The removal of the persistent “Read Aloud” speaker button is a major accessibility and workflow failure.
Accessibility impact: Many users (including those with vision impairments, reading difficulties, or focus/attention challenges) rely on continuous audio playback to consume content. By tying audio to the active chat window and stopping it mid-sentence if I copy, paste, or navigate, you’ve effectively removed ChatGPT’s usability for anyone who needs hands-free listening. This is a massive regression in accessibility.
Productivity impact: The old persistent read-aloud enabled true multitasking. I could listen while taking notes, copying instructions, or switching tabs. That’s how I worked with ChatGPT every day. Now, the moment I try to interact, playback cuts off. This destroys the flow and makes ChatGPT unusable for real work.
This is not a small tweak — it’s a devastating screw-up that undermines both accessibility and productivity. If the goal is to make ChatGPT more useful, this change does the opposite.
Please restore the persistent speaker button or add an option for continuous background playback. Without it, Chat GPT is no longer functional for many of us who paid for Plus and depended on this feature.
My report:
The accessibility barrier occurs in the ChatGPT web interface following the September 4, 2025 update.
Barrier: The “Read Aloud” feature has been moved into the overflow (⋯) menu. Playback immediately stops if the menu loses focus, which prevents users who rely on text-to-speech from listening to responses while navigating or multitasking. Previously, Read Aloud was accessible as a persistent, pinned control, which allowed uninterrupted use.
Impact: This change creates a major barrier for disabled users who rely on Read Aloud for accessibility. It effectively breaks continuous access to ChatGPT’s content. Accessibility features should not be hidden in transient menus or tied to menu focus.
-–
In addition the moving of other elements such as the Project Instructions to a small ••• in the top right as well as the collapsing of files to a small section above the right of the new chat field not being noted in the Sept 4th change log left those of use that relay on the ARIA markers trying to figure out what else changed. Those are just what I have found so far as I treat my Project Instructions as a living memory.
UX changes need to be screened for potential friction before deployment. GPT-5-Thinking can even point out “some” of these friction points if you run the code and screenshots through it against the WCAG Accessibility guidelines. Below it was able to pull 3 of the 5 guidelines I was going to cite. So the model(s) itself could have helped flag this before deployment.
Relevant Accessibility Guidance:
WCAG 2.1.1 & 2.1.2 (Keyboard accessibility) — features must remain operable without reliance on pointer-specific actions.
WCAG 2.1.3 (Media alternative for text) — text-to-speech functions serve as a non-synced media alternative, and should be consistently available.
WCAG 2.5.2 (Pointer cancellation/abort) — functionality tied to a transient menu risks unintentional interruption of critical assistive playback.
ADA alignment: Accessibility controls are expected to be persistent and reliable for equal access.
For context, the update that broke the functionality is the one that introduced this “Branch in new chat” option and relegated Read Aloud to the 3 dot context menu. Not only does switching tabs or windows stop it from playing immediately, but even just clicking off of the context menu immediately stops playback. Very annoying and frustrating to use.
I’m also not happy with this change. The Read Aloud function was one of the most genuinely helpful features in ChatGPT and losing its persistence across tabs makes it far less useful. I used it regularly and found it invaluable, so this feels like a step backwards in terms of accessibility and overall user experience.
It would be great if this could be fixed at the earliest, because what was once an excellent and reliable feature has now become frustrating to use.
I completely agree with everyone in this post — it needs to be reverted back to how it was. One of the most basic features I use in ChatGPT is dictation for my longer emails, along with having it read back while I’m doing other things. Losing this really cuts into my productivity and flow. I don’t even mind having it behind a “More Actions” tab, although the pop-up still ends up covering the text I’m trying to read while it’s dictating on shorter notes.
I absolutely agree this Bug needs to be fixed ASAP.
This is really messing with my Workflow Multitasking, and the Advanced Voice Mode doesn’t have half of the Accessibility Functionality that I typically use Read Aloud for.
I really hope this was an accident that gets cleared up immediately in the next Patch.
I am having the same issue, and to be frank - the way it is right now (Audio stops as soon as the context menu loses focus - which basically means any movement out of the component e.g. different browser tab as a very common example) render this very important ChatGPT feature useless.
As someone with web development background, it also seems to look like a bug that should be pretty easy to fix - not the type which requires 5 meetings over three weeks and a CTO approval to push.
this is actually logically true- as the “share” button for sure is less used in compare to the playback button. the least used buttons need to be shifted into a hidden menu, while the top 2-3 buttons visible for efficient working
I 100% agree. This function is integral for those who multi-task between projects, as well as for those of us who assist with clients who are visually impaired. It is imperative that we are able to toggle between tabs especially when explaning other content. We need this fixed ASAP or we will not be able to continue using this platform.
Is this a Bug or intentionally? Could OpenAI response to this? I hope this can be corrected and restore it to the original ASAP, thank you for looking into this matter.
This is a huge step back in terms of usability for me. It basically makes multi tasking impossible. Please revert the change or allow us to pick which elements we see outside of the more actions sub menu. I don’t think I have ever clicked thumbs up or down, I click read aloud constantly. Having it burred in a sub menu kills my regular workflow.
Please guys, don’t sugarcoat this issue. It is simply STUPID, STUPID nonsense mistake. they should bring it back otherwise we should all stop using this platform.
Removing this feature has killed my work flow tremendously. Being able to hear playback while multi-tasking in other areas was half the reason I have pro. They really need to fix this, or I’m done with GPT
I can also confirm this is a widespread issue. Whenever the modal component closes, the Read Aloud functionality is disabled. A silly oversight perhaps, but hugely disruptive to those that use it. The Android App is still fine, as it doesn’t use any modal menus. There’s plenty of space on the UI to have all the icons displayed without arbitrarily hiding some behind an extra interaction. I’m not sure what the though process here was, but I hope this is recognised and addressed soon.
Just wanted to come back and thank the crew at GPT for the fix. I can click into another tab without disruption. While the pop up is still annoying I’m a lot happier that read aloud is persistent. Cheers
Accessibility + Workflow Feedback