ChatGPT, Gemini and Claude all remember you now — and deleting that memory is no longer as simple as trashing a chat. As of November 2026, each service has restructured how it stores, uses and lets users remove personal context, with critical differences in what gets wiped, where it lives, and whether deletion is even possible without resetting everything.
This isn’t just about privacy hygiene. It’s about control: who decides what stays, what vanishes, and what lingers in the background after you’ve clicked ‘delete’? And for users searching ‘ChatGPT free’ or ‘Gemini AI’, the answer affects how much of their digital life sticks around, even when they think it’s gone.
OpenAI still ties saved memories to your account separately from chat history. Delete a conversation? Your saved memory remains untouched. To scrub it fully, you must manually delete the memory entry itself, plus the original chat, any archived versions, files in your library, and connected apps like Gmail (if enabled).
Worse, OpenAI says it may retain logs of deleted saved memories for up to 30 days. The ‘Don’t mention this again’ toggle? Not deletion. Just suppression.
And temporary chats, once saved, become regular chats and feed into memory. There’s no undo. Anthropic overhauled Claude’s memory entirely. Since summer, it no longer summarizes chats after the fact.
Instead, it saves discrete topics as you type: your job title, ongoing projects, people in your life, writing style preferences. Memory is on by default across Free, Pro, Max and Ultra plans. But here’s the shift: deleting a chat no longer triggers automatic removal of those topic entries. You must go into ‘Manage Memory’, open each topic individually, and delete them one by one, or choose ‘Reset memory’, which wipes everything, including project memories, and cannot be reversed.
Incognito chats (ghost icon) avoid history and memory entirely. Turning off ‘Memory’ before sending your first message keeps the chat in history but stops saving, but only if you do it before the first message. Google’s Gemini takes the most fragmented approach. In the US, memory is tied to saved chats, and since November 9, a memory summary has begun rolling out, letting users view and delete remembered details by category.
Outside the US? No summary. No per-item deletion. Just delete every chat containing the detail, and hope Gemini stops using it after a ‘short delay’.
Even then, connected apps like Gmail or Google Photos aren’t enough to trigger memory. You must both delete the chats and disconnect the app. Disconnect alone? Gemini may still pull from past chats.
Delete chats alone? It may re-fetch from the app. Chrome autofill shows how deep this integration already runs. None of these systems tie memory deletion to model training.
Opting out of training is a separate setting, and turning off memory does nothing to stop your chats from being used to improve the underlying models. So what’s next? A growing number of users are asking not just ‘what does it remember?’ but ‘where did it learn it, and can I really unteach it?’ For now, the answer depends on which AI you’re using, where you’re located, and how carefully you manage settings before you start typing.
CDFiled by The Computing Desk
Laptops, desktops, processors, graphics cards and the operating systems that run on them.
More from this desk →
Be the first to comment
Join the argument. No password, just your email or a passkey.