AI SEARCH IN WINDOWS

Goal
Find what you need faster with AI in Windows search
Product
Windows - Microsoft
Team Size
1 PM, 3 Eng, 1 Designer
AI search in Windows: searching “White dress” shows suggested filters, matching files and photos, and a preview of the bridal collection presentation with an AI summary

CURRENT EXPERIENCE

SEARCH IN WINDOWS

Windows search uses a lexical approach, matching results by exact words rather than context. It functions across various areas, including File Explorer, Settings, and the main Windows search box.

File Explorer search for “screen”, highlighting only file names that contain the exact word
File Explorer
The Windows search bar showing best match, apps and settings for “screen”
Windows search bar
Settings search for “power” with a list of matching settings
Settings
Photos app search results for “screen” in the gallery
Photos

PAIN POINTS

Three pain points. Precise keyword: Windows relies on exact word matching, lacks context or concept-based understanding. Inaccurate results: results don’t always match the user intent, often showing unrelated items. Web suggestions: web results appear even when users are searching for local files.

ENABLING AI IN SEARCH

SEMANTIC SEARCH IN WINDOWS

A search experience that aligns more closely with how humans naturally seek information context insight data

Three capabilities of semantic search. Related word search: expands search results beyond exact word matches by identifying relevant terms and concepts. Deep content search: enables users to discover specific information within documents without needing to open them. People in semantic search: integrates information about individuals to enhance the relevance and connections in search results.

DEEP CONTENT SEARCH

Semantic search understands user intent, not just keywords, allowing for more relevant document retrieval. For example, searching for “white dress” can yield results like “wedding dress” or “wedding planning.”

It also identifies relevant content within a single file, scanning across slides or pages. This makes it easy for users to find all the information they need without opening each file individually.

A search for “car”: lexical search lists file names, while semantic search finds an AutoExpo presentation and the car photos inside it

Deep Content Search in File

Deep Content Search in Video

SEMANTIC SEARCH ACROSS VARIOUS SURFACES ON WINDOWS

Deep Content Search in File Explorer

File Explorer search for “vacation” finding photos and videos by what they show, such as kids on a beach File Explorer search for “white dress” showing top results and all media with matching photos of wedding dresses

Deep Content Search in Windows Search bar

The Windows search box on the desktop with results for “white dress”: files, photos and a summary of a bridal collection presentation

Deep Content Search in Recall

Search results for “white dress” across text matches and visual matches from earlier activity

PEOPLE SEARCH

Semantic search also enables recognition of people in images by identifying contextual clues and linking them to known profiles.

For example, if Tejasvi appears in a group photo or event-related image, semantic search can infer her presence through context – like her associations with specific projects, teams, or frequent collaborators. This level of context-aware tagging makes it easier to locate relevant files and provides a fuller picture of someone’s involvement across various media.

A search for “car”: lexical search lists file names, while semantic search finds an AutoExpo presentation and the car photos inside it

People Video


NEXT PROJECTS