Perplexity Labs

Perplexity Labs is a project-building mode inside Perplexity, the AI-native answer engine, that turns a single natural-language prompt into a finished deliverable — a research report, spreadsheet, interactive dashboard, presentation, or small web app — instead of just a chat reply. Where a normal Perplexity search returns a sourced answer in seconds, a Lab runs for roughly 10 minutes or longer: it browses the live web for research, writes and executes Python and JavaScript to structure data and build charts, and compiles everything into a final output, still backed by the same cited-source standard as regular search.

Labs launched in May 2025. As part of Perplexity's February 2026 product update, the feature was renamed "Create files and apps" and moved into the app's "+" menu alongside Deep Research, Model Council, and Learn step by step — though it continues to be widely referred to by its original name, Labs. It is built by Perplexity AI, Inc., founded in 2022 in San Francisco by Aravind Srinivas, Denis Yarats, Johnny Ho, and Andy Konwinski.

1. Open Perplexity on web, iOS, or Android (Mac and Windows app support is expanding) and start a Lab — reachable from the mode selector, or since the February 2026 update, from the "+" menu under "Create files and apps."
2. Enter a detailed prompt describing the end result you want, for example "build a dashboard tracking AI startup funding rounds with charts and a watchlist." More specific prompts produce meaningfully better results than vague ones.
3. Wait while Labs works — a project typically takes about 10 minutes or longer. The Tasks pane shows the exact subtasks it's running and how information flows between them; you can tap "Notify me" for an alert instead of watching the progress bar.
4. Check the Sources pane to see what was consulted during research, and the Assets pane for every file produced, including charts, images, CSVs, and the underlying code — Python scripts, or the HTML/CSS/JS behind a web app.
5. If Labs built an interactive app or dashboard, open the Apps tab to use it live inside the project, or share it via its own dedicated URL.
6. Refine the result with follow-up prompts. Asking for the smallest possible change tends to work better than broad revision requests, which can cause the system to over-engineer and break functionality that already worked.

• Turning a business's financial CSV exports into a mini dashboard with charts that surface key metrics automatically, instead of building it by hand in a spreadsheet tool.
• Building an interactive tracker around a specific dataset — for example AI startup funding rounds, a sports team's results and fixtures, or global conflict monitoring — complete with maps, filters, and watchlists.
• Producing a multi-section research report backed by cited sources on a topic that would otherwise take several manual searches to piece together.
• Assembling a marketing campaign plan, a meal-planning spreadsheet, or a personal finance breakdown from a single prompt.
• Rapid prototyping of a small, self-contained web app — a calendar, a data explorer, a simple presentation — without opening a separate coding tool.
• Getting a downloadable, ready-to-share deliverable (a PDF/DOCX report, an XLSX spreadsheet, a PPTX/HTML presentation, or a hosted web app URL) directly from a chat-style prompt.