Where you go when you need a direct, cited answer — not ten blue links or a chatbot hallucination. Perplexity is replacing search as the default way people ask the web a question.
You type a question. You get 10 SEO-optimized results. You click, skim, cross-reference, and synthesize. The work happens in your head.
LLMs produce fluent text, but without citations you can't tell what's real. You have to verify everything, which defeats the speed.
Every analyst, student, lawyer, and operator runs the same cycle 30–50 times a day: question → search → read → synthesize. A productivity tax paid in minutes.
GPT-4 class models crossed the threshold where multi-source synthesis is reliably better than human-eyeballing. Citations + reasoning = trust.
Gen Z doesn't Google — they ask. TikTok, Reddit, ChatGPT. The "default answer source" is up for grabs for the first time since 1998.
Inference cost dropped ~90% in 24 months. A cited, synthesized answer now costs cents, not dollars. Advertising + subscription both viable.
Perplexity is the answer engine built for how people actually want to use the internet. You type a question in natural language. We retrieve the most relevant, recent sources in real time. An LLM synthesizes them into a direct answer with inline citations.
Every answer is grounded, cited, and editable. You can ask follow-ups that inherit context. You can browse the sources. You can share the thread. It's the workflow millions of people hacked together with ChatGPT + Google — rebuilt natively.
Our conviction: the default question-asking surface on the internet is about to change. We're building the thing it changes to.
The web-first answer engine. Real-time retrieval, inline citations, follow-up threads, copilot mode for deeper queries.
The subscription tier with frontier models, unlimited Pro Searches, file & image uploads, and Spaces for research workflows.
SSO, audit, internal knowledge sources. API gives developers the answer engine as a backend primitive for any product.
Free tier is the acquisition engine. Shareable answer threads carry the brand into every knowledge workflow. Word of mouth among students, analysts, writers.
Free → Pro conversion via quota, model selection, Spaces. Credit-card-only, no sales. Runs on product signal (query depth, return rate).
Named accounts in F500 research-heavy orgs (finance, law, consulting, gov). Pro usage signals drive outbound. Team-landing → expansion in quarters not years.
Answer-engine-as-a-service for teams building AI products. Every customer built on Sonar is a distribution partner for the consumer brand.
Former research scientist at OpenAI (GPT models) and DeepMind. PhD Berkeley. Published on retrieval-augmented generation — the technique Perplexity is built on.
Former research scientist at Meta AI. PhD NYU under Yann LeCun. Core contributor to the retrieval-ranking stack that differentiates Perplexity's answer quality.
Led by [REPLACE: lead]. [REPLACE: round structure]. Existing investors committing to pro-rata.