Personalize Responses with Memory Sources

## What changed On May 5, 2026 OpenAI began rolling out memory improvements to ChatGPT Plus and Pro users. The system now pulls context from past chats, saved memories, uploaded files, and connected Gmail accounts to shape responses. A new Memory Sources view appears below each answer so users can inspect the exact memories, chats, or files that influenced the output.

Users can edit, delete, or mark items as irrelevant directly in that view. The feature is live on web and scheduled for mobile soon. Memory sources stay private to the account and do not appear in shared chats.

## Why it matters The change reduces the need to restate context across conversations, which matters for people who keep ongoing projects in ChatGPT. It strengthens OpenAI's position against tools that still treat every chat as stateless. The bet is that tighter personalization will increase daily usage and make switching costs higher.

For solo builders this creates a practical advantage when tracking decisions, preferences, and project details over weeks. It also raises the stakes on data hygiene because outdated memories now affect answers more directly.

## How to use it Open any chat on web as a Plus or Pro user and look for the Sources icon under the response. Click it to see the list of memories, chats, files, or Gmail references that shaped the answer. Edit or remove entries on the spot.

Go to Settings to turn memory off entirely or disconnect Gmail. Storage management for related files sits at Settings > Storage. The feature requires a paid plan and is not available to Free or Go users.

## Watch for Confirm the bet if response relevance improves measurably without extra prompting. The bet breaks if users see repeated irrelevant memories or if privacy concerns trigger mass opt-outs. Expect OpenAI to add more connected apps next so the same memory layer can reach calendars and task tools.