OpenAI Introduces 'Record Mode' to ChatGPT, Granting the AI a Persistent Memory

SAN FRANCISCO – OpenAI is rolling out a significant update to ChatGPT that aims to make interactions more personal and efficient: 'Record Mode'. Officially announced on June 18, 2025, the new feature allows ChatGPT to remember and recall information across multiple conversations, effectively giving the AI a persistent memory of user preferences and history. Currently being deployed to users on Pro, Enterprise, and Education plans via the macOS desktop app, Record Mode functions as a memory toggle. When activated, users can explicitly tell ChatGPT to remember specific details, such as their writing style, coding preferences, project details, or personal interests. This information is then stored and referenced in future sessions, eliminating the need for users to repeat context or instructions in every new chat. For example, a developer could instruct ChatGPT to always provide code snippets in Python with detailed comments, and the AI would adhere to this preference in all subsequent interactions. Similarly, a marketer could have the AI remember brand guidelines to ensure all generated copy is consistent in tone and voice. Another key aspect of Record Mode is its ability to transcribe and summarize audio. Users can record meetings, brainstorming sessions, or voice notes directly within the ChatGPT app. The feature provides a live transcript of the conversation and can then generate structured summaries, identify action items, and even draft follow-up emails or reports based on the discussion. The recordings can be up to 120 minutes long. The introduction of a memory function represents a major step towards making ChatGPT a more capable and personalized assistant. Until now, the AI's memory was limited to the context of a single conversation thread. This update moves ChatGPT closer to the concept of a true digital assistant that learns and adapts to its user over time. However, the new feature has also brought privacy considerations to the forefront. OpenAI has stated that it does not use the audio recordings themselves to train its models and that they are automatically deleted after transcription. However, the generated transcripts are saved to the user's chat history. Users on Pro and Plus plans must manually opt out of sharing their chat data for model training in their settings to ensure their transcripts remain private. The company also explicitly reminds users of their responsibility to obtain consent from all parties before recording a conversation, as legal requirements can vary by location. While Record Mode is currently limited to English and the macOS app, OpenAI has indicated that accuracy for other languages is improving and that a wider rollout is planned. The initial feedback has been focused on the potential productivity gains, with users on team-based plans highlighting its utility for streamlining meeting workflows and maintaining project continuity.