have 25 years of digital images in my local disk/local NAS. I used to trust Apple Aperture for indexing and searching. (I still miss it). And I won’t pay Adobe rent. So I ended up migrating to a folder structure and I just use NX Studio (which I can’t upgrade because the new version not only requires an account but it is VERY aggressive about phoning home and won’t load even after a few days offline).
Other sources are scans. Photos taken with my F6 are EXIF-indexed and in a DB (managed by Meta35 but trivial to edit without it). Digital photos and the F6 photos are serially-numbered. Larger formats depend on a name structure for me to know subject, film and date.
I even considered at some point Haiku (due to the file system being metadata pervasive) but now I just end up using a bunch of scripts when I need to search for something. Windows search is miserable. I can be in the folder where a file is, type its name, and it won’t find it.
And I will never trust anything proprietary to handle my archive. It took me one year to migrate away from Aperture. When we are talking about multiple TBs of huge files from different sources, a decent local search option would be great. I may want to find photos by rating or search via EXIF tag. Not always I exactly remember when a photo was taken or from which source/camera.
edit: I can see from the new Windows search that it is still lacking options to narrow scope. If I remember something was done between 2010 and 2015, with certain tags, would be nice to narrow the scope of the search to not have the search running through the whole thing. Images are less trivial to index, and I even got some of my archive already identified and tagged using image recognition and indexing the EXIF tags to make searching faster and easier. But otherwise it really has to read whole files and it can take a while.
Shiunbird,
What you describe is the way I like to do searching as well. I like data and searches to be structured, nailing down search scope. Like you I also remember the old windows search and frankly I find it’s evolved backwards. For better or worse, search engines like google popularized a one size fits all search box. This is considered “user friendly” but can be inferior for parametric searches. It’s not helpful that modern search engines return whatever they feel like even if it means ignoring what I deliberately asked for. Google for one has gotten worse about matching exact quotes over the years. It’s one of my gripes with modern windows search too – cramming more sources into the search feature without putting sufficient thought into things like scope/organization/structure. I’d like to see a return to more structured search interfaces.
I am really curious what you are doing for image recognition because I think that would be useful to have myself.
nextcloud recognize. I have my photo storage mounted as external folder. It works really well! Depending on the place and popularity, it even gets right mountain names.
Shiunbird,
Thanks for that. I don’t use next cloud, but I wonder if I could hack the recognition tools into what I need. I have a large collection and it’s categorized by trip and all, but the individual photos aren’t tagged so it’s hard to search them effectively.
https://github.com/nextcloud/recognize
You probably can. The models are not unique to nextcloud and, if I recall correctly, it calls some ordinary binary and creates a JSON for each file with the tags.
My film shots are tagged manually (I tag them while I scan so it takes no time). But didn’t have a lot of discipline when I was a teenager so recognize helped me sort out the chaos.
I’m not sure what Windows search is like now, but in the Windows 7-era, I used to have what I think were called “Live Search Folders”, where basically you save a query, it appears in Explorer as a folder, and you could set up the search for things like “image files greater than 1000×1000 pixels” or whatever, and they’d all appear there. Not sure if Windows can still do that, but it used to index not just filesystem metadata, but also file metadata (such that you’d find in EXIF or ID3 tags).
It was very useful.
Search is BeOS ? :popcorn:
I use “Everything” from https://www.voidtools.com/ on my Windows XP and Windows 7 hobby machines.
Best search tool I’ve ever found on any platform, thanks to how it can keep itself up to date in a hyper-efficient manner by watching the NTFS journal.
I’d say Microsoft should just buy and integrate it like Apple did for so many popular extensions in System 7.5, but then support for older versions of Windows in new versions of the tool would be dropped.
I’ve been using Everything for search on Windows machines for as long as I can remember. This tool, along with WinSCP, PuTTY, and Notepad++ help me keep my sanity at work as I have to use and support Windows machines all day every day.
Morgan,
I install PuTTY, notepad++, filezilla on my windows machines to handle the basics 🙂
I’ve never tried “Everything”, it looks useful but since I don’t store anything permanently on windows the need to search for anything with windows is kind of redundant for me. I rely on a NAS for just about everything. I wish android were more flexible in terms of mounting NAS though. I hate how android forces me to copy files locally rather than mount them remotely. I think the reason neither google nor apple support user controlled CIFS is because they’d rather tether users into cloud services.
Heck I’ve found even microsoft is going this direction with windows. Some of the clients I support were unknowingly sending documents to onedrive and their own files were being deleted from the local machine without permission.
They did install onedrive, but never gave microsoft permission to delete the files locally. And it’s kind of alarming because the users didn’t know this was happening until their subscriptions expired. The death of window as a local operating system is underway 🙁
It’s probably not a real solution to this specific issue, but I’ve found FE File Explorer on both Android and iOS to be phenomenal at transferring files across networks nearly effortlessly. It has support for pretty much all network file systems; I use it at work to transfer photos I take with my phone directly to a shared folder on my workstation. As usual, the Android version of the app is far more flexible and featureful than the iOS version; even so, the iOS app allows you to do things with files that the native Files app won’t, and opens up file transfers to non-Apple devices. You do need to spring for the paid version if you want to really take advantage of it, but even the free version lets you do far more than Apple thinks you should with your own files.
Over the last few decades, every time Microsoft touches any core system in Windows, they made it twice as slow, and about 10 times as bloated. But don’t worry – now they’ll vibe code the “improvement” – what could go wrong!!
Well said. If only search was the only problem with Windows. (Well, to be fair I quite like the fact that there are problems with everything in Windows, since it encourages people to seek out alternatives.)