Independent public-interest evaluations of how systems respect dignity and support agency.
HXR Evaluation
ChatGPT Read Aloud Interruption Recovery and Playback-Position Preservation
How does this compare with your experience?
Good, difficult, mixed, and different experiences all matter. Share a non-sensitive summary through the manual contact route; evidence and permissions are reviewed separately.
Nothing is published automatically or sent to OpenAI. The private contribution steps and limits are explained below.
An anonymous HXR contributor reports a recurring ChatGPT Read Aloud problem: an ordinary phone interruption can stop or strand playback, and the observed experience provides no practical way to continue from the prior listening position. Recovery can require restarting Read Aloud, repeatedly using the 15-second forward control, and additional phone/app navigation when hands-free use has been lost. The contributor estimates roughly two minutes or more of recovery work per interruption and says the pattern can occur several times per day. Independent public reports describe materially similar restart-from-beginning or failed-resume behavior. HXR does not infer a common technical cause, population prevalence, or behavior across every current ChatGPT platform or version.
- Evaluation ID
- HXR-EVAL-0041
- Status
- Open — gathering evidence
- HXR Incident connections
- 1
- Last updated
- August 25, 2026
Current HXR status
Current work and participation
- Evaluation work
- Open — gathering evidence
- Contribution intake
- Pilot
- Reform and verification
- No Current Reform Work
Evidence, independent corroboration, outreach, organization response, implementation, and verification are tracked separately. An open contribution doorway does not imply any later state.
What would help now: When ChatGPT Read Aloud is paused or interrupted, does it resume where you left off, restart, or behave differently? Share what interrupted playback, whether your place was preserved, and what you had to do next. Please omit private prompt and account details.
Who and what this Evaluation covers
Naming an organization, product, service, or workflow identifies the subject of this Evaluation. It does not by itself prove which internal team caused or controls the issue.
- Organization in scope
- OpenAI
- Product, service, or workflow
- ChatGPT — Read Aloud
Evaluation scope
ChatGPT Read Aloud interruption recovery in the mobile-app context: preservation of listening position and playback state after ordinary phone interruptions; ability to continue or resume at the prior point; persistence and clarity of any interruption/error state; manual restart/seek burden; and the practical effect on hands-free or eyes-busy use. This Evaluation is distinct from HXR-EVAL-0014, which evaluates progressive static/audio integrity during continuous playback. It excludes unsupported claims about operating-system audio focus, notifications, timers, networking, buffering, text-to-speech infrastructure, backend state, internal technical cause, population prevalence, or all ChatGPT platforms.
Question being evaluated
How well does ChatGPT Read Aloud preserve a listener's playback position and provide low-burden continuation after ordinary mobile interruptions, rather than forcing the listener to restart and manually seek back to the prior position?
Current findings
What works well or deserves recognition
Read Aloud provides an auditory path through an existing ChatGPT response and can support hands-free or eyes-busy listening while playback continues normally. The observed playback interface includes manual forward navigation that can approximate recovery after a restart.
Difficulties and opportunities to improve
The contributor reports no practical one-step continuation from the prior listening point after interruption. Recovery can require retrieving or unlocking the phone, returning to ChatGPT, clearing the failed playback state, restarting Read Aloud, and repeatedly pressing the 15-second forward control. This converts a hands-free listening task into a hands-on recovery task and transfers avoidable temporal, attentional, and physical interaction burden to the listener. Independent public reports describe materially similar restart-from-beginning or lost-resume-position behavior.
Mixed findings and limits
The latest described interruption was a phone timer/alert, while the contributor reports that multiple kinds of events can interrupt playback. The exact interruption classes have not yet been independently reproduced by HXR. Exact app version, operating-system version, interruption/error wording, and whether the listening position remains internally available but inaccessible to the user are unresolved. Public reports support the evaluation question without proving one technical mechanism or current universal behavior.
Organization information and response status
No authority-verified OpenAI response to this specific HXR interruption-recovery Evaluation is recorded. User-authored posts on OpenAI-hosted community forums are treated as public user reports, not as official OpenAI findings, acknowledgments, or roadmap commitments.
How HXR would test performance
Begin Read Aloud on a sufficiently long response and interrupt playback using ordinary supported mobile events such as a timer/alert or temporary loss of foreground/audio attention. After the interruption clears, the listener should be able to continue automatically or with one clear resume action at the same sentence or near the prior listening position. Recovery should not require replay from the beginning, repeated fixed-interval forwarding, re-finding the conversation, or reconstructing the playback point. If playback cannot resume immediately, the prior position and an understandable recovery state should remain durable until the user can act.
Evidence coverage and limits
One direct HXR contributor report establishes the bounded current experience. Three independent public report sources materially corroborate restart/lost-resume behavior, and one same-day public feature request independently asks for playback-position memory after interruption. These sources do not establish prevalence, a shared cause, or behavior on every current version.
Connected HXR Incident reports
1 public HXR Incident report is connected to this Evaluation. Canonical roles: 1 supports; 0 partially supports; 0 narrows; 0 context only; 0 challenges. Connections are not votes and do not establish prevalence.
Next observation or verification
Accumulate matching, different, corrective, and successful interruption-recovery reports; preserve interruption type, app/OS version when known, approximate playback point, visible post-interruption state, recovery steps, and whether position was preserved. Continue bounded review for authoritative OpenAI documentation or verified reform evidence.
Technical record details
- Evaluation type
- Product or Application
- Benchmark posture
- Developing Evidence
- Blue Score readiness
- Developmental Data Only
- Reform / positive practice
- No Current Reform Work
- Public revision
- 1