Perplexity
by Perplexity AI
AI answer engine with supervised research, agentic browsing, and shopping
Last reviewed 2026-06-18
Perplexity is an AI answer engine: a conversational search product that runs live web searches and uses LLMs to synthesize a direct, cited answer instead of returning blue links. It has expanded from simple Q&A into increasingly agentic features: a supervised multi-step Research mode, a project workspace (Labs) that produces reports, spreadsheets, dashboards, and mini-apps, an AI-native browser (Comet) that executes browsing workflows, and a shopping experience with in-app checkout. Its autonomy is feature-dependent. Baseline search is assistant-grade; the headline features (Deep Research, Labs, Comet) are supervised agents that a human kicks off and reviews; nothing runs fully unattended. Its differentiator is citation-grounded answers plus model flexibility. It targets consumers, knowledge workers, researchers, students, and, via Enterprise and the Sonar API, businesses and developers.
What it can do
Answer questions with cited web search
AssistantRuns real-time web searches and synthesizes a direct, source-cited answer.
sourceConduct multi-step Deep Research
SupervisedIteratively searches, reads many sources, reasons across subtopics, and produces a cited report; the human kicks it off and reviews.
sourceBuild deliverables in Labs
SupervisedRuns extended self-supervised work (deep browsing, code execution, chart and image creation) to produce reports, spreadsheets, dashboards, and simple deployed mini-apps.
sourceExecute browsing workflows in Comet
SupervisedAn AI-native browser whose assistant acts across pages to compare products, book meetings, send email, fill forms, and complete basic transactions under human supervision.
sourceResearch and buy products (Buy with Pro)
CopilotA shopping assistant with cited products and one-click native checkout that the human authorizes per purchase.
source
Strengths
- +Citation-grounded answers are verifiable, a real advantage over uncited chatbots
- +Model flexibility: not locked to one LLM, with a proprietary search-tuned Sonar family plus frontier models
- +Genuine breadth from quick answers to research, deliverables, and agentic browsing in one product
Limitations
- −The agentic features are supervised, not autonomous, and hallucination risk persists despite citations
- −Tier proliferation with the best features gated behind paid or US-only access
- −Comet's agentic browsing raises prompt-injection and account-access security concerns
Overview
Perplexity is an AI answer engine that runs live web searches and synthesizes direct, cited answers. It has grown a supervised agent layer: Deep Research, the Labs workspace, the Comet browser, and a shopping experience.
What it does
Core search returns cited answers (assistant). Deep Research iteratively reads many sources and produces a cited report. Labs runs extended work to build reports, spreadsheets, dashboards, and mini-apps. Comet is an AI-native browser that acts across pages (compare, book, email, fill forms) under supervision. Buy with Pro adds cited shopping with one-click, human-authorized checkout.
Integrations & setup
MCP connectors (Slack, GitHub, Notion, Gmail, Google Calendar), an official MCP server, and a Sonar API for developers. Models are a hybrid of proprietary Sonar plus frontier models.
Pricing
Freemium: a free tier, Perplexity Pro at $20/month or $200/year, plus higher Max and Enterprise tiers (reported) and usage-based API pricing.
Best for / not for
Best for anyone who wants verifiable, cited answers and a single tool spanning search to research to agentic browsing. Less suited to users who need fully unattended task automation or who are wary of agentic-browser security exposure.
Traction
Perplexity reportedly raised $200M at a roughly $20B valuation (TechCrunch, Sept 2025), with ARR approaching $200M per the same reporting; figures are reported, not audited.
Alternatives
Manus and Genspark are more autonomous task agents; Gemini and Microsoft Copilot offer research within larger ecosystems; Glean is the enterprise-search analog.
What people are saying
Loved for
- +chatgpt
- +content
- +search
Common gripes
- −content
Recent mentions
“The best AI tool depends on the task. Here's a stack that works. Most people default to one tool for everything. That's not a strategy. It's a habit. After months of testing, I've landed on six tools. Each one has a spec”
“Brainstorming: Gemini, ChatGPT, Claude Image generation: Midjourney, Ideogram, OpenAI Image-1.5 Video generation: OpenAI Sora 2 Pro, Kling 3.0, Google Veo 3.1 Audio: ElevenLabs, Suno, Udio Support: Dante, Zendesk AI, For”
“Everyone rushed to kill the blog. I watched it happen in real time. Keyword stuffing killed trust. AI content flooded the internet. Brands chased short-form. Newsletters became the new gospel. And the humble blog left fo”
“Lewis Hamilton is the Barcelona Grand Prix Winner "The Miracle of insassant Conviction" Formula 1 Ferrari Ferrari Group Ferrari Trento Scuderia Ferrari HP Ferrar”
“I checked if ChatGPT mentions my brand when someone asks about my industry. It didn't. Not once. Not even close. That uncomfortable discovery pushed me to build a complete Claude workflow for AEO - Answer Engine Optimiza”
FAQ
Is Perplexity an autonomous agent?+
Mostly no. Its baseline search is an assistant, and its headline features (Deep Research, Labs, Comet) are supervised agents that a human starts and reviews. Comet can take actions in the browser, but under human supervision. Nothing runs fully unattended.
What models does Perplexity use?+
A hybrid: a proprietary search-tuned Sonar family plus model-agnostic switching across frontier models from OpenAI, Anthropic, Google, and others, with a multi-model option on higher tiers.
Sources
- Introducing Comet · accessed 2026-06-18
- Introducing Perplexity Labs · accessed 2026-06-18
- Introducing Perplexity Deep Research · accessed 2026-06-18
- Perplexity reportedly raised $200M at $20B valuation (TechCrunch) · accessed 2026-06-18
Last reviewed 2026-06-18