Exploring AI actions in Windows 11 - Semantic Search

Have you ever spent minutes searching for a file you know exists somewhere on your PC - but you cannot remember the exact name, folder, or even what you called it?

Finding files has always been one of those friction points in daily work. Traditional search requires you to remember exact file names, specific keywords, or at least which folder you saved something in. When you are juggling dozens of documents, presentations, and images across local storage and cloud services like Microsoft OneDrive, that approach breaks down quickly. I have been in that situation more times than I care to admit - searching for “that document about the deployment plan” by guessing file names until something sticks.

Semantic Search on Copilot+ PCs changes this entirely. Instead of relying on exact matches, you can describe what you are looking for in your own words - and Windows understands what you mean. Based on my experience with this feature over the past months, it has completely changed how I find and interact with files. What follows is a look at what Semantic Search is, how to enable it, and practical examples that show its value. It is one of the most overlooked AI features in Windows 11.

What is Semantic Search?

Semantic Search is an AI-powered upgrade to Windows Search available on Copilot+ PCs and Windows 365 AI-enabled Cloud PCs running Windows 11 version 24H2 or later. Unlike traditional search that matches exact file names and keywords, Semantic Search understands the meaning behind your query. It finds files based on their content and visual elements - even when your search terms do not match the file name or contents word for word.

When you use Semantic Search, Windows uses the on-device Neural Processing Unit (NPU) to understand the intent behind your query. For example, searching for “compliance policy for corporate devices” can surface a file named intune-win11-baseline-config.docx - even though none of your search words appear in the file name. The same applies to images: searching for “screenshot of a failed deployment” can find matching images regardless of their file names.

Semantic Search is accessible from Windows Search, File Explorer, the Photos app, Settings, and the Microsoft Copilot app. You can use it from wherever you happen to be working.

Image viewer

The key capability that sets Semantic Search apart from traditional indexing is its ability to understand relationships between concepts. It does not just look for the words you typed - it understands what you mean. Searching for “my resume” can find a file called cv-jesper-2026.docx. Searching for “recipe with chicken” can locate a document with a tostada recipe that mentions chicken in the content.

All processing happens locally on your device using the NPU - your search queries and file contents are not sent to the cloud. This is an important distinction for professionals handling sensitive documents and for organizations with strict data residency requirements.

The quickest way to confirm Semantic Search is available on your device is to look for the sparkle icon in the Windows Search box. If you see it, you are good to go. If the icon does not appear, the details section below walks through verifying your hardware and AI components.

Understanding the semantic index

The semantic index is the foundation that makes natural language file search possible. Understanding how it works helps you get the most out of the feature and troubleshoot when searches do not return expected results.

How semantic indexing works

Traditional Windows Search creates an index based on file names, metadata, and text content - matching your search terms against stored keywords. The semantic index goes further by creating vector representations of your files that capture meaning, not just words. This allows Windows to understand that “quarterly financial summary” and “Q3 budget overview” refer to similar concepts.

The semantic indexing process runs in the background using the NPU, which means it does not noticeably affect CPU performance or battery life. When you add, modify, or delete files, the semantic index updates as files change - you do not need to rebuild it manually.

Image viewer

What gets indexed

Semantic Search creates its index from files in your configured search locations. The following content types are supported:

For images, “visual content” means the AI analyzes what the image actually shows - objects, scenes, text within the image, and spatial relationships. A photo of a whiteboard gets indexed based on its visible content, not just its file name or EXIF metadata. This is what makes queries like “whiteboard with architecture diagram” possible.

For Microsoft OneDrive files marked as on-demand (cloud-only), Semantic Search can still surface them in results based on available metadata and content that has been previously synced. Files that have never been opened locally may have limited semantic indexing until they are downloaded.

File types not listed in the table - such as source code files, compressed archives, or proprietary formats - are not semantically indexed. They still appear in traditional keyword search results based on file names and metadata, but Semantic Search will not understand their content.

Semantic index vs. traditional index

The semantic index works alongside the traditional Windows Search index - it does not replace it. The two indexes complement each other:

When you search on a Copilot+ PC, Windows queries both indexes at once and merges the results. This means exact file name matches still appear instantly (traditional index), while conceptually related files surface alongside them (semantic index). You do not need to choose between the two - the system uses whichever index produces the most relevant results for your query.

The traditional index still excels at precise lookups. If you know the exact file name or a specific keyword that appears in the document, traditional search finds it faster. Semantic Search shines when you cannot recall those specifics - when you remember what the file was about but not what you called it. In practice, the two approaches cover each other’s weaknesses.

Enabling Semantic Search

Semantic Search is available automatically on Copilot+ PCs - there is no separate toggle to enable it. The feature activates when the required AI components are installed and the search index is configured. To verify and optimize your search settings, follow these steps:

1. Open Settings (Windows key + I)
2. Navigate to Privacy & security > Search
3. Under Find my files, choose your preferred indexing scope - Classic indexes your Documents, Pictures, Music, and Desktop folders, while Enhanced indexes your entire PC

For most users, Classic mode is sufficient. It already covers the standard user profile folders - Documents, Pictures, Music, and Desktop - where you typically store your files. Only switch to Enhanced if you regularly store files outside these locations.

Image viewer

Configuring search locations

To get the most complete results from Semantic Search, consider which folders and locations are included in the search index:

1. Open Settings > Privacy & security > Search
2. Under Find my files, select Enhanced to index your entire PC, or keep Classic and customize locations
3. If using Classic mode, select Customize search locations > Modify to add specific folders
4. If you use Microsoft OneDrive, verify that your cloud folders are included in the indexing scope

If you keep files in non-standard locations outside your user profile, consider switching to Enhanced or adding those folders manually. This ensures Semantic Search covers everything you might need to find.

Using Semantic Search

Using Semantic Search does not require learning new tools or changing how you work - it meets you where you already are. The feature is accessible from Windows Search, File Explorer, the Photos app, Settings, and the Microsoft Copilot app. You can use natural language queries regardless of what you are doing at the time. The table below summarizes each entry point and when it works best:

Search tips for better results

Semantic Search responds best to natural, descriptive language. Here are some tips for getting the most relevant results:

• Describe the content, not the file name - Instead of guessing file names, describe what the file contains. “presentation about endpoint security” works better than trying to remember if you named it security-deck.pptx or endpoint-overview.pptx.
• Use qualifying details - Add context like time, topic, or file type to narrow results. “budget spreadsheet from January” is more specific than just “budget”.
• Search for visual content in images - Describe what the image shows rather than guessing file names. “screenshot of an error message” or “photo of a whiteboard with architecture diagram” takes full advantage of the AI image understanding.
• Try different phrasings - If your first query does not return what you need, rephrase using synonyms or different descriptions. Semantic Search understands concepts, so “my CV” and “resume document” should return the same results.

Practical examples

Understanding what Semantic Search can do is one thing - seeing how it fits into real work is another. The following examples reflect scenarios I encounter regularly. They cover finding documents I cannot name, locating images buried in generic file names, and searching across storage without thinking about where I saved something. Each example demonstrates a different entry point and search style.

Finding documents by content description

When you need to find a specific document but cannot remember what you named it:

1. Press the Windows key to open Windows Search
2. Type a description of the document content, for example: “deployment plan for the new office”
3. Review the results - Semantic Search surfaces files whose content matches your description, even if the file name is completely different
4. Click the result to open the file directly

This is especially useful for professionals who work with many similarly named files. It also helps when you receive documents from others with naming conventions you did not choose.

Searching for images by visual content

When you need to find a specific image but your photos folder contains hundreds of files with generic names:

1. Open the Photos app on your Copilot+ PC
2. Type a description of the image content in the search bar, for example: “sunset over water” or “screenshot with a blue error dialog”
3. Semantic Search analyzes the visual content of your images and returns matches based on what the images actually show
4. Browse the results and select the image you need

This removes the need to scroll through hundreds of thumbnails or maintain detailed file naming conventions for images.

Image viewer

Locating cloud files in Microsoft OneDrive

When you store files across both local storage and Microsoft OneDrive:

1. Press the Windows key to open Windows Search
2. Type a description of the file you are looking for, for example: “project proposal shared with the team”
3. Semantic Search searches across both your local files and Microsoft OneDrive files that are synced or available on-demand
4. Results indicate whether the file is stored locally or in the cloud

This unified search experience means you do not need to remember whether you saved a file locally or in Microsoft OneDrive - Semantic Search finds it regardless of location.

Finding settings you cannot locate

When you need to change a Windows setting but do not know where Microsoft placed it:

1. Open Settings (Windows key + I)
2. Type a natural language description in the search bar, for example: “stop apps from running at startup” or “change what happens when I close the laptop lid”
3. Semantic Search understands your intent and surfaces the relevant settings page, even when your description does not match the official setting name

This is especially helpful for IT professionals who need to quickly locate settings across the ever-growing Windows Settings app.

Searching within the Microsoft Copilot app

When you want a more conversational search experience with follow-up capabilities:

1. Open the Microsoft Copilot app on your Copilot+ PC
2. Type a query like “find images of bridges at sunset on my PC” or “find the file with the chicken tostada recipe”
3. Copilot uses semantic file search to locate matching files on your device
4. You can then ask follow-up questions or take actions on the results, such as summarizing a document or copying an image

The Microsoft Copilot integration adds a conversational layer to Semantic Search. You can refine your search through natural dialogue rather than rewriting queries by hand.

Privacy and data handling

Privacy is a key concern for any AI-powered feature that processes your files. Semantic Search addresses this with a clear on-device processing model.

Semantic Search handles your data with these privacy principles:

• On-device processing - On physical Copilot+ PCs, all semantic indexing and search processing happens locally on your device using the NPU. Your file contents and search queries are not sent to the cloud.
• No data sharing - Microsoft does not access, collect, or share your semantic index data. The index is stored locally and encrypted.
• User control - You can enable or disable Semantic Search at any time through Settings. Disabling the feature removes the semantic index from your device.
• Organizational control - IT administrators can manage Semantic Search through Microsoft Intune policies, allowing them to enable, disable, or configure the feature across managed devices.

For organizations with strict data governance requirements, the on-device processing model means that Semantic Search can be used without introducing new data residency concerns.

Other AI features in Windows 11 on Copilot+ PCs

Beyond Semantic Search, Copilot+ PCs include several on-device AI features that cover productivity, creativity, and communication. If you are exploring the AI features available on your Copilot+ PC, consider these related features:

• Click to Do - instant actions on screen content: Take instant actions on anything visible on your screen - extract text from images, search the web for selected content, or summarize text without switching applications. For a detailed walkthrough, see my post on Exploring AI actions in Windows 11 - Click to Do.

• Recall - searchable timeline of what you saw: Periodic, encrypted local screenshots and on-device indexing let you find previously seen content. Strong privacy controls exist. Great for retrieving information you forgot to bookmark or capture.

• Cocreator - image generation and transformation: Integrated into Paint and select apps; combine sketches and prompts to produce new visuals or restyle existing images for documentation or marketing assets.

• Windows Studio Effects - better video and audio: AI-powered camera and microphone enhancements (background blur/replacement, automatic framing, eye-contact correction, voice focus) for higher-quality meetings without extra hardware.

• Live Captions with translation - real-time, local translation: Captions for any audio with on-device translation between many languages - useful for meetings with distributed teams.

For more details on these features, see Copilot+ PCs  on the Microsoft website.

Final thoughts

Semantic Search marks a major shift in how we find files on our PCs. The old approach of remembering file names and folder structures has always been a workaround for a search system that could not understand what you actually meant. With Semantic Search, that limitation disappears - you describe what you are looking for, and Windows finds it.

IT professionals often manage large collections of documentation and configuration files. For them, searching by content description rather than file name removes a major source of daily friction. For knowledge workers dealing with files spread across local storage and Microsoft OneDrive, the unified search experience means one search covers everything.

The key to getting value from Semantic Search is building the habit of searching descriptively. Instead of trying to guess the file name, describe the content. Instead of browsing through folders, type what you remember about the file. After a few days of intentional use, you will find yourself wondering how you managed without it.

For organizations that cannot yet refresh their hardware fleet, Windows 365 AI-enabled Cloud PCs offer a path to deliver these same capabilities to users on any device. Semantic Search and Click to Do are streamed securely from the Microsoft Cloud. The feature is currently available through the Frontier Program as a public preview, with requirements starting at the 8 vCPU Windows 365 Enterprise SKU.

If you have a Copilot+ PC or access to an AI-enabled Windows 365 Cloud PC, I highly encourage you to verify that Semantic Search is enabled. Give it a try with your own files. And if you have not yet explored the other AI capabilities on Copilot+ PCs, there is a lot to discover. That includes Click to Do, Recall, and the features I mentioned above.

–Jesper

Header image attribution: Image created with help from Adobe Firefly