company profile
Perplexity: The Answer Engine Story
Perplexity was founded in August 2022 by Aravind Srinivas. It pioneered the 'answer engine' format: a chatbot that cites sources. Pro is $20/month; Comet is its AI browser; Nvidia is a major investor.
5 min read · Updated Apr 17, 2026
The chatbot that replaced Google for a lot of people
Perplexity bet early on a different shape for AI chat — not an assistant, but a search engine with a soul. Founded in August 2022 by Aravind Srinivas and three co-founders, it was the first mainstream chatbot to ship with inline citations for every claim, and that one decision has shaped the entire research-assistant category.
Timeline
- August 2022 — Founded by Aravind Srinivas, Denis Yarats, Johnny Ho, Andy Konwinski.
- December 2022 — Public launch of perplexity.ai.
- 2023 — Perplexity Pro launches at $20/month; first round at $25M valuation.
- 2024 — Valuation crosses $3B; Nvidia becomes strategic investor.
- 2024 — Perplexity for Enterprise and Perplexity Spaces launch.
- 2025 — Comet browser releases in public beta.
- 2025 — Deep Research mode ships, competing directly with OpenAI’s equivalent.
- 2026 — Comet goes mainstream; Pro stays at $20 with an expanded lineup of selectable models.
Leadership
- Aravind Srinivas — CEO. Ex OpenAI research scientist, PhD from Berkeley.
- Denis Yarats — CTO. Ex Meta and Facebook AI.
- Johnny Ho — Chief Strategy Officer.
Flagship products
- Perplexity.ai — free answer engine.
- Perplexity Pro — $20/month.
- Perplexity Spaces — saved workspaces and shared threads.
- Comet — AI-native browser.
- Sonar — in-house fine-tuned models optimized for search.
- Perplexity API — search-grounded LLM answers for developers.
Business model
Three lines:
- Pro subscriptions — $20/month is the core.
- Enterprise seats — growing; Amazon, Zoom, Dell, and others have deployed.
- Ads and shopping — Perplexity has experimented with sponsored answers and a shopping layer; direction still evolving.
What makes Perplexity distinct
- Citations by default — every answer links to sources inline.
- Model choice on Pro — subscribers can route the same question to GPT-5, Claude Opus 4.6, or Gemini 3. No other $20 plan gives access to all three frontier models.
- Search-first UX — faster, more focused, less “conversational overhead” than ChatGPT.
- Comet browser — a real contender against ChatGPT Atlas for the AI-browser category.
Why this matters for picking a subscription
Perplexity Pro at $20 is the best deal in AI if you think of yourself primarily as a researcher rather than a writer or coder. Being able to swap between GPT-5, Claude Opus 4.6, and Gemini 3 inside a single subscription is uniquely useful. Writers still tend to prefer Claude; coders tend to prefer ChatGPT or Claude Code; but for “I need to find and understand information,” Perplexity is very hard to beat.
Going deeper
Frequently asked questions
Who founded Perplexity?
Perplexity AI was founded in August 2022 by Aravind Srinivas (CEO, ex OpenAI), Denis Yarats (CTO, ex Meta/Facebook AI), Johnny Ho (ex Quora), and Andy Konwinski (ex Databricks).
What is an 'answer engine'?
Perplexity's framing for a search-replacing chatbot: you ask a question, it searches the web, reads the top sources, and writes a cited answer. Unlike ChatGPT, every factual claim links to a source by default.
How much does Perplexity cost?
Perplexity Free is $0 with basic access. Perplexity Pro is $20/month (or $200/year) and unlocks GPT-5, Claude Opus 4.6, Gemini 3, and $5 of monthly API credits. Enterprise tiers exist for teams.
What is Comet?
Comet is Perplexity's AI-native web browser. It embeds the answer engine into every page, supports agent-style workflows, and directly competes with ChatGPT Atlas.
Who owns Perplexity?
Perplexity is private. Investors include Nvidia, Jeff Bezos, IVP, NEA, Elad Gil, Nat Friedman, and Accel. Nvidia is a particularly visible strategic backer.
Does Perplexity train its own models?
Perplexity fine-tunes open-weight models (Sonar family) for search-grounded answers but relies on OpenAI, Anthropic, and Google models as selectable options on Pro. The answer engine layer, not the base model, is the product.