Every Perplexity answer is built from somebody else's reporting. A user types a question, Perplexity searches the live web, reads a handful of pages and writes a short answer with a footnote against each claim. The publisher behind each footnote has done the work of answering, and its name is on screen, but the user rarely needs to click. That design explains both sides of Perplexity's relationship with media owners in 2026: it has built the only revenue pool that pays for a citation without a click, and it is defending more copyright suits from news organisations than any other AI start-up.
What is Perplexity?
Perplexity is an AI-powered answer engine founded in 2022. It sits between a search engine and a chatbot: like a search engine it retrieves current information from the web for every query, and like a chatbot it replies in prose rather than with a ranked list of links. Its product line now includes a free answer engine, paid Pro and Max tiers priced from 20 to 200 dollars a month, the Comet AI browser and agent features that carry out tasks for users.
The business has also changed shape this year. In February 2026 Perplexity dropped the sponsored follow-up questions it had tested since 2024 and told the Financial Times it had no plans to bring advertising back, because even labelled ads could undermine trust in its answers. It now makes its money from subscriptions, enterprise sales and distribution deals.
How big is Perplexity in 2026?
Perplexity is mid-sized among AI assistants: large enough to matter, far smaller than ChatGPT or Gemini. The company told the Financial Times in spring 2026 that it had more than 100 million monthly users across its search and agent products, up from roughly 30 million a year earlier. Its last published query figure was 780 million queries in May 2025.
Revenue has grown faster than usage. According to The Information, as reported by Reuters and SiliconANGLE in August 2026, Perplexity's annualised revenue reached about 750 million dollars, up from under 250 million at the start of the year, and Nvidia was in talks to invest at a valuation above 30 billion dollars. Its last confirmed round, in September 2025, valued it at 20 billion dollars. Distribution has widened too: Perplexity's search now sits inside Samsung's Bixby assistant, which the company says reaches around 800 million devices.
How does Perplexity find and cite sources?
Perplexity uses retrieval-augmented generation. When a question arrives, the system rewrites it into one or more searches, pulls candidate pages from its own index and from live fetches, ranks the passages, and hands the strongest ones to a language model that writes the answer and attaches numbered citations. Paying users can choose which underlying model writes the answer.
Because citations come from pages read at the moment of the question, ranking well on Google is not enough to be cited. Pages that win tend to answer a specific question directly near the top, use clear question-led headings, state facts precisely, show recent dates and serve their text as plain HTML that a fetcher can read without running heavy scripts. A page that buries its answer under preamble or loads its body client-side is far less likely to be quoted.
How much traffic does Perplexity send to publishers?
Not much, and its share is shrinking. SE Ranking's June 2026 study of 101,574 websites found Perplexity accounted for 7.23 per cent of AI referral traffic in 2026, behind ChatGPT at 74.78 per cent and Gemini at 11.56 per cent. Statcounter's worldwide figures for September 2026 put Perplexity at 5.84 per cent of AI chatbot referrals, down from a peak of about 12 per cent in April 2025, as larger assistants added web search of their own.
The structural point matters more than the percentage. Perplexity's answer and its sources sit on Perplexity's own page, so the user's question is usually settled before any click. A citation is real brand visibility, but it produces no page view, no ad impression and no first-party data for the publisher that supplied the facts.
Does Perplexity pay publishers?
Perplexity has tried two models, and the first has effectively lapsed. Its Publisher Program, launched in July 2024 with partners including TIME, Der Spiegel, Fortune and Automattic, promised a share of advertising revenue from answers that drew on partner content. With advertising discontinued in February 2026, that revenue line no longer exists in its original form, and Perplexity has not said what, if anything, replaced it.
The second model is Comet Plus, announced on 25 August 2025. It is a 5 dollar monthly subscription, also bundled into some paid plans, that funds a publisher pool seeded with 42.5 million dollars. Publishers receive 80 per cent of the revenue and Perplexity keeps 20 per cent. Payouts are allocated across three kinds of activity: human visits, citations in search answers and actions taken by Perplexity's agents. Launch partners named in October 2025 included Conde Nast, The Washington Post, the Los Angeles Times, Fortune, Le Monde and Le Figaro.
The citation category is the genuinely new idea. Under almost every earlier arrangement a citation without a click was worth nothing; Comet Plus puts a price on it. The limitation is transparency. As of late 2026 Perplexity had not published how many subscribers fund the pool, how many publishers are enrolled or what any individual publisher has been paid, so it is impossible to judge whether the model is material to a publisher's revenue. CNN, named as a launch partner, sued Perplexity in May 2026 after a reported payment arrangement appears to have fallen through.
Why are so many publishers suing Perplexity?
Because Perplexity reproduces current journalism at the moment of the question, publishers argue it competes directly with their own pages. The main cases as of November 2026 are:
- Dow Jones and the New York Post (News Corp): filed in October 2024 in New York. A motion to dismiss was denied in August 2025 and fact discovery runs to December 2026.
- Japanese publishers: Yomiuri Shimbun sued in August 2025, followed by Nikkei and Asahi Shimbun, who allege robots.txt instructions were ignored.
- Encyclopaedia Britannica and Merriam-Webster: filed September 2025 in New York.
- Reddit: filed October 2025 against Perplexity and several scraping firms, alleging Reddit content was harvested from Google results by circumventing Google's anti-bot protections. In July 2026 a Manhattan judge allowed the core anti-circumvention claims to proceed.
- Chicago Tribune and The New York Times: filed in December 2025. The Tribune's complaint adds a claim that the Comet browser gets around its paywall; the Times targets the retrieval process itself.
- CNN: filed in May 2026, alleging copyright and trademark infringement. Perplexity's response was that "you can't copyright facts".
None of the US publisher cases has reached a final ruling. Together they are the clearest test yet of whether live retrieval and summarisation, as opposed to model training, is covered by fair use.
What does the Comet browser change?
Comet moves Perplexity from reading pages on its own servers to acting inside the user's browser. Launched in July 2025 and free worldwide since October 2025, it can summarise the page in front of the user, fill forms and shop on their behalf. That blurs the line between a bot visit and a human one.
The legal consequences are already visible. Amazon sued in November 2025 under US anti-hacking law to stop Comet's shopping agent and won a preliminary injunction in March 2026. On 4 August 2026 the Ninth Circuit vacated it, holding that when Comet acts at a user's direction through the user's own machine, it is the user, not Perplexity, who accesses the site. Law firm Cooley noted that the decision leaves contract and terms-of-service claims open, but it suggests publishers will struggle to use computer-access statutes against agents that operate through a reader's browser. For publishers, that makes detection and commercial terms more important than legal blocking.
Can publishers block Perplexity?
Partly. Perplexity declares two user agents. PerplexityBot indexes pages for search results and is documented as respecting robots.txt. Perplexity-User fetches pages when a person asks a question, and Perplexity's own documentation says it generally ignores robots.txt because the fetch is user-initiated. Comet, running in a real browser, may be indistinguishable from an ordinary visit.
Trust is also an issue. In August 2025 Cloudflare reported that, when its declared bots were blocked, Perplexity used undeclared crawlers with generic browser user agents and rotating IP addresses, observing 3 to 6 million such requests a day across tens of thousands of domains, and removed Perplexity from its verified-bot list. Perplexity disputed the findings. Either way, a robots.txt line is a statement of preference rather than a control, and dependable identification has to come from behaviour observed at the server or CDN edge.
What should publishers do about Perplexity?
Treat it as a reader, not a referrer. First, measure it: separate PerplexityBot, Perplexity-User and suspected undeclared fetches in server or CDN logs, because client-side analytics will either miss them or count them as people. Second, decide deliberately whether Perplexity citations are worth the reads behind them, and if you allow access, structure pages so the citation is accurate and carries your brand. Third, look at compensation from more than one angle. Comet Plus may pay something for citations, licensing deals pay for access, and litigation is setting the rules for everyone else.
The read itself is the asset that remains in the publisher's control. Every Perplexity-User fetch is a live request for current content, arriving at the publisher's infrastructure before any answer is written. blankspace works at that point, detecting live search agent traffic at the CDN edge so publishers can see it, price it and decide whether to serve, limit or monetise each read.
Frequently asked questions
Is Perplexity a search engine or a chatbot?
Perplexity is best described as an answer engine. It retrieves live web pages for each query like a search engine, but it replies with a written answer and numbered citations, like a chatbot. The defining feature is that its answers are grounded in sources it names rather than produced purely from a model's memory.
Does Perplexity still run ads?
No. Perplexity phased out the sponsored follow-up questions it had tested since 2024 and, in February 2026, told the Financial Times it had no plans to bring advertising back. It now relies on subscriptions, enterprise sales and distribution deals, which means brands can only appear in its answers through organic citations.
How does Comet Plus pay publishers?
Comet Plus is a 5 dollar monthly subscription that funds a publisher pool seeded with 42.5 million dollars. Participating publishers share 80 per cent of the revenue, allocated across human visits, citations in answers and agent actions. Perplexity has not published enrolment numbers or individual payouts, so its value to any one publisher is hard to assess.
Does robots.txt stop Perplexity?
Only partly. PerplexityBot, which builds the search index, respects robots.txt. Perplexity-User, which fetches pages when a user asks a question, generally does not, by Perplexity's own description, and Cloudflare reported in 2025 that undeclared Perplexity crawlers reached sites that had blocked the declared bots. Reliable control needs behavioural detection at the server or CDN edge.
How much referral traffic does Perplexity send?
Very little relative to its reading. SE Ranking's 2026 study of more than 100,000 websites put Perplexity at 7.23 per cent of AI referral traffic, a distant third behind ChatGPT and Gemini, and Statcounter showed its share falling to under 6 per cent by September 2026. Publishers should model Perplexity as a source of citations and reads, not visits.
