Overview
Perplexity is an “answer engine”: you type a question, it reads the web and writes you a paragraph with numbered citations underneath. As of September 2026 it reports roughly 45 million monthly active users, and its annualized revenue has climbed past $750 million, up from about $232 million at the end of 2025. Nvidia was reportedly in talks in late August 2026 to invest at a valuation above $30 billion, more than 50% above the $20 billion mark Perplexity set in a June round. Comet, its Chromium-based AI browser, launched in July 2025 for Max subscribers, went free worldwide that October, and arrived free on iOS in March 2026.
The product is built as the anti-Google, and on one axis it genuinely is. Perplexity stopped taking new advertisers in October 2025 and confirmed a full exit from advertising in February 2026, with executives arguing that sponsored answers would make users “suspicious of everything.” Ads had been under 0.1% of revenue, so the sacrifice was small, and the resulting business model is subscriptions and enterprise contracts, plus a growing bet on an AI agent that runs on your own hardware.
That agent, called Computer, is most of why this entry’s grade moved this year. Perplexity shipped an always-on Mac version, Personal Computer, in April 2026, opened it to $20-a-month Pro subscribers in May, and shipped a local, on-device version called Portable Computer to Windows RTX PCs on September 14. The company credits the Computer line with most of its 2026 revenue growth, and it changes what this entry is rating: not just a search box anymore, but a system that acts inside your files and accounts.
What It Knows About You
The web product knows your queries, your account, and your search history. Comet knows considerably more, because an AI browser reads the page you are looking at and can act inside sessions you are already logged into. Computer, in its cloud form, and Personal Computer and Portable Computer in their local forms, go further: they get standing access to your files and native apps so they can carry out multi-step tasks, such as assembling a slide deck or pulling revenue by vertical from a data warehouse, without you re-prompting at each step.
AI training on consumer plans (Free, Pro, Max) is on by default; opting out means finding the “AI data retention” toggle in account settings, and it applies only going forward. On July 2, 2026, Perplexity replaced its privacy policy with a restructured privacy notice, updated again on July 8 to add detail on cookies and advertising measurement. Three things that had been written into the earlier document are absent from the current one: the AI-training opt-out, the 30-day account-deletion timeline, and the promise to notify users of material changes. New versions are now simply posted to the site. The corresponding controls still exist in the product, but a commitment written into a policy and a switch in a settings panel are different kinds of promise, and only one of them is hard to withdraw.
The same mid-2026 update expanded what the notice says the company may collect: Comet browsing data, voice, and, for anyone enrolled in Perplexity Health, health, genetic, biometric, and neural data. Perplexity states it does not sell or share sensitive personal data for cross-context behavioural advertising and transfers it to third parties only when instructed to.
Retention still differs sharply by tier. Consumer uploads are generally deleted after about 30 days. Enterprise Pro/Max is not used for training and holds uploads roughly 7 days, with custom retention available. The Sonar API is zero-retention. For the Computer product line specifically, Perplexity has not published a separate retention timeline for the audit logs that record what the agent did in your files and apps, a notable gap for a product whose main selling point is that you can review its actions after the fact.
The Real Risks
How the content is acquired, and who is suing over it. On August 4, 2025, Cloudflare published findings that Perplexity was reaching sites that had blocked PerplexityBot by falling back to undeclared crawlers using a generic Chrome-on-macOS user agent and rotating IP addresses, putting the volume at 20 to 25 million daily requests from the declared crawler plus 3 to 6 million more from the undeclared one. Perplexity denied the findings and called the report a publicity stunt. The docket since has only grown: Dow Jones and the New York Post sued in October 2024, Yomiuri Shimbun in August 2025, Encyclopaedia Britannica and Merriam-Webster in September 2025, Reddit in October 2025, the Chicago Tribune and the New York Times in December 2025, and CNN on May 28, 2026, its first AI copyright suit against any company. A July 31, 2026 ruling in the Reddit case dismissed some secondary claims but let Reddit’s core anti-circumvention claims proceed. This is a risk to you as a user, not only to the company: an adverse ruling can change what a search product is permitted to show you.
Agentic surfaces keep finding new ways to misbehave. Brave’s security team documented indirect prompt injection in Comet in August 2025, instructions hidden in a page that could steer the assistant into reading saved data and exfiltrating it to an attacker’s server. Zenity Labs went further in March 2026 with PleaseFix, where accepting a calendar invite was enough to trigger a zero-click leak of local files, and LayerX’s CometJacking showed a single malicious link could hijack a session; Perplexity initially closed that report as “Not Applicable” before shipping a fix. Amazon sued in November 2025 over Comet’s assistant operating inside logged-in Prime sessions; a district judge enjoined it in March 2026, and the Ninth Circuit vacated that injunction on August 4, holding that under the CFAA it is the user, not Perplexity, who “accessed” Amazon’s computers. That is a landmark for agentic commerce, and it also means that when an agent acts inside your account, the law currently treats that as you doing it.
Computer removes the confirmation step, not just the browser tab. Unlike Comet, which still needs you looking at a page, Computer and its local siblings run standing tasks against your files, apps, and connected services around the clock. Independent reviewers have noted it does not ask clarifying questions before acting: it starts executing the instruction it was given, correct or not. Perplexity’s own pitch names the jobs it targets, the recurring work of financial analysts, marketing strategists, and corporate researchers. Co-founder and CEO Aravind Srinivas has said the goal is to help businesses “run as autonomously as possible,” and, asked about the job losses that follow, argued that “most people don’t enjoy their jobs” anyway. Whatever the merits of that view, it is a plain statement of intent from the person building the product.
Accuracy. Columbia’s Tow Center tested 1,600 queries across eight AI search engines in March 2025 and found them wrong more than 60% of the time. Perplexity was the best performer, at 37% wrong. Best in class still means a bad citation in more than one answer in three, delivered in the same confident register the Computer agent now uses when it decides what belongs in your spreadsheet.
Alternatives
- Kagi, paid, ad-free search that returns links rather than conclusions. You do the judging; nothing is trained on you.
- Brave Search or DuckDuckGo, free, less profiling than Google, and Brave runs its own index rather than reselling someone else’s.
- Perplexity Enterprise or the Sonar API, if the tool has to be Perplexity because it is already in your team’s workflow, these tiers do not train on your data and retain far less.
- Comet is optional. Most of the browser-specific exposure in this entry disappears if you use perplexity.ai in an ordinary browser instead.
- So is Computer. Nothing about the underlying search product requires an agent with standing access to your files. If the appeal is automation, weigh a narrow, single-purpose tool against a system that can act across your whole machine without pausing to check with you.
- The low-tech option: a reference librarian, a subject database, or the primary document itself. Slower, and right substantially more often than 63% of the time.
Our Verdict
D, use with caution. As of September 2026, Perplexity is still one of the less predatory products in this Index on the advertising axis, and over the past year it has built an agent that runs on your machine and inside your accounts, which is a different category of risk than a search box carries. Both are true at once.
The scores reflect that shift. Privacy holds at 7: the July 2026 notice moved key protections out of the document while widening what the company says it may collect, and Comet and Computer both see a great deal, though Portable Computer’s local-only mode is a real mitigation when you choose it. Bias stays at 5 for the citation-accuracy problem. Job threat rises from 6 to 7: Computer is explicitly built to automate the recurring work of analysts and researchers, and the company’s own CEO has framed the resulting job losses as a trade worth making. Autonomy rises from 6 to 8: an always-on agent with standing file and app access, that does not pause to confirm an ambiguous instruction, erodes user judgement more than a browser extension does.
Lock-in moves from 3 to 4. That is still not high. There is no proprietary format holding your work, and closing Comet costs nothing. But once Computer is running recurring workflows against your own data, the cost of leaving is the automation you would have to rebuild elsewhere, not just a subscription to cancel.
What would move the grade. Down: a documented security incident in Computer or Personal Computer on the scale of Comet’s, evidence that health or agent-session data is used beyond the feature that collected it, or a ruling that forces the product to strip citations or sources. Up: published retention limits for Computer’s audit logs, a durable licensing settlement with the major plaintiffs, and local-only processing as the default rather than an opt-in most users never find.