TikTok: Add a Feature - AI Search Assistant
Enhancing Content Recall & Discovery Through Conversational AI
Researcher & Designer
Hodo Yusef
Methodologies
Interviews, Surveys, Competitor Audits
Target Audience
TikTok Users (Ages 16+)
1. Defining the Problem & Opportunity
TikTok is a premier social media platform for discovering short-form video content. While TikTok introduced "Collections" to allow users to bookmark content into categories (a major upgrade from simply liking videos), significant usability friction remains.
Users frequently save videos for specific real-world references—such as recipes to cook, places to visit, or products to buy. However, as collections grow, locating a specific previously viewed or saved video becomes overwhelmingly tedious. Users continuously save new content but forget about or struggle to find past saves.
Point of View (POV) Statement:
TikTok users need an efficient way to find videos they’ve watched or saved because they use TikTok as a point of reference and require a reliable method to retrieve specific content.
How Might We (HMW) Statements:
• How might we make the process of finding saved videos easier?
• How might we enhance search features to allow for intuitive ways to locate content?
• How might we leverage conversational AI to identify user intent and match saved items?
2. Research Methodology
To evaluate user behavior, organization habits, and mental models surrounding video retrieval, a multi-method research strategy was executed:
1. User Interviews
Qualitative 1-on-1 interviews conducted to uncover behavioral reasons behind saving content, pain points, and current retrieval walkthroughs.
2. Quantitative Surveys
Distributed to active TikTok users to gather preference metrics, ease-of-use scores, and frequency of revisits.
3. Competitor Audits
Evaluated content categorization and retrieval features across Instagram, YouTube, Facebook, and Yelp.
3. Key Research Findings & Analysis
Data gathered from user interviews and survey responses revealed recurring friction points in content retrieval:
Lack of Organization
The vast majority of users do not manually categorize saved videos into collections due to high effort, leading to unorganized piles of saved content.
Reference Point Usage
Users primary save videos as an active reference point—looking for inspiration on what to cook, where to visit, or products to buy.
Retrieval Friction
Users report severe difficulty re-locating videos when they only remember visual details or general concepts rather than exact account names or video titles.
Competitor Feature Benchmark
Auditing competing platforms revealed that most apps hide saved collections deep within menus:
Platform
Saved Content Access
Search Within Saved
AI Assistant Retrieval
Instagram
Buried in Profile Menu (3 Clicks)
Basic Collections Only
None
YouTube
Library Tab (Playlists)
Keyword Search
None
TikTok (Current)
Profile Tab (Bookmarks/Collections)
Global Search Only
None
TikTok (Proposed AI)
Direct Search Bar / Chatbot
Conversational Natural Language
Integrated AI Bot
4. User Persona
Synthesized research findings into a primary user persona representing the active TikTok content collector
Maya
The Inspiration Collector • Age 22 • Active TikTok User
Background: Uses TikTok daily for entertainment and lifestyle ideas (recipes, fashion recommendations, NYC places to visit).
"I save dozens of videos every day, but when I actually need to find that one pasta recipe or jacket review, I give up because scrolling through my saves takes forever."
Pain Points
• Overwhelmed by hundreds of unorganized bookmarked videos.
• Cannot remember creator handles or exact video captions.
• Time wasted scrolling through general "Liked" and "Saved" lists.
• Cannot remember creator handles or exact video captions.
• Time wasted scrolling through general "Liked" and "Saved" lists.
Goals & Expectations
• Quickly retrieve specific recommendations using simple descriptions.
• An intelligent search bar that understands context, not just keywords.
• An intelligent search bar that understands context, not just keywords.
5. Interface Solution: AI Search Assistant
To eliminate manual collection organization, an AI-powered conversational search bot was integrated directly into TikTok's search workflow. Users can query using descriptive prompts based on memory (e.g., "video I watched about Amazon clothing").
AI Search entry point and conversational retrieval flow based on watch history and saved content.
6. Next Steps
1. Prototype Refinement
Develop interactive prototypes testing multi-turn conversational prompts and error recovery states.
2. Usability Testing
Run task-based usability testing measuring time-to-retrieval compared to manual collection browsing.
3. Search Optimization
Refine NLP (Natural Language Processing) query parsing based on common user memory attributes.
