How to Actually Get Real Deliverables Out of Perplexity Labs (Not Just Prettier Reports)

1. Know which mode you’re actually in, Search, Research, or Labs

This is the mistake I see more than any other. Someone opens Perplexity, types a question, and gets annoyed because “it didn’t build me anything.” Of course it didn’t. They were in Search.

Deep Research, Model Council, Create files and apps (previously known as Labs), and Learn step by step are all now accessible using the ”+” menu in Perplexity’s updated input bar. Before you type a single word, click the ”+” and pick the right lane.

The rule is simple. Search is for “what’s the answer.” Research is for “give me a report with citations.” Labs, now labeled “Create files and apps,” is for “spend ten minutes and produce a thing I can download.” Deep Research remains the fastest way to obtain comprehensive answers to in-depth questions, typically within 3 or 4 minutes, while Labs is designed to invest more time (10 minutes or longer) and leverage additional tools, such as advanced file generation and mini-app creation.

If your deliverable is a spreadsheet, a dashboard, a slide-worthy chart, or a mini web app, you want Labs. If it’s a written brief, you want Research. Getting this wrong is why your outputs feel underpowered.

2. Ask for the artifact, not the analysis

The single biggest habit shift happens at the prompt itself. Vague research prompts get you vague research reports. Concrete artifact requests get you actual files.

Labs writes and executes code to handle tasks like structuring data, applying formulas, and creating charts, text documents, or spreadsheets. All files created during your workflow, from generated charts and images, to CSV and code files, are organized in the “Assets” panel below your query, where you can view and download everything. That Assets panel is where the value lives. Aim your prompt at it.

Bad: “Research the top five companies in AI data infrastructure.” Good: “Research the top five companies in AI data infrastructure, compare positioning and funding, and turn the results into a one-page strategy memo.”

See the difference? The second one names a deliverable. Labs will go build it, drop the memo in Assets as a downloadable file, and probably throw in a comparison chart while it’s at it. Every prompt should end with what you want to walk away with: a CSV, a dashboard, a slide deck, a PDF, a working prototype. Name the format. Name the length. Name the audience.

3. Use the App tab for anything interactive

Here’s a Labs feature almost nobody uses on purpose: Labs can develop and deploy simple interactive web apps directly below your query, enabling the creation of basic dashboards, slideshows, and websites without external development tools. That’s not a report. That’s a working prototype you can click through.

Each Lab features an Assets tab that lets you view or download all generated materials, including charts, images, CSVs, and code files. Some Labs also include an App tab that can render basic web applications directly within the project environment.

When you’d normally ask for a “report on X,” ask for an interactive version instead. “Build me a clickable dashboard comparing the five plans, with a filter for company size.” “Turn this survey data into an interactive slideshow with one insight per slide.” “Make me a mini-app where I enter a target job title and get back a ranked list of the top employers hiring for it.”

You won’t get production software. You will get something you can demo to a stakeholder in the next meeting, which is often more useful than a static PDF nobody opens.

4. Set up a Project (formerly Space) for anything you’ll do twice

If you’re only ever going to ask Perplexity one question, fine, fire off a one-shot Lab. But the moment a topic becomes recurring (a client, a beat, a research area, a course), stop working in the void. Set up a Project.

Projects are dedicated workspaces that let individual users and teams organize, collaborate, and manage research more effectively. A Project acts as a centralized knowledge hub, powered by custom AI instructions, file search, and Computer task management.

The real payoff is the custom instructions field. Projects let you configure custom instructions for every interaction within that Project, so the assistant can match the tone, domain, or reasoning style your project requires. This is where you tell it “always cite peer-reviewed sources,” or “responses in a casual tone under 200 words,” or “focus on European market data only,” and never repeat yourself again.

Upload your reference files once, Perplexity Pro subscribers can upload up to 50 files per project, while Enterprise Pro and Max subscribers can upload up to 500 files per Project, and Enterprise Max subscribers can upload up to 5,000 files per Project , and every Lab you run inside that Project inherits the context. It’s the difference between briefing a new freelancer every morning and having one who already knows the account.

5. Wire up your connectors before you need them

Nobody wants to hear this, but the tool is dramatically more useful if you take the ten minutes to hook it up to where your work actually lives.

Perplexity Enterprise users can sync files from connected apps including Google Drive, SharePoint, OneDrive, Box, and Dropbox. Any member with access to the project can query those synced files. If you’re on Enterprise, do this on day one, not the day you finally need to find something and realize your best notes are stuck in a Drive folder Labs can’t see.

The connectors also change what you can ask for. Instead of “here’s a PDF, summarize it,” you can say “look across everything in the Q3 Client Folder and tell me which pitches used the phrase ‘AI-native.’” That’s the Labs superpower nobody talks about. It isn’t just what the model knows, it’s what you’ve given it access to.

6. Be the “one-shotting ninja”, or fork and try again

Here’s the honest weakness. Labs is spectacular on the first run and mediocre on follow-ups. One Reddit user put it bluntly: “The biggest problem with Labs is that it doesn’t handle follow-ups very well. It basically requires you to be a one-shotting ninja.”

That was true, and it’s partly still true. The fix: write the entire brief up front (audience, format, length, data sources, deliverable, tone) instead of iterating in twelve tiny messages. Treat the prompt like a design brief, not a chat.

The other fix, newer, is forking. Perplexity added forking, improved inline confirmations, and enterprise controls like a Computer Analytics API to the platform. When a Lab goes 80% of the way there and you want to try a different direction without losing the good version, fork it. It’s faster than trying to redirect the same thread, and infinitely faster than starting over.

7. Chain it with Research and Model Council for high-stakes work

If the deliverable actually matters (a board memo, an investment thesis, a launch plan), don’t rely on a single Labs run. Chain the modes.

Start in Research. Deep Research is Perplexity’s premium research feature. It does not just search once, it searches dozens of times across hundreds of sources, reads full articles, compares data, and writes a detailed report.

A typical Deep Research query takes 2 to 5 minutes to complete. In that time, the system may visit 100 or more web pages, cross-references data points, flags contradictions, and produces a structured report with sections and citations. That report is your factual foundation.

Then pressure-test the interpretation with Model Council. Model Council lets you run a single query through three different AI models at the same time, you see all three responses and can compare them. You pick which three you want, and the system runs your query through all of them in parallel. When three frontier models disagree on your reading of the data, you’ve found the seam you need to think harder about.

Then take the reconciled version into Labs and ask it to produce the artifact: the dashboard, the deck, the memo. Research finds the facts. Model Council stress-tests the take. Labs builds the thing. That’s the workflow. Skipping straight to Labs on a real deliverable is how you end up with something that looks great and cites the wrong number.

One more thing: know what tier you’re on before you commit

If you’re going to lean on Labs, know the ceiling. Pro users get Labs but with monthly caps. Perplexity Max provides unlimited Labs usage per month , and Perplexity Computer, the company’s autonomous agent that executes complex workflows independently using 19 different AI models and can create subagents to handle specific problems, is available only on the $200/month Perplexity Max tier.

Most people don’t need Max. But if you’re running multiple Labs a day for client work, the Pro caps will bite you at the worst possible moment. Know before you start.

The habit that ties it all together: treat Labs like a junior analyst you’re briefing, not a search box you’re querying. Tell it the deliverable, the audience, the format, and the constraints up front, the same way you’d brief a real person. Give it the files. Give it the instructions. Then let it work for ten minutes. The people getting great outputs out of Perplexity aren’t luckier than you. They’re just briefing it like they mean it.