Read the official docs
We start from OpenAI's Memory FAQ and Data Controls FAQ, noting exactly what is stated, what is implied, and what is left unspecified.
Trust & methodology
How we research ChatGPT memory.
ChatGPT memory touches your personal data, so accuracy matters. This page explains exactly how we research it: we build every page on OpenAI's official documentation, test the actual in-product toggles, hedge wherever OpenAI doesn't publish details, update as features change, and stay independent. We never publish unverified internal numbers.
Our principles
How we decide what to publish
Five rules govern every page on this site. They keep our guidance grounded in what's verifiable rather than what's plausible.
| Principle | What it means in practice |
|---|---|
| Official sources first | We base everything on OpenAI's Memory FAQ, the Data Controls FAQ, and the in-product settings — not secondary blog posts. We link to the source so you can verify. |
| We test the toggles | We open ChatGPT and exercise the actual controls — saved memories, chat history, Manage memories, Temporary Chat, the training opt-out — to confirm behavior. |
| Hedge where details are unpublished | If OpenAI doesn't publish a number or mechanism (e.g. the exact memory size limit), we say so and point to the help center instead of inventing a figure. |
| Update as features change | Memory features and data policies evolve. We revisit the docs and settings and revise pages when behavior shifts. |
| Independent and transparent | We're not affiliated with OpenAI and don't publish unverified internal benchmarks or leaked specs. |
Our two primary references are OpenAI's Memory FAQ and the Data Controls FAQ.
The process
How a page gets written and checked
Every guide goes through the same four steps before it's published, and the same steps again whenever OpenAI changes something.
Test in the product
We open Settings → Personalization → Memory and Data Controls and exercise each toggle to confirm the documented behavior matches the live UI.
Write with hedges
We describe what's verified plainly, flag what OpenAI doesn't publish as unknown, and avoid specific figures (like exact storage size) that we can't source.
Revisit and revise
When features change, we re-check the docs and settings and update the affected pages, keeping the official links current.
See it applied: how memory works · saved vs chat history · memory & training data · the full FAQ
What we won't do
The lines we don't cross
| We avoid | Why |
|---|---|
| Inventing numbers | We don't cite a specific saved-memory size or retention figure OpenAI hasn't published. See memory limits for how we handle the "memory full" state without a fake number. |
| Unverified internal benchmarks | Leaked specs and speculative performance claims don't belong in a guide people use to make privacy decisions. |
| Speaking for OpenAI | We're independent. For authoritative policy, we point to OpenAI's Data Controls FAQ. |
| Stale guidance | Behavior changes, so we date our claims to the official docs and re-verify rather than assume yesterday's answer still holds. |
Beyond one provider
Why independence matters for AI tools
Because we're not tied to any single AI company, we can say plainly when concentrating all your context in one assistant's memory is a trade-off. If you'd rather spread your usage and choose which model sees which prompt, a multi-model workspace gives you that option while you keep sensitive topics in memory-free sessions.
All-in-one
MultipleChat AI
multiplechat.aiUse ChatGPT, Claude, Gemini and other models in one place — choose where each prompt goes instead of feeding everything to a single assistant.
Privacy-first: MultipleChat doesn't save your chats to memory and doesn't share your data with model providers or let them train on it. Other privacy-focused options include Duck.ai, Proton Lumo, Brave Leo and Kagi Assistant; for full local control, self-hosted tools like Ollama or Jan keep everything on your device.
OpenAIChatGPT
chatgpt.comOffers Temporary Chat (no history, not used for training), a training opt-out under Settings → Data Controls, and editable memory. By default, consumer chats may be used to improve models unless you opt out — verify on OpenAI's help center.
AnthropicClaude
claude.aiAnthropic gives you data and training controls and has historically taken a privacy-conscious stance on consumer chats, but its training and retention settings have changed over time — check your current privacy settings in the app.
GoogleGemini
gemini.google.comGemini Apps Activity lets you pause history and delete past activity, but sampled conversations can be reviewed by humans, so avoid sharing confidential information. Review your Activity controls.
MicrosoftCopilot
copilot.microsoft.comMicrosoft 365 Copilot (business) doesn't use your organization's data to train models and adds Enterprise Data Protection; consumer Copilot has a privacy dashboard to review and delete your history. Check current Microsoft policies.
PerplexityPerplexity
perplexity.aiPerplexity has an "AI data retention" setting you can switch off so your searches aren't used to improve its models, plus account-level data controls. Confirm the current options in settings.
All guides
ChatGPT memory: every guide
How memory worksThe technical walkthrough.
Saved vs chat historyThe two layers compared.
Turn memory on/offManage the settings.
Delete memoryClear individual or all.
Memory & privacyRisks and how to reduce them.
Memory & trainingIs it used to train models?
Temporary ChatMemory-free conversations.
Memory not working?Troubleshooting steps.
Memory limitsHow much it can hold.
Team & EnterpriseMemory and data controls.
Memory vs custom instructionsWhat's the difference?
Memory FAQEvery common question.
FAQ
Methodology — your questions
How we source, test, hedge and stay independent.
What sources do you rely on?
We base everything on OpenAI's official documentation — primarily the Memory FAQ and the Data Controls FAQ — together with the in-product Memory and Data Controls settings as they actually appear in ChatGPT. Where OpenAI's help center is the authority, we link to it directly rather than paraphrasing from secondary write-ups.
Do you test the features yourselves?
Yes. We open the actual settings in ChatGPT and exercise the toggles — Reference saved memories, Reference chat history, Manage memories, Temporary Chat and the Data Controls training option — to confirm the described behavior matches what OpenAI documents. Hands-on testing is how we catch wording or UI changes.
What do you do when OpenAI doesn't publish a detail?
We hedge. If OpenAI doesn't publish a specific number, mechanism or timeline — for example the exact size of the saved-memory limit — we say so plainly and point readers to OpenAI's help center rather than inventing a figure. We treat unpublished internals as unknown, not as something to estimate.
How do you keep the guidance up to date?
Memory features, data policies and controls change over time, so we revisit the official FAQs and the in-product settings and update our pages when behavior shifts. Because of this, we encourage readers to confirm any detail that matters to them on OpenAI's official help center before relying on it.
Are you affiliated with OpenAI?
No. ChatGPTMemory.com is an independent guide. We are not affiliated with, endorsed by, or operated by OpenAI. Our goal is to explain how ChatGPT memory works clearly and accurately for users, not to speak on OpenAI's behalf.
Do you publish internal benchmarks or leaked figures?
No. We do not publish unverified internal benchmarks, leaked specifications, or speculative numbers. If a claim cannot be grounded in OpenAI's official documentation or directly observed in the product, we either omit it or clearly mark it as unknown.