ChatGPT Temporary Chat Can Now Save and Personalize

Most software autosaves everything. ChatGPT's Temporary Chat did the opposite — disposable by default, no history, no new memories — which was great until a throwaway thread became a real project and you were stuck copy-pasting or screenshooting.

On August 27, 2026, OpenAI's ChatGPT release notes shipped the missing middle ground: context without permanence by default, plus an intentful save when the conversation earns a place in your sidebar.

TL;DR

What changed for Temporary chats?

OpenAI framed the update as a direct response to a common workflow failure: people start Temporary for privacy, then discover the thread is worth keeping.

According to the August 27 release notes ("More controls in temporary chat"):

1. Personalize (optional) — when you create a Temporary chat, you can choose personalized responses that use your existing memories, custom instructions, and plugins from regular chat settings.
2. Save (optional) — when a conversation becomes useful, you can add it to chat history.
3. Still Temporary until you save — Temporary chats do not create new memories and stay out of your sidebar/history unless you choose to save them.

That is a real product shift. Before this, Temporary Chat was closer to a hard privacy mode. After it, Temporary Chat is a lifecycle: start disposable → optionally bring context in → optionally promote to history.

The older Temporary Chat FAQ still describes the classic "no memory access / no history" model and says Temporary Chat "will still follow your custom instructions if they're enabled." Treat the August 27 release notes as the source of truth for the new dual-mode behavior: default Temporary is non-personalized, and custom instructions are part of the opt-in personalization path — not an always-on Temporary default.

Does personalizing a Temporary chat write new memories?

No. OpenAI's wording is explicit: personalized Temporary chats can use your memories, custom instructions, and plugins, and they do not create new memories.

That distinction matters for daily work:

If you later save the Temporary chat, OpenAI says it becomes a regular chat and then follows your account-level personalization settings. With memory enabled, that saved history can be referenced for future personalization — so Save is the privacy boundary, not just a UX convenience.

For health and other sensitive contexts, that same "no new memories unless you want permanence" instinct already shows up in explainx.ai's ChatGPT Health privacy guide. Temporary Chat is still the right starting mode when you want a one-off discussion that should not grow your memory summary.

What's the difference between personalized Temporary and regular chat?

Think in three modes, not two:

Personalized Temporary is not "regular chat with a Temporary label." It is read-only personalization for one session: your past context can shape answers, but the session itself does not expand memory while it stays Temporary.

That sits next to the broader industry push toward finer memory scopes — Anthropic's same-week Claude unified Chat/Cowork memory is a different architecture (shared topic list across surfaces), but the product pressure is the same: users want useful context without an all-or-nothing retention story.

If you are comparing how ChatGPT's ambient memory feels versus Claude's more project-shaped context, our Claude vs ChatGPT for work breakdown still holds — Temporary Chat is ChatGPT's escape hatch when ambient memory would pollute an answer.

When should I use Temporary vs regular?

Practical rule of thumb for daily ChatGPT users:

Start Temporary (non-personalized) when:

• You want an unbiased first draft (job offers, negotiation language, creative options) without brand voice or remembered preferences steering the answer.
• You are pasting something you would rather not see in the sidebar tomorrow.
• You are testing a prompt pattern and do not want it to become "how ChatGPT thinks you work."

Start Temporary + personalize when:

• You need your usual tone, role, or plugins, but you are not sure the thread deserves history yet.
• You are doing 20 minutes of exploratory work that might become a project — the QoL win is exactly this: start disposable, save later, instead of screenshot archaeology.

Start regular when:

• You already know this is recurring work (a client project, a research series, a standing weekly brief).
• You want the session to be eligible for future memory and history search.

Plugins are part of the personalization opt-in. If your workflow depends on ChatGPT's Plugin Directory / Skills surface, see how to use Skills in ChatGPT — personalized Temporary is the mode that can bring those tools into a still-unsaved thread.

How do I save a Temporary chat that became useful?

OpenAI's release notes describe the intent, not a pixel-perfect UI walkthrough: when a conversation becomes useful for the future, add it to your chat history. Saving converts it into a regular chat.

What changes after Save (per OpenAI):

• The chat follows your account-level personalization settings.
• It follows your model-improvement preferences.
• If memory is enabled, the saved chat can be referenced to personalize future responses.

What does not change until Save:

• Temporary chats stay out of the sidebar/history.
• Temporary chats do not create new memories.
• Per the Data Controls FAQ, Temporary chats aren't used to train models, may be reviewed only for abuse monitoring, and are deleted from OpenAI's systems after 30 days while they remain Temporary.

Community reaction to the announcement clustered around one theme: saving requires intent — the reverse of most consumer software. That is the feature, not a bug. The other recurring request — a per-chat "disable memory" toggle on regular chats for unbiased opinion work — is still a user wish, not something this Temporary Chat update ships.

Privacy implications of opt-in save + no new memories

This update is a privacy win if you treat Save as a deliberate act.

What Temporary still protects (while unsaved):

• No sidebar history entry
• No new memories written
• Not used to improve models (per Help Center)
• 30-day retention window for safety/abuse purposes — Temporary is not "zero server copy the instant you close the tab"

What personalization reintroduces (if you opt in):

• Existing memories can shape answers
• Custom instructions apply
• Plugins can run — and any plugin/action that sends data to a third party still follows that recipient's privacy policy (same Temporary Chat FAQ caveat as before)

What Save reintroduces:

• Full history membership
• Account-level memory and training preferences
• Future personalization that can reference the saved thread

For workplace use, Temporary Chat remains one of the few consumer controls that reduces accidental retention — relevant to the shadow AI pattern where people paste work into personal accounts and hope the sidebar never reminds them. Temporary does not fix policy or DLP. It does give individuals a cleaner default for exploratory prompts.

If you care about how ChatGPT-style memory compares to more structured memory systems, Perplexity Brain's memory approach is a useful contrast: ChatGPT's Temporary Chat is a session-scoped privacy mode, not a separate memory graph.

Honest limitations

• Personalization is start-only. OpenAI says you cannot change the personalization choice after the Temporary conversation begins. Wrong mode means start over.
• Default Temporary is colder than before. Non-personalized Temporary does not use custom instructions. If you relied on the older FAQ's "still follows custom instructions" behavior, verify what your account shows after the rollout.
• Save is irreversible in spirit. Converting to a regular chat brings account-level memory and model-improvement rules back into play.
• No shipped per-chat memory off for regular chats. Users still want that for "one unbiased regular thread"; Temporary Chat remains the workaround.
• Help Center pages can lag. Memory FAQ and Temporary Chat FAQ language about Temporary never using memories may not fully reflect the new personalized Temporary option — prefer the dated August 27 release notes for this feature.
• Plugin/third-party actions still leave Temporary's bubble. Data sent through GPT actions or plugins follows the recipient's policy.
• Not a compliance product by itself. Enterprise retention and Compliance API behavior for Temporary chats is separate; consumer Temporary is a UX privacy mode.

Practical workflow: start Temporary, personalize, save

A pattern that matches how people already work:

1. Open Temporary (Help Center: pill-shaped Temporary control on a new chat).
2. Decide personalization at the start — leave it off for clean-room thinking; turn it on if you need voice, remembered facts, or plugins.
3. Work for as long as the thread is exploratory.
4. If it becomes a project, Save — accept that it now behaves like a regular chat under your account settings.
5. If it stays disposable, close it — no sidebar clutter, no new memories, Temporary retention rules apply.

That is the workflow Temporary Chat was always supposed to support. August 27 made the middle states explicit.

Related reading

• ChatGPT Health: Apple Health, records, and privacy
• How to use Skills in ChatGPT
• Claude unified memory: Chat + Cowork (Aug 26)
• Claude vs ChatGPT for work
• Shadow AI workplace privacy risk
• Perplexity Brain memory system
• What is memory.md for AI agents?
• Sam Altman on ChatGPT watching your screen

Official sources: ChatGPT release notes — August 27, 2026 · Temporary Chat FAQ · Memory FAQ · Data Controls FAQ

Feature behavior, Temporary Chat retention, and personalization rules verified against OpenAI Help Center release notes and FAQs as of August 27, 2026 — confirm the controls in your ChatGPT account, because rollout and Help Center copy can lag each other.