### Suggestions for Improving ChatGPT
Hello,
I have prepared 16 suggestions for improving ChatGPT based on my practical experience using the product. They address the mobile interface, voice interaction, conversation navigation, and certain behavioral patterns of the assistant.
Since the suggestions include specific descriptions of problems and proposed solutions, I recommend using an advanced ChatGPT analytical model for an initial assessment. The text of the suggestions can simply be copied and submitted to the model. It can evaluate each proposal in terms of user value, potential impact on the product, and possible implementation approaches.
### Suggestions
1\. Settings for opening the chat sidebar with a horizontal swipe
On mobile devices, even a slight horizontal finger movement across a conversation screen can unexpectedly open the chat list. This interrupts the user’s work and forces them to close the sidebar and return to the conversation.
Proposed solution: Add a settings toggle that allows users to enable or disable opening the chat sidebar with a horizontal swipe. This feature should be disabled by default, while remaining available to the few users who genuinely find it convenient. For everyone else, the existing large and clearly visible menu button in the upper-left corner of the screen is perfectly sufficient.
2\. Preserving voice input when the internet connection is lost
When the internet connection drops, voice input stops, and users cannot resume recording after the connection is restored. This results in lost time and dictated content.
Proposed solution: Enable local recording and intermediate transcription storage, allow users to resume dictation after the connection is restored, and preserve the completed message until it can be successfully sent.
3\. Eliminating the pattern of taking credit for users’ ideas
Sometimes the assistant repeats an idea the user has just expressed as though it were its own initiative. For example, a user suggests avoiding a particular word, and the assistant responds, “I wouldn’t use that word.”
Proposed solution: Recognize and respect the user’s authorship of suggestions. Avoid presenting the user’s ideas as the assistant’s own, and continue the discussion naturally without reintroducing an already expressed idea as a new recommendation.
4\. Reducing the overuse of the formulaic “It’s not X, it’s Y” construction
The assistant frequently uses contrasting formulations such as “It’s not a problem, it’s an opportunity” or “It’s not one thing, but another.” Excessive repetition makes its language predictable, artificial, and monotonous.
Proposed solution: Reduce the frequency of this construction, expand the variety of phrasing, and use contrasts only when they genuinely strengthen the meaning.
5\. Eliminating the pattern of undermining the value of completed work in the concluding part of responses
The assistant may deliver a high-quality, complete result and then add a closing remark that diminishes its value and creates an unpleasant impression. Examples include formal disclaimers, unjustified expressions of uncertainty, or unnecessary offers to check or improve something further when the task has already been completed.
Proposed solution: Preserve the value and integrity of completed work in the concluding part of a response. Avoid unnecessary disclaimers and self-undermining statements. If genuine limitations exist, explain them specifically without diminishing the value of the work already delivered.
6\. Control over repeated audio playback
Unintended automatic repetition of voice responses should be eliminated.
Proposed solution: Add a separate loop-playback toggle that allows users to enable or disable continuous repetition themselves.
7\. Placing the read-aloud button at the beginning of a message
Currently, users have to wait until the response generation is complete before starting playback.
Proposed solution: Place the read-aloud button at the top of the message so users can start listening as soon as the first playable text becomes available.
8\. Automatic voice playback as responses are generated
Add a setting for permanently enabling automatic playback of new responses.
Proposed solution: When this feature is enabled, playback should begin as soon as the first available text fragment appears and continue as new fragments are generated, without waiting for the entire response to be completed. A similar approach is implemented in DeepSeek.
9\. Making message actions available before generation is complete
Currently, users have to wait for the assistant to finish its entire response before accessing certain functions.
Proposed solution: Make the main actions — read aloud, copy, and share — available as soon as the corresponding content appears. Priority should be given to enabling audio playback before generation is complete.
10\. Rewinding and navigating through spoken messages
When listening to long responses, there is no convenient way to jump directly to a specific section. If users get distracted or want to listen to a particular part again, they have to rewind manually or restart playback.
Proposed solution: Add a visual playback progress bar that users can move with their finger. Include forward and backward navigation, the ability to jump to any position or to the end of the message, and the option to resume playback from the selected position.
11\. Specifying the time zone when displaying usage-limit reset times
When a usage limit is reached, ChatGPT displays the time when it will reset but does not specify the time zone. This can lead to confusion about when access will be restored.
Proposed solution: Automatically specify the time zone and display the reset time according to the user’s device-local time.
12\. Adding an upward navigation button with multiple destination options
When working with long conversations, it is useful to return quickly to previous messages. However, there is currently no sufficiently flexible upward navigation.
Proposed solution: Add an upward navigation button with three destination options:
* The beginning of the latest message.
* The beginning of the previous message.
* The beginning of the entire conversation.
The button could appear when the user scrolls upward, in the same way as the existing button for scrolling to the bottom.
13\. Adding a vertical conversation navigation slider
In long conversations, it is difficult to determine one’s position relative to the entire conversation and quickly navigate to a specific section.
Proposed solution: Add a vertical slider on the right side of the screen. It should indicate the user’s position within the entire conversation and allow quick navigation up and down, including jumps to the beginning and end.
14\. Collapsed message previews when scrolling through long conversations
In lengthy conversations, large messages take up considerable screen space, making it slower to find information and navigate through the discussion.
Proposed solution: Introduce a collapsed-message display mode with previews of approximately the first 200 characters and an option to expand each message in full. Add a setting that allows users to choose between full and collapsed message display.
15\. Duplicating the “Copy code” button at the top of code blocks
In long code blocks, the copy button is located at the bottom. Users have to scroll through the entire block to access it.
Proposed solution: Add a duplicate copy button at the top of each code block while retaining the existing button at the bottom.
16\. Making the message interaction menu easier to access
For long messages, reaching the action menu at the bottom is inconvenient. This complicates interaction with messages and requires unnecessary scrolling.
Proposed solution: Duplicate the action menu at the top of each message and/or allow users to open it by long-pressing anywhere within the message.
### Conclusion
I believe implementing these 16 suggestions could significantly improve the quality of the ChatGPT user interface and strengthen its competitive position against products such as Claude, Gemini, Grok, and Meta AI.
A more convenient and thoughtfully designed user experience directly affects customer satisfaction, loyalty, retention, and willingness to purchase and renew paid subscriptions. These proposals therefore deserve careful consideration in terms of their practical value to OpenAI.
In return, I would also appreciate being provided with an annual ChatGPT subscription for my account:
Access to the most advanced models would allow me to conduct deeper product analysis and develop stronger solutions. Where promising opportunities arise, I would also be able to share further collections of improvement proposals with the OpenAI team on a regular basis.
I would also appreciate being given an email address or the contact details of a relevant team member to whom I could send such proposals directly, ensuring efficient and reliable communication.
Best regards, Kasianov Pet