Hi everyone, I’d like to report a significant issue with the latest updates to the ChatGPT mobile app (Android, in my case — not sure if this applies to iOS too). Until recently, the voice dictation feature worked in a very convenient way: I could tap the microphone button, speak, and then see the transcribed text in the message box before sending. This allowed me to review, copy, or edit the text before submitting it to the chat.
However, in one of the latest updates, this behavior has changed. Now, after finishing a voice recording and pressing the send button, the message is sent immediately without showing the transcribed text beforehand. This completely removes the possibility to correct or even copy the text before it gets submitted, which is a major regression in usability — especially for those of us who use ChatGPT primarily as a voice-to-text tool rather than for direct conversation.
Additionally, if an error occurs during recording (e.g., network glitch or voice not recognized), there’s no way to reprocess or retry the transcription based on the existing recording. You have to record again from scratch, which is very frustrating.
I’ve also noticed the new “press-and-hold to speak” feature, where the message is sent as soon as you release the button. That’s fine as a new option. But I strongly believe that the previous tap-once behavior — where you could record and review the transcription — should still be available for those who need more control over what gets sent.
Please, OpenAI team: bring back the preview/edit functionality after voice input. It was one of the best features of the app, and its removal seriously degrades the experience for users like me.
Thanks for considering this. Has anyone else experienced the same issue?
Yup same issue here. It really is a questionable change.
The voice recording function now looks more fancy, but actually has less functionality. I also relied on it quite a lot, especially since I can’t trust the classic (non-advanced) voice mode anymore, whenever I speak for longer times (about 3 minutes or so), since it would almost every time just give me a network error (“our servers are having issues right now, could not understand what you said” reply) and have noting of what I said recorded.
So I started relying on the voice recorder for longer messages, basically recording for up to 2 to 3 minutes and then pressing the button for transcription, since recording for too long with the voice recorder also can give you a “network error”, where it just isn’t able of transcribing the speech at all. And after recording the 3 minutes, I could just do another voice recording in addition to the already transcribed text.
And, as mentioned, it oftentimes is really important to have the ability to check certain parts of the transcription, since errors really do happen, which can make ChatGPT misunderstand one’s input.
So overall, really an unnecessary change which solves less than it breaks. Please Open AI, reverse the change!
Agreed. The way it used to be was far more functional. And chatGPT has, by far, the best speech to text. All other applications I’ve tried for STT do not compare. I used it a lot to create complex messages in the chatGPT app because I, like everyone else, am incredibly busy. What would become a time-sink typing on my phone is instead done in a matter of seconds or a couple of minutes. I really hope they revert back or let us choose the older feature.
Yes, Javifeeble - I am experiencing the same issue and it is incredibly frustrating. You’re not exaggerating when you say this is a highly significant usability regression - I am deeply unhappy with this change and I hope OpenAI will hear us out and revert this change. As an avid user of the service and a subscriber for plus, this feature along with the others ChatGPT offers are huge for me. Please revert this change!
Hi everyone — just wanted to chime in and fully support the concerns raised here. I’m also a frustrated user who relied heavily on the original voice input features in the ChatGPT app.
The removal of the separate microphone button on iPad and the change to how voice dictation works (especially losing the ability to review or edit the transcription before sending) has been a real step backward in both usability and accessibility. Being able to dictate at my own pace, edit with voice, and build prompts incrementally was a key part of my workflow — and losing that has made the app feel much more rigid and frustrating to use.
This update has not only affected productivity, but it’s also made the app significantly less accessible for users like me who have an accent and who depend on voice input.
Please, OpenAI — bring back the separate mic button and the ability to review/edit transcribed text before sending. These were essential features that made the app powerful, flexible, and user-friendly.
Thanks to those who raised this — I hope the team is listening.
I also got this new version for a couple of days and now, after another app update it’s gone back to how it was. I’m on the Android beta program, so I’m not sure if that’s why I got it.
Anyway, thank goodness they reverted it back to how it was. As described here, the new version was really not good. Also, I feel like the transcription itself, despite notoriously faster, was not as accurate. I got several mistakes in the transcription, whereas usually I never do… Anyone knows of any details on this side of things?
Yes, this is a major problem for the way I use ChatGPT. I definitely want to read the message before sending. Or I want to add to it. It worked well and now it doesn’t. Worse: it is confusing because it’s such a big change from before. And there are now three similar voice modes (dictate+send, advanced voice, classic conversation voice mode) and the app does not provide a good explanation of what each does.
How to fix it?
I suggest adding a simple checkbox to Settings/Voice that will say “Auto-send dictated message”. If it is switched on, message is sent immediately after dictated. If it is deactivated, text is transcribed but stays in the text field and user can edit it or add more text or add more dictation. That last one is pretty useful and was possible just a week ago.
I’m very glad others have brought this up – I thought I was the only one who missed the previous behavior.
I completely agree with everything you’ve described. I’m also using the ChatGPT app on Android, and this recent change to the voice dictation function has made it much harder to use in the way I was used to.
Previously, I loved being able to dictate several thoughts, see the transcribed text, edit or build on it, and then send the message when I was ready. That workflow was perfect for someone like me who thinks best when speaking and needs a moment to organize ideas before submitting them.
Now, with messages being sent immediately after voice input, it feels rushed and inflexible. It’s especially frustrating when there’s a small error, and I can’t even fix it before it’s sent.
Please bring back the ability to review and edit the transcription before sending. Even as an optional setting, it would make a huge difference for users who rely on voice input for thoughtful communication.
Thanks to everyone here for expressing this so clearly.
Absolutely, I made an account for this forum just to issue this complaint. You summarized the issue completely, so I’m just here to add volume to the complaint.
Just adding my complaint. I never come to forums but this is so bad! I feel like someone has chopped off my left arm!
Losing the ability to edit before sending has made it far less usable for me, and it’s honestly ruined my workflow. I hope you’ll reconsider this change or provide an alternative.
I have been one of the vocal users about the problem, opening a thread about it and commenting on some posts. But, for the last few days, the app has gone back to its old functionality at least in my case on Android, which honestly made me happy.
The voice dictation feature now disables the Deep Research mode, so it is impossible to ask for a Deep Research using a dictated message.
If transcription fails for whatever reason, the message is lost. The previous version allowed to try again, now it just tells you it lost the message.
The main problem remains: if the transcription is successful, it still sends it immediately without the user having any chance to read it first, add to it or correct misspelled words.
This is made it pretty much impossible to use ChatGPT on the phone. Can you please fix it by adding an option to the Settings to use Classic dictation (or whatever you choose to call it)? This new way is just objectively not working well.
Well, I won’t deny that it feels really strange to me, since for me the change reverted two days after I opened a thread and posted a couple of messages in the forum. In fact, I got pretty excited thinking it was the developers who had read the posts and decided to bring back the new voice dictation mode. Right now, the interface shows that blue bar that appeared when they first introduced the change, but once I send the voice message, it lets me check it before actually sending. I insist, I’m using the latest version of the app.
It’s also worth mentioning that I’m a beta tester for the app, so maybe that has something to do with it. Maybe the newest changes haven’t rolled out for me yet, or I’m on a different testing track. I don’t know. I just hope it stays the way it is now, or at least that they give us the option to switch it, honestly.
And just to be clear—this skepticism isn’t meant to imply that anyone’s lying. It’s just that, well, it strikes me as odd because, in my case, the change went back to the normal state.
Hi guys, just saw the update on Android device for the chat app.
The voice typing feature has changed. Now I see a sort of a blue interface when it is listening to me with a button to send. That’s not practical. Because you cannot edit the miss-recognized text (for someone like me who has an accent in English that’s a bummer).
Previous interface had a close button which converted the voice into text inserting it into the text area. Which allowed you to combine manually typed text with a voice recognition text and edit before sending.
Now this ability is gone with the new interface which makes the app less practical.
This auto-send voice-to-text change is a UX disaster.
There’s no review, no chance to fix errors, no option to cancel. Just automatic send the moment I stop speaking, like the app assumes I’m always ready to ship half-formed thoughts. This is a tool built on precision and you’ve stripped away the ability to control that.
There was a moment where it reverted. Text showed up before sending, like it used to. It worked. Then it was yanked again without explanation.
Not every message is ready the second I stop talking. Sometimes I need to pause, rephrase, or scrap the whole thing. That’s how real communication works. I shouldn’t have to play whack-a-mole with the stop button just to avoid sending something unfinished.
If this is A/B testing, it’s a failure. If it’s intentional, it’s out of touch. If it’s a rollback, it’s poorly communicated.
Give us the option. Let people choose whether voice messages send automatically. You made it worse and gave no way to fix it. That’s not progress. It’s sloppiness.
I’m another advanced user affected by this. I used ChatGPT while training, speaking ideas out loud The system used to convert my voice to text, then I could edit it, expand it, and send a powerful, structured message. That was my entire thinking workflow. Now that’s broken. The new update sends the voice input immediately and removes all control Please bring back the edit-before-sending feature — it’s critical for high-performance users Not everyone uses ChatGPT like a chatbot. Some of us build real-time cognition here.
