Perplexity

## What it is Perplexity is an AI-powered answer engine: you ask a question, it runs live web searches, and it returns a synthesized paragraph-style answer with numbered citations linking back to the sources it used. That citation-first format is the product's whole pitch, and it's the thing that most separates Perplexity from a plain chatbot — every claim in an answer is (in theory) traceable to a specific page. Perplexity Inc. was founded in 2022 by Aravind Srinivas, Denis Yarats, Johnny Ho, and Andy Konwinski, several of them ex-OpenAI and DeepMind researchers, and the company has since raised over $1.7 billion, reaching a reported valuation above $20 billion by early 2026. Beyond the core search product, Perplexity now offers Deep Research (a slower, multi-step research mode that browses more pages before answering), Labs (agentic task execution), and Comet, a standalone AI-native browser that became free for all users in October 2025. ## Who it's actually good for Perplexity is a solid fit for people doing quick research who want an answer with a paper trail: students checking a claim before citing it, analysts scanning several sources on a topic, or anyone who's tired of scrolling past ads and SEO filler to find the actual answer on a search results page. Independent coverage from outlets like Tom's Guide and TechRadar consistently frames it as a genuine alternative to a Google search for research-style questions, and third-party comparison testing (imperfect as most of it is) tends to rank Perplexity's citation accuracy and grounding ahead of general-purpose chatbots on factual lookup tasks. If your job involves synthesizing information from multiple web sources and you want to check the receipts, it's a reasonable tool to have open in a tab. It is a weaker fit for anyone who needs a stable, predictable monthly allowance, anyone doing work where a single fabricated citation is a real liability (legal, medical, or published journalism), and anyone who has been burned before by subscription products that quietly redefine what "unlimited" or "included" means after you've paid. ## Strengths, specifically The citation model is the real differentiator. Rather than a wall of prose asking you to trust it, Perplexity attaches inline numbered links to its claims, which at minimum makes it fast to spot-check a specific assertion — a meaningfully different experience from a chatbot that states things with no attached source. The company's own pricing page lists frontier models (including GPT-5-class and Claude Opus-class options) available inside Pro, so subscribers aren't locked to one underlying model. Comet, the company's AI browser, drew real interest when it went free in October 2025 — multiple outlets covered it as one of the more usable entries in the nascent "AI browser" category, with an assistant that can summarize pages and, with more mixed results, attempt agentic tasks like shopping or form-filling directly in the browser. ## What's actually wrong with this The most substantiated and recent problem is usage limits. Multiple independent write-ups (MakeUseOf, PiunikaWeb, and community trackers aggregating user reports) document that between late 2025 and mid-2026, Perplexity repeatedly cut Pro-tier usage caps — Deep Research runs falling from roughly daily use to about 20 a month for some users, and Pro Search limits being roughly halved — without a direct notification to subscribers, including people who had prepaid for a full year. That is a real, checkable pattern, not a one-off complaint, and it directly affects the value of the plan you'd be paying for. Second, support and billing draw sustained, specific complaints. Perplexity's Trustpilot page currently sits at roughly 1.5 out of 5 across several hundred recent reviews, with the dominant themes being difficulty reaching a human for support, denied or delayed refunds, and confusion over how credits and quotas are calculated. Trustpilot review pools skew toward people who had a bad experience and went looking for a place to vent, so this number should not be read as "most users hate it" — G2, where reviewers tend to be more measured, puts Perplexity closer to 4.5–4.7 out of 5. But the specificity and consistency of the billing and support complaints across both platforms is the kind of pattern the constitution asks us to flag rather than average away. Third, Perplexity is currently the subject of active copyright litigation from several major publishers, including the New York Times, Dow Jones, the Chicago Tribune, and Japanese outlets Nikkei and Asahi Shimbun. The core allegations are that Perplexity has both reproduced substantial portions of paywalled articles in its summaries and, in some documented cases, fabricated content and falsely attributed it to a named publication. Those are allegations in active litigation, not settled fact, but the underlying behavior — an answer engine occasionally inventing content and citing a real outlet for it — is a real accuracy risk for a product whose entire value proposition is trustworthy citations. Independent analysis cited in the Times' complaint found paraphrasing and partial plagiarism of paywalled content in tested summaries. Fourth, Comet itself has drawn specific, credible performance complaints beyond the usual new-browser rough edges: high CPU and memory usage with only a handful of tabs open, sluggish or freezing tabs on a Chromium base that's supposed to be lighter, and an agentic assistant that handles simple tasks fine but breaks down on multi-step ones — hallucinated dates while trying to book a parking spot, and failures partway through checkout or hotel-booking flows in reviewer walkthroughs. ## Pricing reality As of this review, Perplexity's own pricing page lists Free ($0), Pro ($20/month or $200/year), and Max ($200/month or $2,000/year) for individuals, plus a discounted Education Pro tier, and Enterprise Pro ($40/seat/month) and Enterprise Max ($325/seat/month) for organizations. The catch, per the usage-limit reporting above, is that the number attached to "Pro searches per week" or "Deep Research queries per month" has moved more than once in the past year, generally downward, and not always with clear notice to people already paying. Treat any specific quota figure as something to verify at time of purchase rather than something to trust from a review written months earlier. ## Verdict Perplexity's core idea — search with receipts — is genuinely useful and is executed better than most competitors attempting the same thing, which is why independent coverage keeps ranking it near the top of the AI-search category. But the company's recent pattern of quietly shrinking what a paid plan includes, combined with specific and sustained billing and support complaints, means the product is more trustworthy on any given query than it is as a subscription to commit money to without checking back regularly. Use it, don't assume the plan you signed up for is the plan you'll have in six months. ## Quick take **Verdict:** A genuinely useful cited-answer search tool undercut by a pattern of quietly shrinking paid plan limits and sustained billing/support complaints. **Choose it if:** - You want quick, source-linked answers for research instead of a page of search-result links - You want to try a free AI browser (Comet) without committing to a paid plan - You're comparing AI search tools and citation traceability matters more to you than raw model power **Skip it if:** - You need a subscription with usage limits that stay fixed for the length of your billing term - Your work can't tolerate an occasional fabricated citation attributed to a real publication - You've had billing or refund problems with a SaaS product before and that's a dealbreaker - You need reliable multi-step browser automation today rather than an early-stage agentic feature **Watch out:** Perplexity has cut Pro-tier usage limits (Deep Research and Pro Search caps) multiple times between late 2025 and mid-2026, including for users who had already paid annually, generally without direct notice — verify current limits on the pricing page before buying, don't rely on what a review (including this one) says the caps are.