Perplexity memory: the complete guide

The three layers

Threads. A Thread is a single research session — multiple turns, a shared context across those turns, and citations that persist. Threads don't cross-pollinate. Opening a new Thread starts fresh, even on the same topic.

Spaces. A Space is a container for many Threads plus system-level context: a description, custom instructions, uploaded files, and shared links. Every Thread inside a Space inherits that context. Spaces are the closest thing Perplexity has to "topic memory."

Memory (Pro). A lightweight cross-session layer that carries preferences and a small handful of facts about you across all your Perplexity use, similar in spirit to ChatGPT's saved memories but much less aggressive.

The configuration that works

1. Turn Memory on in Settings → Personalization.
2. Create a Space for any topic you research more than once. A Space beats "I'll just start a new Thread and re-explain the context" every single time.
3. Put the durable facts about the topic (the frame, the constraints, the goal) in the Space's custom instructions field, not in individual Threads.
4. Upload any reference PDFs to the Space so they're implicitly available to every Thread inside it.
5. Use Threads for actual sessions. When a Thread's context gets confused, start a fresh one — the Space's context carries.

Where Perplexity beats ChatGPT — and where it doesn't

Beats. Spaces are the honest version of "topic memory": you can see exactly what context is being applied and edit it. Citations survive across Threads. Custom instructions scope cleanly to a Space rather than being global. For research workflows, this is much better ergonomics than ChatGPT.

Doesn't. Perplexity Memory is shallow. There's no long-running "who you are as a person" layer that grows with you across a year of use. If you want an AI that gradually learns your voice, preferences, and history, Perplexity isn't optimizing for that — it's optimizing for research quality on the current query.

Making it work across your other AI tools

Because Spaces are so explicit, Perplexity is one of the easiest tools to sync into a portable memory layer — the context you'd paste into a ChatGPT chat or a Claude Project is basically a Space's custom instructions with light massaging. If you keep a canonical version of that context outside any one tool, updating it in one place propagates everywhere. See our Perplexity memory field note for the specific patterns we've seen work.

Related reading

• Perplexity memory — field note
• Best AI for long-term memory
• Memory vs context window