Perplexity is an AI-native answer engine. Every response is synthesized from live web sources with numbered inline citations (no list of ten blue links, no scraped snippets stripped of attribution).
Perplexity’s core product is cited AI search. Paid subscribers can select among recent GPT, Claude, Gemini, and other supported models, while the Sonar API exposes search-grounded models for developers.
Recent developments (March-May 2026)
- May 8: Morningstar and PitchBook added Perplexity integrations for finance research workflows, strengthening Perplexity’s pitch as a cited, source-backed work surface for analysts.
- May 6: Snap said its planned $400 million Perplexity AI search deal ended amicably, making Perplexity’s direct app, Comet, enterprise, and Computer strategy more important than Snapchat distribution.
- April 24: A retrieval-poisoning demo exposed how thin, circular sources can contaminate AI search answers. This is directly relevant to Perplexity because citation UX needs source independence, not just visible links.
- April 16: Perplexity Personal Computer ships for Mac, a desktop-native research agent that works across your local files and browser tabs alongside web search.
- March 18: Comet browser paywall dropped. Free across iOS, Android, Mac, and Windows.
Max adds Model Council and Perplexity Computer. The Comet browser is free on every major platform.
System Verdict
Pick Perplexity if you need cited, verifiable answers from the live web. Nothing else in the consumer-AI market surfaces sources as aggressively or as cleanly. Pro at $20/mo or $200/year is the right default for researchers, journalists, analysts, and anyone doing due diligence who needs to see where a claim came from.
Skip it if your workload is writing, coding, or creative generation. Claude Opus 4.7 wins on long-form coherence and terminal coding; ChatGPT wins on feature breadth, image generation, and the Codex agent. Perplexity has no native image or video generation and does not pretend to be a general-purpose assistant.
Who pays which tier: Free for casual fact-checks, Pro $20/mo or $200/year for regular cited research (the sweet spot), Enterprise Pro $40/seat/mo or $400/year for teams that need SSO, shared files, and admin controls, Enterprise Max $325/seat/mo or $3,250/year only for heavy reasoning, multi-model research, larger files, and high-volume Computer workflows.
Key Facts
| House model | Perplexity-native cited answer engine; Sonar / Sonar Pro on the API |
| Switchable models (Pro+) | Recent GPT, Claude, Gemini, and other supported models |
| Deep Research | Reads hundreds of sources, 2-5 min per query; exports to PowerPoint, spreadsheets, dashboards, and websites |
| Model Council (Max-only) | Runs the same query across 3 frontier models; synthesizer (Claude Opus) resolves convergence/divergence |
| Perplexity Computer | Agentic work surface for paying subscribers; Pro shows 500 credits/mo and Enterprise Max shows 15,000 credits/mo on the May 2026 pricing page |
| Spaces | File-grounded research workspaces. Pro: 50 files. Enterprise: 500 files. Scheduled Tasks for recurring research |
| Comet browser | Free on iOS (Mar 2026), Android (Nov 2025), Mac and Windows (Jul 2025). No iPad build |
| Voice Mode | Pro and Max |
| Pricing | Free · Pro $20/mo or $200/year · Enterprise Pro $40/seat/mo or $400/year · Enterprise Max $325/seat/mo or $3,250/year |
| API | (search-augmented LLM, 2× citation density on Pro) |
What it actually is
One product surface (answer engine plus a thin agent layer) with consumer and enterprise pricing. The free tier is for light answer-engine use. Pro adds model selection, more Pro searches, Deep Research, Spaces, and Computer credits. Enterprise tiers add team governance, data controls, and higher usage ceilings.
The moats are specific and narrow. First, citation UX: every numbered claim links to the underlying source inline, which neither ChatGPT nor Claude nor Gemini matches in default behavior.
Second, the Sonar+frontier-model architecture. Perplexity routes default search answers through Sonar and lets paid users pick among supported frontier models when the question calls for a different reasoning style. That’s a neutrality play no single-vendor chatbot can replicate.
Third, Deep Research’s direct export to decks and spreadsheets eliminates the copy-paste step that kills most research-to-deliverable workflows. Fourth, Comet is the only AI-native browser available free on every major platform.
The open question is whether any of these survive once Google AI Overviews, ChatGPT browsing, and Gemini Deep Research close the gap.
When to pick Perplexity
- You need verifiable answers with live citations. Every sentence is traceable to a source. That’s the exact behavior default ChatGPT and Claude chats do not produce without explicit browsing mode prompts.
- You want to A/B frontier models on the same question. Pro’s model switcher and higher enterprise tiers’ multi-model research mode are a clean way to compare recent GPT, Claude, Gemini, and other supported models against similar research prompts without stacking separate subscriptions.
- You run research-to-deliverable workflows. Deep Research exports directly to PowerPoint, spreadsheets, and dashboards. That’s the step that kills throughput when researchers use a chatbot and then paste into slides manually.
- You want a browser-layer AI assistant. Comet is free on iOS, Android, Mac, and Windows; it reads the page, summarizes tabs, and answers in context without leaving the browser.
- You’re building search-augmented LLM apps. The Sonar and Sonar Pro APIs ship real-time web-wide research with citations as a primitive. Cheaper than stitching your own Bing/Brave + RAG stack.
When to pick something else
- Long-form writing, coding, or nuanced reasoning: Claude Opus 4.7. Better coherence on 10K+ word outputs. Claude Code CLI is the strongest terminal coding agent.
- One tool for everything (text + images + voice + agents): ChatGPT. Broader feature surface, GPT Store, Codex, Operator agent.
- Neural web search for technical and research-heavy queries: Exa. Embeddings-based; best when you want the raw ranked document list, not a synthesized answer.
- Ad-free traditional search with privacy controls: Kagi. Keeps the classic ranked-links UI and lets you re-rank or block domains.
- Community-grounded answers with agent workflows: You.com. Thinner citations but broader consumer feature set.
- Peer-reviewed scientific literature: Consensus or Elicit. Both index and score published papers. Perplexity covers the open web but cannot enter most paywalled journals.
Pricing
Consumer pricing via Perplexity pricing:
| Plan | Price | Pro Searches | Models | Deep Research | Who’s it for |
|---|---|---|---|---|---|
| Free | $0 | ~3-5/day | Sonar only | Limited | Casual fact-checks |
| Pro | $20/mo or $200/year | Up to 200 Pro queries/week | Recent GPT, Claude, Gemini, and other supported models | Up to 20/month | Most individuals should land here |
Enterprise pricing via perplexity.ai/enterprise/pricing:
| Plan | Price | Fit |
|---|---|---|
| Enterprise Pro | $40/seat/mo or $400/year | Team search, files/apps, SSO/SCIM, admin controls |
| Enterprise Max | $325/seat/mo or $3,250/year | Enterprise Pro plus advanced reasoning, deep research at scale, larger files, multi-model research, retention configuration, audit logs, and team insights |
Sonar and Sonar Pro APIs are priced separately at perplexity.ai/api-platform. Verify current per-token rates there before building on it.
Prices verified 2026-05-13 via Perplexity pricing and the Perplexity API pricing docs.
Against the alternatives
| Perplexity Pro | ChatGPT Plus | Claude Pro | |
|---|---|---|---|
| Source citation quality | Strongest · inline numbered citations by default | Good in browsing mode, absent elsewhere | Present when browsing is invoked, less granular |
| Research depth | Deep Research reads 100+ pages, exports to deck/sheet | Browsing + Deep Research mode, no direct export | Projects + web search, no direct deliverable export |
| Pricing | $20/mo or $200/year | $20/mo | $20/mo |
| Best viewed as | Cited answer engine and research specialist | Generalist default with broadest feature suite | Reasoning, writing, and coding specialist |
Failure modes
- Citations don’t guarantee correctness. Perplexity occasionally misrepresents what a linked source actually says. The citation looks authoritative but the underlying claim is a paraphrase drift or outright hallucination. Spot-check before quoting.
- Source quality varies by query. For thin-coverage topics the engine cites low-tier SEO content and blog roundups with the same confidence as primary sources. Depth is only as good as the open web on the topic.
- Paywalled content is mostly inaccessible. Academic journals, gated news, and subscription research do not load. For scientific literature use Consensus or Elicit instead.
- Model switching adds a decision tax. Pro users with six models on tap tend to ping-pong. The default (Sonar auto-routing) is the right choice for most queries; override only when you have a specific reason.
- Perplexity Computer credits deplete quietly. Pro shows 500 credits/mo and Enterprise Max shows 15,000 credits/mo on the May 2026 pricing page. Multi-step agent runs can burn credits fast, so watch the live credit tracker before kicking off a long deploy or scraping job.
- Comet is not on iPad. Mac, Windows, iPhone, and Android ship. iPad users are stuck with the web app or the regular Perplexity iOS app.
- Model Council is web-only on Max. Not available on mobile apps, Enterprise Pro, or the free tier. The synthesizer step adds latency. Expect 30-60s before a unified answer.
- No native image, video, or voice generation beyond Voice Mode input. For multimodal creation you need a second tool.
Methodology
This page was produced by the aipedia.wiki editorial pipeline, an automated system that ingests vendor documentation, verifies pricing and model details against primary sources, and generates the editorial analysis you are reading. No individual human wrote this review. Scoring follows the four-dimension rubric at /about/scoring/ (Utility × Value × Moat × Longevity, unweighted average). Last verified 2026-05-13 against Perplexity Pro, Perplexity Enterprise pricing, the Perplexity API pricing docs, the Sonar model docs, the API privacy/security docs, the May 8, 2026 Morningstar/PitchBook integration announcement, and the May 6, 2026 Snap-Perplexity deal coverage.
FAQ
Is Perplexity free? Yes. The free tier has limited advanced usage and full access to standard cited answers. Pro at $20/mo or $200/year unlocks higher limits, the model switcher, Deep Research with deck and spreadsheet export, Spaces, and Voice Mode.
Which frontier models can Pro users pick per query? Perplexity currently advertises model choice across recent GPT, Claude, Gemini, and other supported models. Exact model availability changes quickly, so check the model picker before buying for one specific model.
What is Sonar? Sonar is Perplexity’s search-grounded answer model family. For developers, the Sonar and Sonar Pro APIs expose search-grounded models with published 128K and 200K context windows.
It is also available to developers as the Sonar and Sonar Pro APIs. Pricing for the API tier is listed at perplexity.ai/api-platform.
What does Deep Research do today? Deep Research runs multi-step research (typically 100+ sources over 2-5 minutes). As of the February 2026 upgrade it exports the result directly to PowerPoint, spreadsheets, dashboards, or a generated website.
The exact model path can vary by plan and selected model, so verify the in-product model setting when the engine matters.
Is Comet browser Mac-only? No. Comet shipped on Mac and Windows in July 2025, Android in November 2025, and iPhone on March 18, 2026. It is free across every platform. There is no iPad build.
What are Spaces? Spaces are file-grounded research workspaces, the renamed and expanded version of the old Collections feature. Queries inside a Space search both uploaded files and the open web.
Pro users get up to 50 files per Space, Enterprise up to 500. Scheduled Tasks let a Space run recurring research automatically.
What is Model Council and who gets it? Model Council runs the same query against multiple supported frontier models and compares their answers. A synthesizer then resolves agreements and disagreements into a unified answer.
It is Max-only (and Enterprise Max), web-only, not on mobile.
What is Perplexity Computer? An agentic work surface for paying subscribers that can use tools, files, connectors, and web context for multi-step tasks: research pipelines, analysis, drafting, and automation. The May 2026 pricing page lists 500 Computer credits per month on Pro and 15,000 credits per month on Enterprise Max; live credit tracking shows consumption as runs progress.
Related
- Category: AI Search · AI Chatbots · AI Writing
