- OpenAI has added a new feature to ChatGPT that allows users to interrupt the AI mid-response
- Users can write context or corrections to a message without restarting, using the “Refresh” button
- The feature is especially aimed at those using GPT-5 Pro or Deep Research.
Accidentally sending a message to ChatGPT only to realize that you forgot to include some detail or copied the wrong phrase from a website is a fairly common occurrence. Still, it’s a real pain if you’re using the GPT-5 Pro model or ChatGPT’s deep research feature, because they take much longer to respond and you may have a limited number of messages to send.
That should be less of a problem now that OpenAI allows ChatGPT users to interrupt the chatbot mid-response. During the minutes that the AI is compiling an answer, you can add new information and redirect its reasoning without having to start over. It is a way to direct the conversation in real time.
While the AI ponders how to respond to your query, you can click “Refresh” in the sidebar and add any new context or clarification, then watch as the AI immediately adjusts what it will say.
No need to reboot. You can simply have ChatGPT change your mind while you’re still typing a response.
You can now interrupt long-running queries and add new context without restarting or losing progress. This is especially useful for refining deep investigations or GPT-5 Pro queries, as the model will adjust its response to your new requirements. Just hit refresh in the sidebar and… pic.twitter.com/kESrkU9hc9November 5, 2025
In practical terms, the ability to edit the intermediate response can save a lot of time for researchers and analysts who use ChatGPT for deep multi-step queries. Imagine someone writing a data analysis report and realizing halfway through that a key source is out of date.
Before this update, they would have to stop the model, rewrite their query, and run everything again. Now, they can simply say “replace 2022 with data from 2024” and watch the text update accordingly.
That’s a small miracle for anyone who’s ever listened to a long, overconfident explanation of AI, waiting for the model to finish so they can correct a premise that was incorrect at the beginning. For a predictive text-based system, this marks a shift from simple back-and-forth messages to a more interactive format.
AI adjustments in real time
Reducing time lost due to errors in a message or AI understanding of it is definitely something that ChatGPT users will like. The update also continues ChatGPT’s gradual evolution into a more adaptable service.
Since GPT-4, users have been able to provide follow-up prompts, but the flow has always been done in turns. The continuous, overlapping feedback loop is more useful because you are no longer waiting for your turn to speak. AI listens while you type.
In a sense, OpenAI is giving ChatGPT a sense of conversational humility. Previous models were famous for moving on, even when they had misread a question or misunderstood the context. The update doesn’t make the AI self-aware, of course, but it does allow it to correct itself in a way that’s closer to actually collaborating with a person.
And of course you can also follow TechRadar on TikTok for news, reviews, unboxings in video form and receive regular updates from us on WhatsApp also.
The best business laptops for every budget
