Tencent- QQ Browser AI Search Experience Design

Internships work in tencent, Designed AI-powered search and Q&A experiences, focusing on structuring query responses and optimizing interaction models.

Company

Tencent

Timeline

May – Aug 2025

Industry

Technology

Role

Product Design Intern

Background

Users often faced long, unstructured AI answers that were difficult to parse.

  • Lack of interaction entry points beyond text input — limited ability for users to refine or expand answers.

  • Users didn’t always trust the AI’s response due to unclear sources and missing context.

  • In complex queries (e.g., multi-step or comparative tasks), users struggled to continue the flow without restarting the query.

Goal

Improve clarity, usability, and trust in AI-generated answers, especially in complex information-seeking scenarios.

Problem Space

  • Users often faced long, unstructured AI answers that were difficult to parse.

  • Lack of interaction entry points beyond text input — limited ability for users to refine or expand answers.

  • Users didn’t always trust the AI’s response due to unclear sources and missing context.

  • In complex queries (e.g., multi-step or comparative tasks), users struggled to continue the flow without restarting the query

Research & Analysis


  • Task Flow Audit: Mapped typical search → Q&A flows in QQ Browser. Identified friction at the “refine” and “multi-step” stages.

  • Competitive Benchmarking: Compared with Google, Bing Copilot, Perplexity, and domestic competitors like Quark, Baidu.

    • International tools excel at structured modules (charts, lists, cards).

    • Domestic products lean toward rich media integration (short video, shopping links, medical cards).

  • User Insights:

    • Users wanted faster scanning (structured formats).

    • Preferred follow-up queries embedded in the answer itself.

    • Needed confidence in reliability (sources, official endorsements).

Design Strategy & Iterations

Before iterating on design, we first audited the query landscape.

  • Collected both high-frequency queries and sampled long-tail queries.

  • Classified them based on user needs and product goals.

  • Identified high-value verticals (high QV) and answer intents that had significant optimization potential.

Key Optimization Categories

Design Exploration – Basic Formats

Goal: Address the limitation of text-only answers, which often fail to deliver clarity and intuitiveness.

Introduce rich media formats to adapt to different types of queries.

Design Exploration – Data Transformation

Q: “How much does the world’s largest mandrill weigh?”

Goal

  • Users struggle to interpret abstract weight numbers without context.

  • Need to make the information more relatable and visual.

Approach

  • Introduced reference objects (e.g., compare to familiar animals).

  • Added visual model of mandrill with key attributes (weight, height).

  • Provided a more direct and intuitive perception of size.

Q: ““How many calories are in a McDonald’s hamburger?”

Goal

  • Text-only calorie numbers are vague and not memorable.

  • Users need clear breakdown of components to trust and understand the answer.

Approach

  • Transformed whole calorie value into a layered visualization.

  • Each ingredient shown as a stack with its calorie contribution.

  • Allowed users to grasp total calories and composition at a glance.

Connecting Information to External Actions

Upgrading from static answers to actionable solutions through interaction and contextual links.

Case 6: Symptom Query → Actionable Interaction

Q: “How to relieve anxiety symptoms?”

Goal

  • Current text-only answers feel abstract and passive.

  • Users need immediate, guided actions instead of static info.

Approach

  • Transformed answer into a multi-sensory interaction (e.g., “Breathe In / Out” animation).

  • Enabled users to engage directly with the advice, turning knowledge into action.

  • Created a more natural and immersive flow, reducing hesitation.

Case 7: Dream Interpretation Query → Visualized Emotional Guidance

Goal

  • Text-heavy interpretations are boring, hard to scan, and lack personalization.

  • Users want clear, digestible insights with emotional context.

Approach

  • Extracted keywords from the dream text.

  • Assigned emotional values (positive/negative), paired with color-coded graphics.

  • Displayed concise cards (e.g., “Wealth Rising” / “Pay Attention to Health”) with icon-based CTA.

  • Made dream interpretations feel more visual, contextual, and trustworthy.

Case 8: Travel Query → Tool Integration

Q:Question
“Can this MUJI suitcase be taken on a flight?”

Goal
Text-only answers with dimensions were hard to interpret.
Users needed a more direct, visual, and verifiable way to check compliance.

Approach

  • Added comparison charts against airline size limits.

  • Integrated an AR measurement tool so users could scan their suitcase directly.

  • Turned abstract numbers into an instant, actionable check in context.

Case9 –Government Query → Structured Process Guidance

Goal

  • Text-only answers were lengthy and hard to follow.

  • Users needed clear steps and direct access to processing entry points.

  • Reduce repeated searches caused by vague instructions.

Approach

  • Broke down text into structured step cards (e.g., “Prepare documents → Submit application → Collect permit”).

  • Strengthened guidance with dual pathways: information + direct service entry points.

  • Integrated service center details and links, reducing user effort and supporting full task completion in one flow.

Case10: Entertainment Query → Resource-Linked Experience

Goal:

  • Transformed “passive list” into actionable recommendations.

  • Improved user satisfaction by reducing extra steps.

Approach:

  • Resource Embedding: Attach video resources (rating, trailer, platform links) directly under each title.

  • Service Closure Loop: Enable one-click viewing or subscription inside the card.