You delete an entry from your memory summary, but in a later chat ChatGPT still seems to know that detail about you. What explains it?

The principle — a memory entry is derived from something, and deleting the derivative does not delete the original.

Why the key is correct — OpenAI is explicit that to fully delete something ChatGPT may know about you, you need to delete every source where it appears, including past chats, archived chats, files and the memory summary, and disconnect any connected apps that may contain that information. Clearing one line from the summary leaves those sources untouched, and memory can form again from what remains.

Why the others are wrong — the removal is not queued, since the entry leaves the summary immediately. There is no hidden duplicate list; the summary is a high-level view of a broader synthesis rather than a decoy. And memory does not have to be switched off for a deletion to work, though switching it off does stop new entries being created from chats that remain.

Remember this — privacy work on this feature is source work. If it matters, delete the conversations and files, not just the summary line, and remember that turning memory back on can rebuild entries from chats still in your history.

Sources — OpenAI's Memory FAQ.