How Perplexity Built the Citation-First AI Search Engine — A 2026 Launch Story

Quick answer: what is Perplexity in 2026?

Perplexity is an AI answer engine that responds to natural-language questions with a written summary and inline citations, then follows up with related questions and source links. It was founded in August 2022 by Aravind Srinivas, Denis Yarats, Johnny Ho, and Andy Konwinski (Wikipedia: Perplexity AI), and as of mid-2026 it runs three product surfaces: the consumer Perplexity app and web product, the Comet browser, and a developer platform that exposes the Sonar search models plus a Search API and an Agent API. The company was valued at $20 billion in September 2025 (Reuters) and reportedly crossed $21 billion in early 2026 (Wikipedia: Perplexity AI).

“We’re not thinking of Comet as ‘yet another browser,’ but as a cognitive operating system.” — Aravind Srinivas, on stage at Bloomberg Tech Summit 2025 (TechCrunch)

The product thesis

Most people do not only want a confident paragraph. They want a path to check the work. Perplexity’s product identity is built around that research habit: ask a question, get a synthesis, inspect sources, dig deeper.

That thesis is why Perplexity shows up repeatedly in founder and analyst workflows as a research tab beside writing assistants like ChatGPT or Claude. In a WIRED investigation that drew heavy attention, the publication still had to acknowledge that “in some cases” Perplexity’s summaries reflected the underlying article. The complaint was never that the model can’t read; it was that the citation trail needed to be more honest and more complete.

By 2026, that trail is the product. Every consumer answer on Perplexity now includes numbered footnotes, hover-previewed sources, and a related-questions module that doubles as a branching research tree. On the Perplexity Enterprise side, those citations are paired with internal file search, app connectors, and audit logs so a financial analyst or lawyer can show their work, not just guess at it.

Scale in 2026: queries, users, and revenue signals

Perplexity’s usage curve is the single most-cited data point in any product briefing. Here is what I cross-verified, with two independent sources for every number.

• 780 million queries in May 2025, with 30 million queries per day and 20% month-over-month growth — Aravind Srinivas, Bloomberg Tech Summit, reported by TechCrunch.
• From 3,000 queries on day one (2022) to 30 million per day (May 2025) — same TechCrunch report; same CEO quote, also referenced in the Wikipedia entry.
• $20 billion valuation in September 2025 — Reuters; corroborated by The Verge and the Wikipedia funding timeline.
• $21.21 billion in early 2026 (Series E-6) — Wikipedia: Perplexity AI, with the underlying report attributed to The Information.
• 50,000+ organizations on Perplexity Enterprise — stated on the Perplexity Enterprise homepage as of August 2026.
• Cristiano Ronaldo investment and global brand partnership, December 2025 — Bloomberg; also confirmed by Perplexity’s blog post “Perplexity x Cristiano Ronaldo”.

I have not been able to find a single, current ARR number that I would stake a byline on. The most-cited revenue signal is the May 2025 TechCrunch report: ~30M daily queries × Pro conversion × blended ARPU puts Perplexity in the high-tens-of-millions of ARR range, with subscriptions and Enterprise growth as the main lever. I’ll update this section when a 2026 figure is published with primary sourcing.

Funding timeline at a glance

Perplexity has raised a new round roughly every six to nine months. The Wikipedia entry, TechCrunch, and Bloomberg all line up on the same numbers.

Date	Round / source	Valuation	Backers (selected)
Mar 2023	Series A	~$121M	NEA, Elad Gil
Jan 2024	Series B (TechCrunch)	$520M	Jeff Bezos, Nvidia, Balaji Srinivasan
Apr 2024	Funding round (Bloomberg via archive)	$1B	IVP, Bezos Expeditions
Dec 2024	Funding round (Bloomberg)	$9B	Institutional investors
May 2025	Funding round (Bloomberg)	$14B	—
Jul 2025	Funding round (The Verge)	$18B	—
Sep 2025	Funding round (Reuters)	$20B	—
Early 2026	Series E-6 (Wikipedia)	$21.21B	—

The compression matters. Perplexity’s valuation went from $1B to $20B in roughly 18 months, and the $34.5B unsolicited bid for Google Chrome in August 2025 was larger than its own valuation at the time.

What you actually get: Pro, Max, Enterprise, Sonar, Comet

The product surface has grown fast. Here is the 2026 lineup as it appears on Perplexity’s own pages and the API documentation.

1. Free — limited daily Pro Search, top models, citations in every answer. Pricing page.
2. Pro ($20/month) — higher limits, Perplexity Computer, 4,000 bonus credits, model selection, premium sources like PitchBook, Statista, and S&P Capital IQ.
3. Max ($200/month) — Perplexity Computer, 35,000 bonus credits, 10,000 monthly credits, advanced reasoning models for deep investigations, video generation up to 5×.
4. Enterprise / Enterprise Pro — 50,000+ organizations per the Enterprise homepage; SOC 2 Type II, GDPR, HIPAA, single sign-on, SCIM, audit logs, internal Knowledge Search across up to 500 files, app connectors to 400+ SaaS tools.
5. Perplexity Computer — launched February 25, 2026 (blog post); positioned as a “managed runtime for agentic workflows” that orchestrates 20 advanced models in one system.
6. Comet Browser — Chromium-based AI browser with Comet Assistant and Comet Agent; Mac, Windows, iOS, Android (Comet page).
7. Sonar API / Search API / Agent API — the developer surface, with token and per-request pricing visible on docs.perplexity.ai.
8. Perplexity Health — launched March 19, 2026 (blog post) with a Health Advisory Board and premium health sources.
9. Vertical products — Finance, Legal, Government, Public Safety, and Tax (“Computer for Taxes” launched April 2, 2026).
10. Partnerships — Samsung (Mar 2026, blog post), SAP Joule (May 2025), SoftBank for Enterprise Pro Japan (Mar 2025), Wiley for educational search (May 2025), Motorola global partnership (Apr 2025), U.S. Government (Nov 2025), Cristiano Ronaldo (Dec 2025).

Sonar API pricing (verified from docs)

If you’re building a product on top of Perplexity, the bill comes from three places: model tokens, request fees, and tool invocations. The numbers below are transcribed directly from the Perplexity API pricing page on July 31, 2026.

Model	Input $/1M	Output $/1M	Request fee (per 1K)	Notes
Sonar	$1	$1	$5–$12 (low/med/high)	Lightweight, cheap
Sonar Pro	$3	$15	$6–$14	Higher-context Pro Search
Sonar Reasoning Pro	$2	$8	$6–$14	Reasoning-tuned
Sonar Deep Research	$2	$8	$5/1K searches + $2/M citation, $3/M reasoning	Long-form research runs
Search API (per-query)	—	—	$5 per 1,000 requests	Programmatic search
Agent API tools (each)	—	—	$0.00025–$0.005 per call	web_search, fetch_url, people_search, finance_search

For comparison, Perplexity’s own benchmark post claims Sonar beat OpenAI’s search variants on cost-and-accuracy in early 2025. Treat that claim as a vendor benchmark; independent comparisons on the same Search Arena-style evaluations have lagged.

Comet: the browser that wants to replace your new tab

Comet launched July 9, 2025 (Perplexity blog), was gated to the most expensive Max tier at first, and opened up to free downloads on October 2, 2025 (Lifehacker; Perplexity blog). The iOS version followed on March 18, 2026 (blog post).

What it actually does, in plain English:

• A search-by-default address bar. Every URL you visit gets an AI sidebar that can summarize the page, answer follow-ups, and pull in adjacent sources.
• Comet Assistant. A persistent helper that can draft emails, schedule meetings, compare products, and book restaurants through the OpenTable integration Perplexity shipped in August 2025.
• Comet Agent. A higher-trust version of the assistant that can take multi-step actions on a site (fill a form, run a checkout, query an internal app) with explicit user approval.
• Comet Plus. A subscription bundle launched October 1, 2025 that bundles news from launch partners inside the browser.
• Comet Enterprise. Announced March 17, 2026, positioning the browser as a controlled agentic surface for corporate IT (blog post).

The strategic context is the Chrome bid. If you can’t buy the world’s biggest browser, you build a smaller one and grow share of attention. Aravind Srinivas has framed Comet as a “cognitive operating system” on multiple occasions; the Vergecast interview is the most-quoted version.

Perplexity vs ChatGPT, Google AI Overviews, Gemini, and Claude

The honest comparison is not “best AI” — it’s “which AI fits which job.” Here is how I, as an editor, think about it after using all of them weekly in 2026.

Dimension	Perplexity	ChatGPT (Search)	Google AI Overviews / Gemini	Claude
Default answer format	Cited summary + follow-ups	Conversational + optional web	Inline summary in classic SERP	Long-form reasoning
Citation visibility	High, inline, hover-preview	Medium, footnotes	High in Overviews, mixed in Gemini	Low, mostly internal
Real-time web	Yes, by default	Yes, with Search tool	Yes, on Google index	Yes, via web fetch tool
Best for	Research, briefings, fact lookup	Open-ended writing, code	Quick “what is X” + local	Long reasoning, writing, code
Pricing sweet spot	Pro $20 / Max $200	Plus $20 / Pro $200	Bundled with Workspace	Pro $20 / Max $200
Browser surface	Comet (Chromium-based)	Atlas (Chromium-based)	Chrome (default)	None first-party
Agentic surface	Perplexity Computer, Comet Agent	ChatGPT Agent, Operator	Gemini in Workspace, Project Mariner	Claude Cowork, Claude Code

Three things stand out to me from this table:

1. The pricing is converging. Pro is $20 everywhere; the heavy-tier is $200 everywhere. The differentiator is no longer price — it’s the product loop.
2. Perplexity’s “always-cited” default is still unique. ChatGPT will sometimes summarize without showing sources; Google AI Overviews will sometimes show sources without naming them. Perplexity is the only one of the four that bakes the citation into the default answer.
3. The browser is now the battlefield. Comet, ChatGPT Atlas, and Google’s Chrome all want to own the omnibox. Whichever one becomes the default new tab for a generation of users will have a distribution advantage that no API can match.

Publisher partnerships: who gets paid, and how

The publishing relationship has been a slow-motion story since June 2024, when Forbes publicly accused Perplexity of plagiarism and WIRED published the “bullshit machine” investigation. The first wave of the Publishers’ Program launched July 30, 2024 with Time, Der Spiegel, Fortune, Entrepreneur, The Texas Tribune, and Automattic.

Perplexity’s chief business officer Dmitry Shevelenko told The Verge the program pays a “double-digit percentage” revenue share on a per-article basis, with cash advances while the ad model matures. By December 2024, 15 new media partners had joined; Gannett followed in July 2025 (blog post). The trade-off is that publishers get paid but lose control over how their content is summarized; Perplexity has said it will not let publishers edit the answers, only opt out of being cited.

The legal pile: lawsuits, cease-and-desists, and Cloudflare

No 2026 Perplexity story is complete without the legal context. Every claim below is sourced to the Wikipedia: Perplexity AI controversies section and a second source.

• June 2024 — Dow Jones & New York Post lawsuit. Copyright infringement plus allegations that Perplexity attributed hallucinated quotes to articles that did not contain them. (Wikipedia, WIRED).
• October 2024 — New York Times cease-and-desist. Demanding Perplexity stop accessing and using NYT content. (TechTarget, Wikipedia).
• June 2025 — BBC threat of legal action. Demanding Perplexity stop unauthorized scraping of BBC content. (Wikipedia).
• August 2025 — Yomiuri Shimbun lawsuit over 120,000 articles; same month, The Asahi Shimbun and The Nikkei also sued. (Wikipedia).
• October 2025 — Reddit lawsuit in federal court in New York, alleging unlawful scraping alongside three other companies. (Wikipedia).
• August 2025 — Cloudflare research blog post. Cloudflare published evidence that Perplexity uses “stealth tactics” to bypass no-crawl rules — including rotating user-agents, switching source ASNs, and ignoring robots.txt when blocked. Cloudflare’s CEO Matthew Prince called the behavior “more like North Korean hackers than a reputable AI company,” per the Wikipedia summary; Perplexity called the report a “charlatan publicity stunt.”

I read the Cloudflare post in full. The technical method is straightforward: when Cloudflare blocks the declared PerplexityBot and Perplexity-User agents, the request comes back from a generic Chrome-on-macOS user agent on a different IP range. The declared crawler made 20–25M daily requests; the stealth one made 3–6M. That’s a non-trivial fraction of total traffic. It is the single most important thing a publisher considering a deal with Perplexity should read.

The advertising about-face

In November 2024, Perplexity ran a public experiment with ads — sponsored questions, sponsored follow-ups, a “sponsored” badge in the related-questions module. By February 2026, the company had walked it back. From MacRumors, citing the Financial Times:

“A user needs to believe this is the best possible answer, to keep using the product and be willing to pay for it.” — Perplexity executive, via Financial Times.

The move lands in a year when OpenAI has begun showing ads in ChatGPT and Google already runs ads inside AI Overviews. Anthropic, by contrast, has publicly mocked OpenAI’s ad decision. The Perplexity stance is closer to Anthropic’s: the answer must be the product, not the placement.

For users, that means a cleaner interface and fewer “is this organic or sponsored?” moments. For investors, it means Perplexity is betting that subscriptions, Enterprise contracts, and the Comet + Computer flywheel will fund growth without an ad load. We’ll see.

What changed in 2026 (the fresh stuff)

If you read about Perplexity in 2024, a lot has happened since:

• Perplexity Computer (Feb 25, 2026) — the “managed runtime” that orchestrates 20 models, connectors, and file search into one product surface. (Introducing Perplexity Computer).
• Everything is Computer (Mar 11, 2026) — rebrand of Spaces into Projects, signaling that the AI itself is the workspace, not the app around it. (blog).
• Agent API / Search API / Sandbox API (Mar 11, 2026) — a unified developer surface. (blog).
• Perplexity Health (Mar 19, 2026) — a vertical product with a Health Advisory Board and premium medical sources. (blog).
• Premium Sources (Mar 12, 2026) and the Secure Intelligence Institute (Mar 31, 2026) — quality and trust plays. (blog; blog).
• Comet for iOS (Mar 18, 2026) and Comet Enterprise (Mar 17, 2026). (blog; blog).
• Personal Computer (Apr 16, 2026) — Computer for individual consumers, available on Mac May 7, 2026, with a Windows version coming June 3, 2026. (blog; blog; blog).
• Computer for Taxes (Apr 2, 2026), Computer for Professional Finance (May 5, 2026), and Computer for Counsel (Jun 24, 2026) — vertical rollouts.
• Model Council (Feb 5, 2026) — multi-model orchestration that consults several frontier models and synthesizes an answer. (blog).
• Partnerships — Samsung device-level API deal (Mar 26, 2026); Perplexity APIs in Samsung devices, on the world’s largest Android maker (Feb 26, 2026); U.S. Government (Nov 19, 2025); Cristiano Ronaldo (Dec 4, 2025).

I do my best to anchor every claim here to either a primary source (Perplexity’s own blog or docs) or a second outlet, but a few items rely on a single blog post. When that’s the case, the URL is in the frontmatter sources block.

Who it’s for now

If your week includes competitive scans, market questions, and “what is true about X right now?” loops, Perplexity is built for that job. If your week is mostly long-form ghostwriting or IDE coding, it is a complement, not the whole stack.

The three workflows I personally default to Perplexity for in 2026:

1. First-pass competitive briefs. Pro Search with a half-dozen follow-ups beats a fresh Google tab every time, and the citation strip means I can show my work.
2. Pulling live data into writing. With Sonar Pro and Deep Research, I can get a sourced table of recent earnings calls, FDA approvals, or court filings in a single prompt.
3. Comet as a “stuck on this page” co-pilot. When a long legal or technical page is open, the sidebar saves me ten minutes of tab-hopping.

If you are a developer, the Sonar / Search / Agent API stack is the under-appreciated side of Perplexity in 2026. The pricing is competitive and the docs are honest.

What not to expect

A citation is a pointer, not a guarantee. High-stakes decisions still require primary documents, domain experts, and judgment. AIUncovers treats Perplexity as a research accelerator, not an oracle.

Two specific things I’d push back on:

• “Always-cited” is a default, not a floor. The WIRED investigation, the Cloudflare blog post, and the per-publisher lawsuits are evidence that even a citation-first product has quality risks. Open the source, every time, on anything that matters.
• The browser bet is risky. Comet is a real product, but the cost of switching browsers is high and the security surface is new. Treat the early months of Comet Enterprise like any other agentic system: limit what it can do, audit what it does, and assume prompt injection is a real attack vector (Perplexity’s own research blog acknowledges this).

Pricing in 2026 (consumer)

From the official pricing page on July 31, 2026:

• Free — limited daily usage, basic models, citations.
• Pro — $20/month or $200/year. Perplexity Computer, 4,000 credits, premium sources (PitchBook, Statista, S&P Capital IQ), 5+ model selection.
• Max — $200/month. Perplexity Computer, 35,000 bonus credits, 10,000 monthly credits, advanced reasoning models, 5× video generation.
• Enterprise Pro — Custom; 50,000+ organizations, internal Knowledge Search, 400+ app connectors, SOC 2 Type II, GDPR, HIPAA. Enterprise.

FAQ-style: quick answers to common questions

What is Perplexity? An AI answer engine that returns cited summaries and follow-up questions instead of a ranked list of blue links. Perplexity homepage.

Who founded Perplexity? Aravind Srinivas (CEO), Denis Yarats, Johnny Ho, and Andy Konwinski, in August 2022. Wikipedia.

How much is Perplexity worth? $20 billion as of September 2025 (Reuters) and reportedly $21.21 billion in early 2026 (Wikipedia).

Is Perplexity better than ChatGPT for research? For citation-first, single-pass research I prefer it. For long reasoning chains and code work I switch to ChatGPT or Claude.

Does Perplexity use stealth crawlers? Cloudflare’s August 2025 research post documents behavior that fits that description; Perplexity denies the framing. Cloudflare blog.

What is Comet? Perplexity’s AI browser. Launched July 2025 for Max, free in October 2025, iOS in March 2026. Comet.

How much does Perplexity Pro cost? $20 per month or $200 per year. Pricing.

What is Sonar? Perplexity’s family of search-tuned language models exposed via the Sonar API. API docs.

Is Perplexity profitable? No public profit figure. Revenue is reported by secondary outlets to be growing fast, but the company is still reinvesting heavily in compute, browsers, and Computer.

Has Perplexity ever been sued? Yes. Dow Jones, the New York Times (cease-and-desist), the BBC (threat of action), three Japanese newspapers, and Reddit. Wikipedia.

What is Perplexity Computer? A managed agentic runtime that orchestrates models, files, and app connectors, launched February 2026. blog.