OpenAI appears to be experimenting with a new “Memory search” feature for ChatGPT, designed to let users quickly retrieve information stored in their memories by issuing direct queries. This capability could address current limitations of the memories UI, which becomes cumbersome when users accumulate a large number of entries. Instead of navigating manually, users will likely be able to prompt ChatGPT to surface relevant memory content on demand, streamlining workflows and making context access more efficient.
The concept closely resembles an existing functionality in Atlas browser, where users can perform searches across their Browser Memory, and even the icons for both features are nearly identical. Although the memory search feature has only been visible temporarily and is not currently accessible, its appearance suggests ongoing internal testing or a phased rollout.
Browser Memory on Atlas
OpenAI’s move to add more advanced memory handling could be part of a broader push to retain and expand its user base, especially as competition in the AI assistant market intensifies. Google’s release of Gemini 3 reportedly allowed it to capture market share from OpenAI, leading to a “code RED” situation internally at OpenAI according to recent reports.
BREAKING 🚨: OpenAI is about to drop another reasoning model next week to challenge Gemini 3, according to The Information. And no ads for you… for now.
Red GPT? 👀 https://t.co/lp6e2SUNrk pic.twitter.com/mHLVlPlOw5
— TestingCatalog News 🗞 (@testingcatalog) December 2, 2025
As a result, there is increasing speculation that OpenAI may accelerate the launch of GPT-5.2 as a direct response, potentially aiming for a December release window. If both the memory search and a new model update arrive in close succession, this would signal a strategic effort by OpenAI to counter growing competitive pressure and reinforce ChatGPT’s role as a productivity tool for users managing large amounts of information.
