Déjà Vu: Chat GPT Horror Story
Introduction
This project focuses on improving chat organization and history management in ChatGPT to reduce unnecessary new chat creation as our main repetitive user behavior. As users struggle to retrieve past chats, they often start fresh instead, leading to fragmented context, lower productivity, and increased data usage. By streamlining organization and retrieval, the project aims to enhance continuity, boost efficiency, and promote more sustainable digital habits.
My Role
I led the end-to-end product design process for this project, from UX research and user interviews to wireframing, prototyping, and usability testing. I translated insights from surveys and interviews into user flows and high-fidelity UI designs, ensuring every decision was user-centered and aligned with business goals.
Project Scope
A team of a Product Designers, a UI/UX Designer, and a developer.
Mentor: Marzie Nadeali
4 Weeks (Remote) – 2025
Context
It didn’t start with a big insight! It started with small, familiar moments:
“Where’s that prompt I had yesterday?” “Didn’t I already ask this?”
That single question led me to rethink how conversation history works in ChatGPT and how we might design a better way to navigate it.
Today, ChatGPT is Everyone’s Daily Assistant
ChatGPT Statistics 2025: Top Picks
*Messages = every interaction (user + AI turns).
Research
What Users Say, and How They Feel…
Common friction points users face
We combined insights from Reddit, ChatGPT forums, app reviews, and our own network. The feedback revealed recurring struggles with chat organization, history navigation, and workflow support.
Discovery through Research
Our research began with a comprehensive survey of +160 ChatGPT users, complemented by 10 in-depth interviews and contextual inquiries to establish baseline metrics and uncover key pain points. The findings revealed a scaling crisis, as usage grows, managing and retrieving information becomes increasingly difficult.
Interviews & Contextual inquiry to validate
While the survey quantified user behavior, interviews and contextual inquiries helped explain why it happens. These sessions provided qualitative depth, capturing frustrations, habits, and unmet needs behind the numbers.
struggled to find chats
Use ChatGPT for work/professional tasks.
Persona
The people behind the scenes..
Our mixed-methods research revealed four primary user patterns based on usage frequency (15-20 to 30-40+ queries daily), chat volume management, and organizational needs. These personas represent the spectrum of ChatGPT users and guided our feature prioritization for the redesign.
Competitive Analyse
Search, Structure, and Sharing: Three Universal Patterns in Conversational Platforms.
We analyzed 3 Direct competitor – ChatGPT Free, Grok, Claude, Gemini and 2 indirect Chat Platform Competitor- Google Workspace Chat, and Notion AI – to see how they organize chats and manage conversation history.
Observation
Insight
What We See
Our competitive analysis reveals three key patterns:
- Advanced search and pinning have become standard solutions.
- In-conversation navigation like chapter scrolling shifts conversations from linear streams to structured documents, helping users reference specific sections without endless scrolling.
- Collaboration features reveal the most transformative insight: AI conversations should be shared organizational knowledge, not isolated artifacts.
Ideation & prioritization
From 70 Ideas to Actionable Solutions
We generated over 70 ideas and carefully evaluated them through the three lenses of Desirability, Viability, and Feasibility.
Using the Impact–Effort Matrix, we identified and prioritized the most promising ideas for implementation.
Key Evaluation Criteria
Solution
Search Function
Simple & Advanced Search: An enhanced search system that makes finding past conversations faster, easier, and more accurate.
Impact
The Advanced Search & Semantic Options feature played a key role in optimizing chat organization. It helped users reduce new chat creation, minimize repetition, and boost retrieval success by making it easier to locate and continue relevant conversations. This shift improved the overall sense of flow and continuity in ChatGPT, turning scattered interactions into a more cohesive and efficient experience.
Collaboration Function
Collaborative Chat (Share to collaborate vs share privately) Enables multiple people to join the same conversation and collaborate in real time, turning ChatGPT into a shared thinking space.
Impact
The Collaborative Chat feature bridges the gap between individual and team use without trying to replace existing tools. By enabling lightweight co-review and discussion within shared chats, it reduces repetitive content generation and keeps collaboration contextually connected. This approach turns ChatGPT into a more efficient, shared thinking space; helping teams stay aligned while maintaining focus within the product.
This makes collaboration smoother and more organized.
Highlights
Chat Highlights lets users highlight key parts of a chat to make important info stand out and find it faster with a quick scan.
Pin
Pin Messages Users can pin messages in a conversation, keeping them visible and sticking at the top of each chat as a Pin Bar.
Chapter Scroll
Chapter Scroll A smart navigation tool that auto-generates section titles from each chat, making long threads effortless to explore.
Impact
The Highlight, Pin, and Chapter Scroll features helped reduce repeated work inside each chat by giving users simple ways to save and return to key moments in a chat. Instead of abandoning conversations and starting new ones, users could pick up where they left off. These all led to improving flow, reducing redundancy, and increasing overall engagement and satisfaction.
Mindmap
Visual Mindmap A visual tool at the top of the ChatGPT workspace that turns chats into structured, interactive maps for easier navigation.
Impact
The Visual Mind-Map feature reimagines how users organize and reflect on their chat history. By turning conversations into structured, visual overviews, it helps users recognize patterns, revisit ideas, and maintain long-term context without scrolling through endless threads. While not designed to replace search tools, it adds emotional engagement and cognitive clarity, transforming history management from a task into an intuitive, meaningful part of the ChatGPT experience.
Testing Impacts
Design clarity isn’t found in pixels, but in perception.
We conducted a 5-second test to evaluate how quickly users could understand the interface at first glance.
In parallel, we ran a preference test to measure users’ emotional response, visual preference, and overall engagement with the design.
5 Seconds Test: How Well Users Identified the Purpose of New Features
User feedback: : How Well Did Users Understand The Features
Sidebar UI Preference Test
After 5-second tests and initial feedback revealed usability issues, we redesigned UI elements and conducted a round of preference testing. Comparing different sidebar layouts and collaboration features, asking users to choose between designs with folder hierarchies and shared workspace views. This iterative approach helped identify which interface elements users found most intuitive for both finding their own past chats and collaborating with team members on shared conversations.
Insight
Users mostly preferred UI 2 for these reasons:
- Cleaner information hierarchy: sections with clear groupings (recent chats, projects, shared chats) vs. exposed chat list in UI 1
- Reduced cognitive load: Progressive disclosure through expandable folders instead of everything visible at once
- Eliminated confusion: Removed ambiguous pin/highlight icons that users found unclear (“What do the icons do?”)
- Preserved context: Users noted pins work better within conversations rather than as separate navigation elements.
Key Finding: Users prioritized functional clarity over visual features, with multiple participants noting that pins and highlights work better as contextual elements within chats rather than standalone navigation items. Also, they complained about the UI No. 2 being crowded as well. And the spacing needs improvement.
Iterations
UI Updates based on the test result
Chapter Scroll
Collaboration function
Second Test: Collaboration UI Preference Study
Insight
Users mostly preferred UI NO.1 because of these reasons:
- Clear decision hierarchy: Two distinct options presented upfront (“Share link” vs “Invite to collaborate”) making the choice obvious
- Reduced cognitive load: Simple two-step process instead of everything on one screen
- Lower error risk: Separate paths prevent accidentally granting wrong permissions
- Better information scent: Descriptive text under each option helps users understand consequences before committing
- Less overwhelming: Clean, focused interface vs dense, multi-element layout in UI NO.2
Takeaway
This project showed how design can meaningfully shape user behavior and contribute to more responsible AI interactions and sustainable LLM use. While improving organization and continuity helps reduce redundant prompts and resource use.
I also acknowledge that the environmental impact of AI extends beyond user behavior; it’s rooted in how models are trained, deployed, and scaled.
Still, thoughtful design has a role in guiding how people engage with these systems, making every interaction a bit more intentional, efficient, and responsible.
Next Step
If given more time, I would focus on refining the Visual Mind Map to make its categorization system more intuitive and easier to interpret. Improving how topics and relationships are visually represented could help users better understand and navigate their chat history.
Additionally, I’d explore integrating a Gallery View for prompts, images, voice notes, and files; creating a richer, more tangible layer of interaction that connects multimodal content within each chat.
