ChatGPT has been quietly building a profile on you — even when you're not talking to it.

The exact clicks

1. Open ChatGPT — the app or chatgpt.com, any plan, free or paid — and start or continue any chat.
2. Type exactly: "What do you remember about me?" and send it.
3. Read the summary it gives back in full before reacting to any one detail in it.
4. Then check the visible list yourself: Settings > Personalization > Memory (some accounts see a further "Manage" step) shows the editable memory summary and, separately, any individually saved memories.
5. Optional, to see more of what it's actually drawing on: ask a direct follow-up, "Is there anything else you know about me that isn't in that summary?" — OpenAI's own help docs say the visible summary can be incomplete, so this sometimes surfaces more.
6. If anything in what it says back is wrong, outdated, or something you don't want remembered, use the same Settings > Personalization > Memory screen to edit or delete it directly — you don't need a special prompt to remove something.

What changes

• Verified live (29.9.2026): "Dreaming" confirmed directly on OpenAI's own blog (openai.com/index/chatgpt-memory-dreaming/, published 4 June 2026), corroborated by multiple independent outlets with consistent details; the memory-summary-is-incomplete claim confirmed against OpenAI's own Help Center content, not inferred; the current Settings > Personalization > Memory path and the "What do you remember about me?" prompt itself both confirmed working today across multiple independent 2026 sources.
• Checked against every prior episode script and build for a repeat angle: no earlier episode on this channel covers ChatGPT memory/profiling specifically — this is a new mechanism and a new claim, not a rehash.
• A real screenshot of the live Settings > Personalization > Memory screen was attempted (headless Chrome, retried this session) and blocked at the infrastructure level (403 from openai.com and help.openai.com through this environment's proxy) — the same class of failure episode 51 hit for a different reason. The real ChatGPT logo mark used in the video was sourced directly from Wikimedia Commons as an actual downloadable image file, not a screenshot; the interface itself is shown as a labeled recreation of the verified real path, not presented as a literal capture.

What it will not do

• This is about ChatGPT's Memory system specifically (Dreaming plus the visible summary) — not a claim about every AI product's memory behavior.
• The exact wording, detail, and accuracy of what comes back varies by account, plan, region, and how much history exists — OpenAI's own rollout was staged by plan and country starting June 2026, so exactly what you see may differ from what's shown here.
• Asking the prompt shows you a summary of what the model has inferred and retained — it is not guaranteed to be a complete, itemized list of every underlying data point OpenAI's systems hold about you.