visual search on bing
year
2025-2026
TOOLS
FIGMA, CLAUDE CODE, ux labs
design TEAM
solo
project overview
Search is no longer text-first As users increasingly combine images and text to express their intent. however Visual Search on Bing remained a standalone feature, confined to the Image Detail Page. This created an opportunity to reimagine Visual Search as a native search modality embedded throughout the Bing search experience.
In my role, I worked on defining the experience strategy, scalable design principles, and shaping multimodal experiences that seamlessly combined visual understanding with traditional web search. I partnered with cross-functional teams to align on solutions that could scale across Bing surfaces.
The projects below are part of multiple initiatives that contributed to the growth of Visual Search's weekly active users from 5 million to 14 million—a 3× increase over 12 months.
key challenges
01
Users struggled to discover where to use Visual Search as it lived outside the primary search experience
02
Search remained text-first even when users had predominantly visual intent
03
Users struggled to understand what Visual Search could do limiting it to percieved image serch instead of translation entity recognition problem solving and ai-powered understanding
KEY PROJECTS
01
visual search entry point (GROWTH)
research and compete analysis, opportunity framing, growth strategy, interaction design, stakeholder alignment
2 entry points have been shipped. The hover framework is being scaled to multiple VS entry points. Overall growth initiatives have added 400K DAU
02
visual search results on serp
designing scalable solutions, cross functional partnerships, data-informed design, influencing product strategy, interaction design
4 experiences have been shipped on SERP with positive engagement metrics
03
imporving visual search input modality
system thinking, influencing product direction, end to end flow, rapid prototyping
Supported the design team with social media graphics, print materials, and brand guideline development
• DISCOVERABILITY
Starting visual search from Video Vertical
" I want to find the shoes in this video, but there is no easy way to search for them."
Prototype tool: VS code + Claude Code
The Brief
How might we meaningfully integrate Visual Search into video experiences, enabling users?
My role
CREATED A CLAUDE SKILL FOR COMPETE ANALYSIS - THIS Explored the opportunity space by benchmarking video commerce, visual search, and multimodal AI experiences, IT ALSO MAPPED ALL INTERACTION PATTERNS AROUND PRODUCT/ENTITY DIsCOVERY FROM VIDEOS
This primilary research helped in the exploration while solving for how visual understanding could be meaningfully integrated into video consumption.Defined the experience strategy by framing key interaction questions—what should Visual Search identify, when it should appear, where it should surface, and how users should transition from video to exploration.
Established design principles to ensure Visual Search enhanced moments of curiosity without interrupting the primary video viewing experience.
Explored multiple interaction models across different trigger points, placements, and navigation patterns, evaluating how each balanced discoverability, user intent, and viewing continuity.
Built high-fidelity interactive prototypes using Claude Code to simulate the end-to-end experience, enabling realistic stakeholder reviews and early engineering feasibility discussions.
Built Claude skill for compete analysis & preliminary research
What does it do?
This skill can be evoked for any project to do the following -
- a landscape study of all competes out there in the domain with this feature inlcuding user flow, heuristic evaluation, IA, user review etc.
- generate a implementation analysis of
comprehensive list of all possible ui patterns and users mental models around the feature
- Generate open questions to think out while solving
• DISCOVERABILITY
Starting visual search from Image answer
" i found a sofa I love in this photo, but I don't know what it's called. I just want to find similar ones"
Prototype tool: Figma
The challenge
Visual Search suffered from low discoverability due to gaps in awareness, capability understanding, and accessibility.
My role
Audited the Bing ecosystem to identify high-impact canvases where users naturally engaged with images, prioritizing moments where Visual Search could deliver meaningful value and drive adoption.
Defined an ecosystem-wide entry point strategy, identifying where and how Visual Search should surface based on user intent, content context, and the goals of each canvas.
Explored scalable interaction models that could adapt to different user intents—such as identification, shopping, translation, and learning—while maintaining a cohesive and recognizable experience across surfaces.
Established the experience strategy and design principles for making Visual Search discoverable through contextual, low-friction entry points, transforming it from a standalone destination into a native search modality.
Developed a scalable interaction framework that balanced consistency with context, enabling each canvas to expose Visual Search in ways that respected its primary user experience while reinforcing a unified multimodal search journey.
Partnered with Product, Engineering, and partner teams to align on priorities, evaluate implementation trade-offs, and define a roadmap for rolling out the strategy across multiple surfaces.
The impact
VS hover on SERP delivered
WAU gain at 20% flight (20-25k), projected WAU from full ship: 100-125K
IRP Hover during experimentation -
40.8k WOW gain in flight 1 and 94K WAU gain in flight 2
• DISCOVERABILITY
Starting visual search from Image detail page
" I like the jeans this celebrity is wearing, I want to find a similar pair."
Prototype tool: VS code + Claude Code
The Brief
Visual Search receives the majority of its traffic from this entry point, making it one of the most critical surfaces in the experience. How can we redirect users to VS on SERP experience without disrupting the experience.
My role
Partnered with the team owning this surface to synthesize user signals, behavoral metrics, uncovering the root causes behind declining user satisfaction despite high Visual Search adoption.
Reframed the design challenge from simply increasing engagement and accomodating redirection to - redesigning the entry point to drive adoption through a more intuitive, predictable, user-controlled, and non-disruptive interaction model.
Defined the experience strategy and design principles by translating key usability insights—ambiguous interactions, intrusive visual treatments, and lack of user control—into clear design objectives.
Explored and prototyped multiple interaction models using Claude Code to validate end-to-end behaviors, interaction patterns, and transitions before implementation.
Facilitated cross-functional design reviews with Product, Engineering, and partner teams, evaluating concepts through trade-off analysis across user value, technical feasibility, scalability, implementation complexity, and business impact.
Drove alignment on the final proposal, balancing discoverability with usability while establishing a scalable interaction pattern for future Visual Search experiences with the partner teams.
INTERACTION:
• INTENT BASED MULTI-MODAL RESULTS
Design intent based visual search results on SERP for multi-modal search
"I want to know the name of the shoes, where I can buy the exact product, see similar shoes, and copy the quote on the ad for my notes"
The Brief
Flyout improvements for text and translate scenario
overview answers for shopping scenario
exact match answer (3 sizes)
text and translate answer
My role
Visual matches answer
Designed a scalable metadata card system for Visual Matches, adding trusted product information and contextual details to improve confidence and support shopping use cases. Partnered with cross-geo teams to align on the experience and drive implementation.
Exact matches answer
I partnered closely with engineering to understand available data, technical constraints, and metadata variations, then designed a flexible list answer that surfaced the most relevant information while scaling gracefully across diverse data scenarios.
The reusable list-view card pattern was incorporated into the mainline answer framework, enabling the solution to scale across multiple scenarios and future content types with minimal design and engineering effort.
text and translate
text and translate feature was accmodated in the visual search image upload flyout. Worked with the product team to define the end to end flow for text and translate scenario
Impact
+4 experiences are shipped on SERP with positive APSAT and engagement metrics
Prototype tool: VS code + Claude Code
Influencing Product direction
The existing Visual Search experience surfaced translations only within a transient flyout, limiting discoverability, persistence, and user engagement.
I partnered with the product team to address this gap and advocate for bringing translation into the main search experience as a mainline answer. After aligning with engineering on technical feasibility, I designed the MVP for a mainline translation answer, introducing interactive elements that were ultimately incorporated into the bing answer framework as reusable patterns.
tHE FINAL ANSWER DESIGN IS IN DEVELOPMENT
• IMAGE INPUT
Reduce the friction of searching visually
"Its cumbersome to save an image to make a query, I wish I could just paste it"
Prototype tool: VS code + Claude Code
The Brief
Bing didn't support image paste as a search input, despite clipboard being a natural way users share screenshots and images. An engineering experiment validated strong user demand but negatively impacted key quality metrics, including query reformulation and DSAT. I was tasked with defining the end-to-end image paste experience, balancing ease of use with clarity, predictability, and a cohesive multimodal search journey.
My role
Benchmarked multimodal interactions across leading AI and search products to understand emerging patterns, user expectations, and opportunities for integrating image paste into Bing.
Explored multiple experience directions for introducing clipboard-based image search, balancing discoverability, familiarity, and integration with Bing's existing search workflow.
Defined the end-to-end interaction model, mapping user journeys, system states, edge cases, and failure scenarios to create a robust and predictable experience that is consistent with search bar interaction patterns on SERP.
Built interactive prototypes to validate key interaction patterns and facilitate design discussions with Engineering, accelerating feasibility assessment and implementation planning.
Scoped and prioritized the MVP, identifying the minimum set of capabilities that delivered user value while establishing a foundation for future multimodal experiences.
Impact
tHE EXPERIENCE WAS FLIGHTED WITH POSITIVE USER ENGAGEMENT BEHAVIOUR WITH
Mapping all the states and defining the interaction
What happens when there is already a query and user pastes an image
What happens if there is some load time for the image to show
What should be the next step after the image is pasted
how should the autosuggest text show when there is no query typed (should the image show with every query?)
what happens if the user presses on clear
What happens if the user clicks on backspace to clear it
what happens if the paste fails
what happens when user clicks on the image
what happens when user hovers on the image
INTERACTION:
