Introduction

On June 15, 2025, OpenAI rolled out a significant update to GPT-5, introducing long-term memory and user personalization features across ChatGPT. The update enables the model to recall past interactions, understand individual preferences, and adapt its responses accordingly, ushering in a more agentic AI experience.

“Personalized AI is the key to relevance and utility,” said Mira Murati, CTO of OpenAI.¹

The memory system retains facts about the user such as name, writing style, project goals, and task history—unless explicitly disabled. Users can review and delete memory items through a new ChatGPT Memory Manager dashboard, giving them transparency and control.

Why it matters now

• Memory and personalization close the gap between general models and task-specific agents.
• It enables contextual continuity across sessions—critical for enterprise use.
• Enhances assistant effectiveness in business, education, and productivity tools.

Call-out: GPT-5 now remembers who you are—and what you want

OpenAI reports a 40% increase in task accuracy when memory is enabled in real-world productivity tasks.

Business implications

• Executives and managers gain AI assistants that remember key decisions, ongoing tasks, and communication tone.
• Developers and analysts can build workflows with session-spanning context.
• Customer support benefits from agents that track user history for continuity and satisfaction.

The new features are currently available in ChatGPT Pro and Enterprise plans, with a broader rollout planned later this summer. Privacy protections include opt-out defaults and encrypted local memory options.

Looking ahead

OpenAI is also testing deeper memory APIs for fine-tuned models and multi-agent systems. It plans to release team-based memory contexts and shared agent collaboration zones for business users by Q4.

IDC forecasts that by 2028, 60% of business software will incorporate AI agents with persistent user context and adaptive reasoning.

The upshot: This latest GPT-5 upgrade signals a shift from stateless AI to adaptive agents—an essential step toward true digital collaborators.

––––––––––––––––––––––––––––
¹ Mira Murati, OpenAI ChatGPT Update Briefing, June 15, 2025.