Role
Role
Role
UX/UI Designer
UX/UI Designer
UX/UI Designer
Team
Team
Team
2 Designers
2 Designers
2 Designers
Tools
Tools
Tools
Figma, Lovable,
Replit
Figma, Lovable,
Replit
Duration
Duration
Duration
5 Weeks
5 Weeks
5 Weeks
Deliverables
Deliverables
Deliverables
High-Fidelity Wireframes
& Prototypes
HF Wireframes
& Prototypes
HF Wireframes
& Prototypes
ABOUT CHATGPT
Let’s Get A Quick Stats Snapshot First!
INTRODUCTION
ChatGPT Is More Than An AI Tool, So Finding Past Chats Should Be Easy. Right?!
It’s a daily assistant for millions of people. From drafting emails and brainstorming ideas to planning trips and even casual conversations, people rely on it for both professional and personal needs. But as usage grows, so do the challenges. Managing past conversations, organizing useful chats, and retrieving important information can quickly become overwhelming. This made us think and find the right solutions.
PROBLEM BREAKDOWN
How To Make Chat Organization And History Management More Intuitive To Ensure Users Navigate With Ease, Maintain Context And Revisit Past Conversations!
Many users engage in several chats on different topics, including personal, professional, and creative ones, but struggle to keep them organized. Without organizational tools, it’s difficult to quickly find and refer back to a specific conversation, especially if it’s buried among many others. This can lead to lost context, repeated questions, or confusion.
PERSONA
Let’s Meet Users Behind The Prompts Who Struggle With Challenges!
How Our Personas Currently Organize Their Chats!
Today, our personas save ChatGPT conversations externally, resulting in scattered storage, lost context, and difficult retrieval. They also rely on visual or verbal workarounds (screenshots, tags, etc.), but ultimately need an integrated ChatGPT solution that enables faster, smarter revisiting with less manual effort and more control in the AI era.
OUR GOAL IS
Increasing User Engagement With Past Chats By Designing AI-Powered Features To Help Users Find Past Conversations Conveniently And Organize Them Effectively And Efficiently!
BUY-IN METRICS
Success? It’s All About Moving The Needle On The Following Metrics...
If this were implemented in a live product, metrics like revisit rates, prompt repetition, and retrieval time could offer strong signals of success. In our case, impact came from user voices.
North Star Metrics:
% of users re-engaging with past conversations
% of users who actively use smart search and organization features weekly (e.g., TOC, tags, folders, search)
Other Supporting Metrics:
Reduction in time spent searching for past chats
Increase in retrieval success rate
Retention rate for thought-partner seekers
North Star Metrics:
% of users re-engaging with past conversations
% of users who actively use smart search and organization features weekly (e.g., TOC, tags, folders, search)
Other Supporting Metrics:
Reduction in time spent searching for past chats
Increase in retrieval success rate
Retention rate for thought-partner seekers
North Star Metrics:
% of users re-engaging with past conversations
% of users who actively use smart search and organization features weekly (e.g., TOC, tags, folders, search)
Other Supporting Metrics:
Reduction in time spent searching for past chats
Increase in retrieval success rate
Retention rate for thought-partner seekers
ASSUMPTIONS
Our Initial Guesses & Thoughts
At first, we were just taking shots in the dark, not really sure what would stick. So, we dove into every source we could find, trying to piece things together and we came up with some assumptions as follows:
At first, we were just taking shots in the dark, not really sure what would stick. So, we dove into every source we could find, trying to piece things together and we came up with some assumptions as follows:
At first, we were just taking shots in the dark, not really sure what would stick. So, we dove into every source we could find, trying to piece things together and we came up with some assumptions as follows:
Then we conducted secondary research and took a look at online sources, user comments, etc. to see if we need to add any other points.
Then we conducted secondary research and took a look at online sources, user comments, etc. to see if we need to add any other points.
Then we conducted secondary research and took a look at online sources, user comments, etc. to see if we need to add any other points.
DESK RESEARCH
Here Is The Reality...
Search functionality plays a critical role in shaping the overall user experience. A poorly designed search experience can drive frustration, while an intuitive search tool can enhance satisfaction and keep users engaged. Beyond its usability benefits, adding search to chat history (assuming text search is computationally cheaper than LLM operations) will result in:
Search functionality plays a critical role in shaping the overall user experience. A poorly designed search experience can drive frustration, while an intuitive search tool can enhance satisfaction and keep users engaged. Beyond its usability benefits, adding search to chat history (assuming text search is computationally cheaper than LLM operations) will result in:
Search functionality plays a critical role in shaping the overall user experience. A poorly designed search experience can drive frustration, while an intuitive search tool can enhance satisfaction and keep users engaged. Beyond its usability benefits, adding search to chat history (assuming text search is computationally cheaper than LLM operations) will result in:
Also, we need to take a look at the risk of hallucinations in chat organization.
If ChatGPT generates incorrect or misleading information (hallucinations), users may struggle to organize their chats effectively. They might have to re-validate past responses or manually sort out useful information from errors, making chat history cluttered and unreliable.
In 2023, research by Vectara found that chatbots invented information in 3% to 27% of news summaries, even in a simple summarization task (The New York Times, 2023).
Also, we need to take a look at the risk of hallucinations in chat organization.
If ChatGPT generates incorrect or misleading information (hallucinations), users may struggle to organize their chats effectively. They might have to re-validate past responses or manually sort out useful information from errors, making chat history cluttered and unreliable.
In 2023, research by Vectara found that chatbots invented information in 3% to 27% of news summaries, even in a simple summarization task (The New York Times, 2023).
Also, we need to take a look at the risk of hallucinations in chat organization.
If ChatGPT generates incorrect or misleading information (hallucinations), users may struggle to organize their chats effectively. They might have to re-validate past responses or manually sort out useful information from errors, making chat history cluttered and unreliable.
In 2023, research by Vectara found that chatbots invented information in 3% to 27% of news summaries, even in a simple summarization task (The New York Times, 2023).
EXPLORE THE VOICES OF USERS
Top Reported Issues: No Way to Sort And Keep Conversations Organized And Managed; Unclear Auto-Generated Titles, And So On
To identify the main issues users face with chat organization on ChatGPT, we reviewed comments on the OpenAI Community and other websites. Here are a few snapshots of users’ comments:
To identify the main issues users face with chat organization on ChatGPT, we reviewed comments on the OpenAI Community and other websites. Here are a few snapshots of users’ comments:
To identify the main issues users face with chat organization on ChatGPT, we reviewed comments on the OpenAI Community and other websites. Here are a few snapshots of users’ comments:
Then We Summarized Key Issues By Stepping Into Users' Shoes
CONTEXTUAL INQUIRY
Users Need Better Visibility and AI-Powered Solutions for Search and Organization Challenges Inside The Platform!
We conducted 10 contextual inquiries with both free and premium ChatGPT users across laptop and mobile app versions, gathering key insights.
We conducted 10 contextual inquiries with both free and premium ChatGPT users across laptop and mobile app versions, gathering key insights.
We conducted 10 contextual inquiries with both free and premium ChatGPT users across laptop and mobile app versions, gathering key insights.
COMPETITIVE ANALYSIS
As A Pioneer, ChatGPT Still Needs To Address Gaps In Chat Organization By Leveraging AI To Deliver Smarter Features And A More Integrated Experience!
We learned that the trend is moving toward smarter, more organized, and integrated chat experiences. Tools like Claude focus on context and memory, Notion and Slack offer structured organization, and Gemini and Copilot emphasize cross-tool integration and smart search.
We learned that the trend is moving toward smarter, more organized, and integrated chat experiences. Tools like Claude focus on context and memory, Notion and Slack offer structured organization, and Gemini and Copilot emphasize cross-tool integration and smart search.
We learned that the trend is moving toward smarter, more organized, and integrated chat experiences. Tools like Claude focus on context and memory, Notion and Slack offer structured organization, and Gemini and Copilot emphasize cross-tool integration and smart search.
IDEATION
Mapping Out The Most Impactful Solutions: From AI-Powered Table of Contents, Visual Timeline To Customization Features Including Tags, Pins, Etc.
After brainstorming ideas inspired by our personas, we sketched them out on sticky notes and mapped them onto an effort-impact matrix. This helped us prioritize what to build first. Using that map, we focused on ideas that struck the right balance: lightweight to implement, heavy with user value. Features like “Recent & Frequent” and “Auto-tagging” rose to the top by promising clarity without complexity and Table of Contents was ranked the most useful and helpful feature by users.
After brainstorming ideas inspired by our personas, we sketched them out on sticky notes and mapped them onto an effort-impact matrix. This helped us prioritize what to build first. Using that map, we focused on ideas that struck the right balance: lightweight to implement, heavy with user value. Features like “Recent & Frequent” and “Auto-tagging” rose to the top by promising clarity without complexity and Table of Contents was ranked the most useful and helpful feature by users.
After brainstorming ideas inspired by our personas, we sketched them out on sticky notes and mapped them onto an effort-impact matrix. This helped us prioritize what to build first. Using that map, we focused on ideas that struck the right balance: lightweight to implement, heavy with user value. Features like “Recent & Frequent” and “Auto-tagging” rose to the top by promising clarity without complexity and Table of Contents was ranked the most useful and helpful feature by users.
VISUAL DESIGN
Let’s Explore How We Brought Our Ideas To Life Through A Tangible Interface.
In the visual design, we focused on simplicity, modern aesthetics, and consistent branding to create a seamless experience tailored for ChatGPT users.
In the visual design, we focused on simplicity, modern aesthetics, and consistent branding to create a seamless experience tailored for ChatGPT users.
In the visual design, we focused on simplicity, modern aesthetics, and consistent branding to create a seamless experience tailored for ChatGPT users.
Recommending Past Relevant Chats When Starting A New One!
When the user starts typing, a box appears that temporarily hides the chat history. Users can click to reveal the messages at any time. This reduces friction for those switching between sessions and supports memory recall for users with fragmented usage patterns.
When the user starts typing, a box appears that temporarily hides the chat history. Users can click to reveal the messages at any time. This reduces friction for those switching between sessions and supports memory recall for users with fragmented usage patterns.
When the user starts typing, a box appears that temporarily hides the chat history. Users can click to reveal the messages at any time. This reduces friction for those switching between sessions and supports memory recall for users with fragmented usage patterns.
Visual Knowledge Map
AI automatically groups related topics and assigns clear titles, organizing subtopics under broader categories. This provides users with a structured and engaging overview of their past conversations with ChatGPT. The improved organization makes it easier to manage chat history, especially for users engaged in learning or ongoing projects. A floating button at the bottom of the screen allows access to this view; when hovered over, a tooltip appears to guide users on how it works.
AI automatically groups related topics and assigns clear titles, organizing subtopics under broader categories. This provides users with a structured and engaging overview of their past conversations with ChatGPT. The improved organization makes it easier to manage chat history, especially for users engaged in learning or ongoing projects. A floating button at the bottom of the screen allows access to this view; when hovered over, a tooltip appears to guide users on how it works.
AI automatically groups related topics and assigns clear titles, organizing subtopics under broader categories. This provides users with a structured and engaging overview of their past conversations with ChatGPT. The improved organization makes it easier to manage chat history, especially for users engaged in learning or ongoing projects. A floating button at the bottom of the screen allows access to this view; when hovered over, a tooltip appears to guide users on how it works.
Table of Contents With Timestamps
We designed a table of contents with timestamps in long chats, allowing users to easily navigate, revisit previous messages, and track their progress. Users can toggle this feature on or off based on their needs; the toggle is located on the right side of the top navigation bar.
We designed a table of contents with timestamps in long chats, allowing users to easily navigate, revisit previous messages, and track their progress. Users can toggle this feature on or off based on their needs; the toggle is located on the right side of the top navigation bar.
We designed a table of contents with timestamps in long chats, allowing users to easily navigate, revisit previous messages, and track their progress. Users can toggle this feature on or off based on their needs; the toggle is located on the right side of the top navigation bar.
Visual Customization
The user can click the settings icon for a specific conversation, then select Appearance from the menu. A settings panel will appear as an overlay, allowing the user to choose from emojis, icons, and colors, or upload a custom image. At the bottom of the panel, a preview of the final visual is displayed before clicking Save.
The user can click the settings icon for a specific conversation, then select Appearance from the menu. A settings panel will appear as an overlay, allowing the user to choose from emojis, icons, and colors, or upload a custom image. At the bottom of the panel, a preview of the final visual is displayed before clicking Save.
The user can click the settings icon for a specific conversation, then select Appearance from the menu. A settings panel will appear as an overlay, allowing the user to choose from emojis, icons, and colors, or upload a custom image. At the bottom of the panel, a preview of the final visual is displayed before clicking Save.
Recent & Frequent Panel
We redesigned the sidebar by adding a Recent & Frequent panel. This feature reduces the time and cognitive load required to find relevant chats, especially for users juggling multiple tasks or themes. It brings both speed and familiarity to the forefront of the interface.
We redesigned the sidebar by adding a Recent & Frequent panel. This feature reduces the time and cognitive load required to find relevant chats, especially for users juggling multiple tasks or themes. It brings both speed and familiarity to the forefront of the interface.
We redesigned the sidebar by adding a Recent & Frequent panel. This feature reduces the time and cognitive load required to find relevant chats, especially for users juggling multiple tasks or themes. It brings both speed and familiarity to the forefront of the interface.
Floating Top Panel (Chat, Files, Pins)
We designed a fixed top section that remains visible as users scroll through the chat. It offers quick access to the main chat, shared files (such as images and PDFs), and pinned messages specific to each conversation.
When users hover over a message, a pin icon appears. Clicking it adds the message to the pinned section, allowing users to easily save and reference important information.
We designed a fixed top section that remains visible as users scroll through the chat. It offers quick access to the main chat, shared files (such as images and PDFs), and pinned messages specific to each conversation.
When users hover over a message, a pin icon appears. Clicking it adds the message to the pinned section, allowing users to easily save and reference important information.
We designed a fixed top section that remains visible as users scroll through the chat. It offers quick access to the main chat, shared files (such as images and PDFs), and pinned messages specific to each conversation.
When users hover over a message, a pin icon appears. Clicking it adds the message to the pinned section, allowing users to easily save and reference important information.
Tag Your Way: Smart Suggestions + Custom Options
Chats and individual messages can be tagged for better organization. Tags are AI-generated based on content, but users can remove or customize them as needed. When a user hovers over a ChatGPT response, a tag icon appears at the bottom; clicking it opens a menu where they can select from AI-suggested tags or create their own.
Chats and individual messages can be tagged for better organization. Tags are AI-generated based on content, but users can remove or customize them as needed. When a user hovers over a ChatGPT response, a tag icon appears at the bottom; clicking it opens a menu where they can select from AI-suggested tags or create their own.
Chats and individual messages can be tagged for better organization. Tags are AI-generated based on content, but users can remove or customize them as needed. When a user hovers over a ChatGPT response, a tag icon appears at the bottom; clicking it opens a menu where they can select from AI-suggested tags or create their own.
When a user hovers over a conversation tag chip, they can see how many conversations share that tag and have the option to filter by it.
When a user hovers over a conversation tag chip, they can see how many conversations share that tag and have the option to filter by it.
When a user hovers over a conversation tag chip, they can see how many conversations share that tag and have the option to filter by it.
5-SECOND TEST
Checking If Our Ideas Made Sense And If Users Could Feel The Difference!
We conducted a 5-second test to validate first impressions of the new Knowledge Map idea. For other features, we returned to our original users and asked, “Does this fix what you told us?” Their answers helped shape what stayed and what changed.
Everyone tested found real value in the feature, and what they wanted most was a way to look back and see how their thinking evolved — not just find things faster. Most people also rated it highly useful, showing solid excitement about the idea overall.
We conducted a 5-second test to validate first impressions of the new Knowledge Map idea. For other features, we returned to our original users and asked, “Does this fix what you told us?” Their answers helped shape what stayed and what changed.
Everyone tested found real value in the feature, and what they wanted most was a way to look back and see how their thinking evolved — not just find things faster. Most people also rated it highly useful, showing solid excitement about the idea overall.
We conducted a 5-second test to validate first impressions of the new Knowledge Map idea. For other features, we returned to our original users and asked, “Does this fix what you told us?” Their answers helped shape what stayed and what changed.
Everyone tested found real value in the feature, and what they wanted most was a way to look back and see how their thinking evolved — not just find things faster. Most people also rated it highly useful, showing solid excitement about the idea overall.
ITERATION
What We Thought Was Done Until Users Spoke!
As we tested early concepts, small friction points emerged that had a big impact on usability. We responded with targeted iterations to improve clarity, privacy, and flow:
As we tested early concepts, small friction points emerged that had a big impact on usability. We responded with targeted iterations to improve clarity, privacy, and flow:
As we tested early concepts, small friction points emerged that had a big impact on usability. We responded with targeted iterations to improve clarity, privacy, and flow:
Redesign The Smart Suggestion Section
Before
In the initial design, when a user enters a word, a list of previously related chats appears.
Before
In the initial design, when a user enters a word, a list of previously related chats appears.
Before
In the initial design, when a user enters a word, a list of previously related chats appears.
After
User feedback about privacy concerns prompted us to add a pop-up before displaying past chats.
After
User feedback about privacy concerns prompted us to add a pop-up before displaying past chats.
After
User feedback about privacy concerns prompted us to add a pop-up before displaying past chats.
Merging Timeline And Table Of Contents For A Seamless User Experience!
Before
Separate Timeline and Table of Contents Caused Confusion and Redundancy.
Before
Separate Timeline and Table of Contents Caused Confusion and Redundancy.
Before
Separate Timeline and Table of Contents Caused Confusion and Redundancy.
After
Combined Timeline and Table of Contents into One Seamless Navigation Tool.
After
Combined Timeline and Table of Contents into One Seamless Navigation Tool.
After
Combined Timeline and Table of Contents into One Seamless Navigation Tool.
IMPACT
Hearing What Worked And What We Would Measure If This Went Live!
Since we don’t have access to real usage data, we relied on qualitative feedback to evaluate our ideas. But if this were implemented in a live product, metrics like revisit rates, prompt repetition, and retrieval time could offer strong signals of success. In our case, impact came from user voices.
Since we don’t have access to real usage data, we relied on qualitative feedback to evaluate our ideas. But if this were implemented in a live product, metrics like revisit rates, prompt repetition, and retrieval time could offer strong signals of success. In our case, impact came from user voices.
Since we don’t have access to real usage data, we relied on qualitative feedback to evaluate our ideas. But if this were implemented in a live product, metrics like revisit rates, prompt repetition, and retrieval time could offer strong signals of success. In our case, impact came from user voices.
RISKS & TRADE-OFFS
Risks And Trade-offs For OpenAI And Product Teams
Although the final designs received positive user feedback, implementing them still poses certain risks that stakeholders need to consider. Below, we’ve outlined some potential challenges and trade-offs the OpenAI team may face during implementation.
Although the final designs received positive user feedback, implementing them still poses certain risks that stakeholders need to consider. Below, we’ve outlined some potential challenges and trade-offs the OpenAI team may face during implementation.
Although the final designs received positive user feedback, implementing them still poses certain risks that stakeholders need to consider. Below, we’ve outlined some potential challenges and trade-offs the OpenAI team may face during implementation.
01
01
01
Technical Debt
Technical Debt
Technical Debt
Complex features like clustering, tagging, and visual maps require long-term AI and UI maintenance.
Potential Impact: Higher development and maintenance costs.
Complex features like clustering, tagging, and visual maps require long-term AI and UI maintenance.
Potential Impact: Higher development and maintenance costs.
Complex features like clustering, tagging, and visual maps require long-term AI and UI maintenance.
Potential Impact: Higher development and maintenance costs.
02
02
02
Scalability Issues
Scalability Issues
Scalability Issues
Systems like “Recent & Frequent” and Knowledge Maps require real-time performance and efficient data handling.
Potential Impact: Increased infrastructure cost, performance lags.
Systems like “Recent & Frequent” and Knowledge Maps require real-time performance and efficient data handling.
Potential Impact: Increased infrastructure cost, performance lags.
Systems like “Recent & Frequent” and Knowledge Maps require real-time performance and efficient data handling.
Potential Impact: Increased infrastructure cost, performance lags.
03
03
03
User Segmentation Drift
User Segmentation Drift
User Segmentation Drift
Advanced features could make things feel too complicated for casual users or first-timers.
Potential Impact: Widening gap in user satisfaction across personas.
Advanced features could make things feel too complicated for casual users or first-timers.
Potential Impact: Widening gap in user satisfaction across personas.
Advanced features could make things feel too complicated for casual users or first-timers.
Potential Impact: Widening gap in user satisfaction across personas.
04
04
04
Mismatched Expectations
Mismatched Expectations
Mismatched Expectations
If manual tools like deleting parts of a chat or editing tags don’t work perfectly, users might lose trust or get frustrated.
Potential Impact: High support load if users run into inconsistencies.
If manual tools like deleting parts of a chat or editing tags don’t work perfectly, users might lose trust or get frustrated.
Potential Impact: High support load if users run into inconsistencies.
If manual tools like deleting parts of a chat or editing tags don’t work perfectly, users might lose trust or get frustrated.
Potential Impact: High support load if users run into inconsistencies.
05
05
05
Legal Compliance
Legal Compliance
Legal Compliance
Any retention, deletion, or AI suggestion mechanisms tied to user data must be compliant with GDPR/CCPA.
Potential Impact: Legal and reputational risks if improperly handled.
Any retention, deletion, or AI suggestion mechanisms tied to user data must be compliant with GDPR/CCPA.
Potential Impact: Legal and reputational risks if improperly handled.
Any retention, deletion, or AI suggestion mechanisms tied to user data must be compliant with GDPR/CCPA.
Potential Impact: Legal and reputational risks if improperly handled.
REFLECTION
Good Organization Feels Invisible Until You Need It!
Three lessons shaped this project.
01
01
01
Designing For Organization, Not Overload
We learned that users don’t want more control; they want less chaos. Our goal shifted from “more options” to “smarter defaults” that make organization feel natural and seamless.
We learned that users don’t want more control; they want less chaos. Our goal shifted from “more options” to “smarter defaults” that make organization feel natural and seamless.
We learned that users don’t want more control; they want less chaos. Our goal shifted from “more options” to “smarter defaults” that make organization feel natural and seamless.
02
02
02
Prototyping With Purpose
Introducing multiple features (tags, pins, the knowledge map and the table of contents) helped us see how interconnected a retrieval system really is. Each element played a role in reducing friction.
Introducing multiple features (tags, pins, the knowledge map and the table of contents) helped us see how interconnected a retrieval system really is. Each element played a role in reducing friction.
Introducing multiple features (tags, pins, the knowledge map and the table of contents) helped us see how interconnected a retrieval system really is. Each element played a role in reducing friction.
03
03
03
Listening Shapes The Roadmap
User feedback changed our direction more than once, from collapsing smart suggestions for privacy to adjusting floating button behavior. Real needs helped us fine-tune what actually mattered.
User feedback changed our direction more than once, from collapsing smart suggestions for privacy to adjusting floating button behavior. Real needs helped us fine-tune what actually mattered.
User feedback changed our direction more than once, from collapsing smart suggestions for privacy to adjusting floating button behavior. Real needs helped us fine-tune what actually mattered.
NEXT STEPS
Looking Ahead, We Are Excited To Explore ...
Copyright 2026