Perplexity Spaces for Recurring Research

TL;DR: Perplexity Spaces doesn't exist under that name anymore. The help-center URL that used to explain it now serves a page titled "What are Projects?", and Projects are the persistent workspace Perplexity documents for recurring research today. Below: what belongs in one, how it differs from Deep Research and Model Council, and how to set one up.

What Happened to Perplexity Spaces?

If you searched "perplexity spaces" hoping for a features page, here is what actually exists on September 3, 2026: nothing under that name. Perplexity's help center still resolves the original article URL, /help-center/articles/10352961-what-are-spaces, but the page it serves now is titled "What are Projects?" We fetched it directly to confirm: the <title> tag reads "What are Projects? | Perplexity Help Center", not "What are Spaces?" (Perplexity Help Center, accessed September 3, 2026). Spaces was renamed, not retired, and the URL slug is the only thing still carrying the old name.

This matters for anyone searching Perplexity Spaces for recurring research specifically, because every setting that used to live under that name now lives under a different one, and a guide written against the old feature will describe menus that no longer exist. We checked Perplexity's full help-center sitemap while researching this piece: of 207 English articles, zero currently document a feature called Spaces, and the sitemap lists 2,424 URLs across every locale with the same pattern. Zero also document something called Perplexity Labs, a separate premise we checked and ruled out before writing a word about it.

What didn't survive the rename is the vocabulary, not the underlying idea. If you used Spaces to keep a running research thread alive across sessions, add files to it, and share the workspace with collaborators, a Project does the same job, with more attached to it. The rest of this post treats "Project" as the working name.

What Is a Perplexity Project?

Perplexity's own definition is the cleanest single sentence available: "A Project is a persistent, shareable workspace in Perplexity that keeps everything for an ongoing effort in one hub — your Search conversations, your Computer tasks, files, custom instructions, connected tools, and the context Perplexity builds up as you work." That's a wider net than "a folder for chats." A Project holds live Search threads, Computer automation tasks, uploaded and generated files, the instructions you write for it, any connected tools such as Gmail, Slack or Notion, and an accumulating layer of memory the article calls Brain, covered below.

Creating one takes three fields: a title (up to 50 characters), an optional description (up to 1,000 characters), and an optional icon or color. Projects are private (Restricted) by default, and you choose who else can see or edit it afterward. Sharing runs on three roles: Owner, with full control; Can edit, who can touch context, files, connectors, skills, people and defaults; and Can view, who can read the Project and its shared sessions or files. Contributor counts are capped: "up to 5 contributors on non-Enterprise Projects; up to 9,999 on Enterprise-owned Projects." Enterprise admins can additionally restrict public sharing, external invites, and org-wide publishing.

You aren't locked into building a Project from a blank page. Perplexity documents starting one from an existing session, where it suggests a title, description and instructions based on what you already did, and inside a Computer task you can use "Convert session to project" to carry files and artifacts straight in rather than re-uploading them.

What Belongs Inside a Research Project's Instructions?

The instructions field is the closest thing a Project has to a system prompt, and it's generous by chat-app standards: up to 8,000 characters "that tell Computer how to work inside this Project — for example, a preferred response format or a status file to keep updated." For a recurring research Project, that's room for an actual brief, not a topic restatement: what counts as a credible source for this specific effort, the citation format you want back, which sections a report should always contain, and what Computer should do when it hits contradictory information rather than silently picking one side.

Files sit next to instructions rather than instead of them. Each Project gets a persistent files store you can add to individually, as folders, or as imports from connected sources, and Computer can create, update and manage files inside it on your behalf, not just read them. Viewers can see project files; only editors can change them.

The Settings tab holds the rest of the configuration, split into Context, Memory, and Connectors. Context is where you manage the standing instructions and prioritized web links and domains, and where you set the Project's default mode between Search and Computer plus a default Computer orchestrator model. Memory is where Brain and project-summary generation get turned on, off, or set to automatic versus manual, and where you decide whether queries in the Project can also draw on your individual, non-project memory. Connectors decides which of your linked apps this specific Project can use, with connector-specific instructions and shared project-level credentials layered on top.

How Does a Project's Memory (Brain) Work?

A Project's memory system has a name that only shows up inside Perplexity's own settings: Brain. Perplexity states it plainly: "Brain helps build ongoing memory for all work done in a project." Separately, it states that Brain "builds a live and updating knowledge based on the project’s activity to give Computer better context". Practically, that means a Project you return to weekly doesn't start cold each time. Findings, decisions and open questions from session three are available to session thirty without you re-pasting them into the instructions field.

Two settings matter here, both under Memory. First, whether Brain runs automatically after each session or only when you trigger it manually, worth deciding deliberately because Brain runs are billed to whoever created the Project. Second, a separate toggle for project-summary generation, which produces a running status update of the work and can lean on Brain for more context when switched on. Neither is the same as the files store: files are documents you or Computer put there on purpose, Brain is Perplexity inferring what mattered from how the Project actually got used.

One caution worth stating plainly: none of this is published as a numeric limit. Perplexity's pages describe what Brain does and how you control it, but at no point publish a token count, a memory-entry cap, or a retention window for what it retains. Treat that as undocumented, not as unlimited.

Are Skills Different Inside a Project?

Skills are a Computer feature, not a Projects feature on their own, but the two connect directly: a Project's Settings tab has its own Skills sub-section that lets you "Upload and manage reusable skills scoped to the project." It also notes that "users will still have access to personal and organization skills within a project as well". So skills exist at three scopes, personal, organization, and project, and a Project just adds one more shelf to the same system rather than a parallel one.

What a skill actually is matters for a research workflow specifically. Perplexity's help center opens its Skills article with an unusually direct claim: "Skills are more than reusable instruction sets — they deploy agents." That's Perplexity's own framing, not ours: a skill works less like a saved prompt and more like a dispatcher for sub-agents. A skill "gives Perplexity Computer a methodology for a type of task, and Computer then dispatches dedicated sub-agents to carry the work out, often several in parallel, each handling one part of the task and reporting back." The built-in Research skill is the relevant one here: it "deploys research sub-agents that work through a rigorous multi-round methodology in parallel, with source validation and citations." Computer often chains it with other skills, handing results to a Research Report skill for formatting or a Slides skill for a deck.

You can also write your own. A custom skill is a .md file with YAML frontmatter carrying a name and a description, the same shape Perplexity documents for its "Create with Perplexity" and "Upload a skill" flows. The description field does the routing: Perplexity activates a skill when your request matches the trigger phrasing you wrote into it, so a vague description is a skill that never fires.

Deep Research, Model Council or Computer: Which Do You Actually Need?

A Project is where recurring research lives. Deep Research, Model Council and Computer are what you run inside it, and they solve different problems well enough that reaching for the wrong one wastes the wait.

The plan-gating split on Model Council is the one worth reading twice: it's Max-and-above on web and in the iOS app, but "In Computer, Model Council runs as a skill and is available in all paid plans." A Pro subscriber can't toggle Model Council in the search bar, but can ask Computer for a Model Council run inside a Project and get one, spending Computer credits instead of a separate query allowance. On web, Model Council itself carries a plan price: it's included with Perplexity Max at $200 per month or $2,000 per year, and with Enterprise Max, with no extra per-query charge within that plan's monthly cap.

How Do You Set Up a Project for Recurring Research?

Setting one up well takes about ten minutes, and the sequence below front-loads the parts that are hard to bolt on later.

1. Start from an existing thread if you have one. "Convert session to project" inside a Computer task, or starting a Project from an existing session, carries your work forward instead of writing the instructions field from memory.
2. Write instructions as a brief, not a vibe. You have up to 8,000 characters; use them for source rules, output format, and what to do on contradictory findings, not a restatement of the topic.
3. Set sharing before you invite anyone. Projects default to Restricted. Decide whether collaborators need Can edit or Can view before you send the link, since only editors can touch files.
4. Turn on Brain deliberately. If you're creating the Project, know that Brain runs are billed to you, and decide manual versus automatic before your first real session.
5. Set the default mode. A research Project set to Computer as default gets skills, scheduling and connectors without switching modes every time; set to Search, it stays a fast Q&A thread.
6. Pin it. Pinned Projects surface in both the overview page and the sidebar, which matters once you're running more than two or three of these at once.

A starting instructions block you can paste into a new Project and edit down:

ROLE: Research assistant for [topic / project name].

SOURCES: Prefer primary sources and named studies over aggregator blogs.
Flag anything from a marketing page or SEO listicle as such.

FORMAT: Every finding gets one sentence of claim, one citation, one date.

ON CONFLICT: If two sources disagree, present both positions and their
sources. Do not silently pick one.

STATUS: Keep a running list of open questions at the top of the latest
report.

Adjust the SOURCES and ON CONFLICT lines first. They're doing the actual work; everything else is formatting.

If you're deciding between Perplexity and another AI search tool for this specific job, our guide to prompting Perplexity with search-shaped prompts covers the query side, and our Perplexity prompt templates for research post has copy-paste starting points for the Search side of a Project. For the mode selector itself, Focus, Choose sources, Research, Model Council, our Perplexity Focus modes guide covers what each one actually does now that Focus modes themselves are gone from the documentation.

The Bottom Line

Perplexity Spaces is a name you can stop searching for. What replaced it, Projects, is more capable than the old feature was ever documented to be: persistent files, scoped skills, its own memory layer, and a direct line into Computer's scheduling and automation. If your research is a single question, Deep Research or plain Search answers it faster than setting up a Project ever will. If it's a standing effort you return to weekly, for a client, a market, a competitor set, a Project is where Perplexity now says that work should live, not a workspace you rebuild from scratch every time the thread gets long.

A Project's instructions field solves this for Perplexity specifically. Prompt Architects solves the adjacent version of the same problem: stop re-explaining yourself every session, everywhere. If Perplexity is one leg of your research stack and ChatGPT, Claude or Gemini are the others, keeping the same source rules and output format synced across all of them, not just the one product with an 8,000-character field, is what a shared prompt library is for.