Perplexity
Perplexity holds place 2 of 14 in our AI chatbots and assistants ranking. The catalogue describes it as: answer engine that puts inline citations on every answer.
Sebastiaan SmitsFounderOn this site vendors can buy visibility, and we say so. Outbound links, including the one to Perplexity, are marked identically. Read more.
Visit Perplexity See it in the AI chatbots and assistants ranking
Perplexity at a glance
| Headquarters | United States |
|---|---|
| Deployment | Left open by the sources we used |
| Pricing model | Free, Pro and Max plans published; Enterprise priced separately |
| Free plan | Yes, per our records |
| API | Perplexity mentions an API in the material we reviewed |
| Best for | Researchers who want every answer shown with its sources and links |
What is Perplexity?
Perplexity is an answer engine from Perplexity AI, a company that also sells a browser and an API platform. Instead of returning a list of links, it searches the web, reads the results and writes an answer with citations. The vendor's developer documentation describes five API products: a Search API that returns ranked web results with extracted page text, an Agent API that returns cited answers or calls third-party models, an Embeddings API, a Decisions API that returns typed answers with probabilities, and an Orchestrator API in private preview.
Its consumer site sits behind a bot check that blocked our fetch, so this profile leans on the developer docs and trust centre and does not restate consumer plan details. Buyers are researchers and developers who want sources shown beside every answer.
Perplexity key features
- Agent API. Returns web-grounded answers with built-in citations in one call, or routes to models from OpenAI, Anthropic, Google, xAI, Z.AI, Moonshot AI and NVIDIA, with web search tools and effort levels.
- Search API. Delivers ranked web results as JSON with extracted page text, for your own model to read. The vendor states that it produces no generated answer and is billed per request.
- Embeddings API. Creates vectors for semantic search and retrieval pipelines, in standard and contextualized variants, each with its own pricing section in the documentation.
- Decisions API. Answers yes or no, multiple choice and rubric questions about text and images with probabilities attached, with an option for web grounding.
- MCP servers and CLI. A remote MCP server at api.perplexity.ai/mcp works with clients that support remote servers, signing in with a Perplexity account or API key. A local open-source server runs over stdio, and a Perplexity CLI exists.
- Tool calls in the Agent API. Documented tools include web search, fast search, image search, URL fetching, people search, finance search and a sandbox container for running code during a request.
- Structured output. The API FAQ says JSON output can be constrained with the text.format field, which matters when answers feed into code. The documentation also lists OpenAI compatibility.
Who Perplexity suits
- A market analyst checking claims. Ask a question, read the answer and open the cited pages to confirm each point. The value is the visible source trail, not a promise that the answer is correct, and the vendor warns that responses can contain mistakes.
- A developer adding web search to a chatbot. Call the Search API for ranked results and page text, then pass them to the model already in use. Billing per request lets the team estimate cost from expected query volume.
- An engineering team building a research agent. Use the Agent API with a chosen effort level, web search and fetch tools, pin an explicit model ID, and keep a fallback model. The docs recommend pinning IDs because retirements are announced in the changelog.
Getting started with Perplexity
For the API, the docs say to create an account on the developer console, generate an API key and make a first call from the quickstart, which the vendor describes as taking minutes. Billing, projects and key management have their own admin pages, and rate limits depend on usage tiers.
SDKs and an OpenAI-compatible interface exist for migration. An MCP server can be added to an assistant by URL. The consumer apps are reached through perplexity.ai, but that site blocked our fetch, so sign-up steps and trial terms for them are not covered here.
What we make of Perplexity
An answer engine built around inline citations, from a San Francisco company that also sells a browser and an API. The Free plan covers limited daily use, and Pro and Max raise limits and add research and premium data sources.
It routes each query across models from other labs, so you choose the answer style, not the model, unless you pay for a Pro or Max selector.
In our AI chatbots and assistants ranking Perplexity stands at #2 of 14: it follows Mistral Le Chat and comes before Claude. The first three places are reserved for challengers where a category has them, which is why we publish no numeric score.
Home market: the United States, one of 6 North America bases among the 14 products here.
Strengths and limits of Perplexity
- An answer engine built around inline citations, from a San Francisco company that also sells a browser and an API.
- Catalogue tag: Model routing.
- Suits: Researchers who want every answer shown with its sources and links.
- Cited answers, one of the three traits the catalogue lists for Perplexity.
- Comet browser, one of the three traits the catalogue lists for Perplexity.
- Enterprise prices are not shown, so organisations must go through sales to compare.
- Poor match for: Teams wanting one fixed model behind every answer.
- Pro adds model selection, so Free users cannot choose which lab's model answers.
Perplexity pros and cons in practice
- Citations are part of the Agent and Search APIs, so developers can show readers the source of each statement.
- API tool prices are listed per call, which makes costs of search, fetch and sandbox use visible before a build.
- The docs state that customer content is not used for training by default, which helps with internal review.
- Remote MCP and an OpenAI-compatible interface cut the work of adding Perplexity to existing assistants.
- Zero Data Retention needs a written agreement and is unavailable on every model, so strict retention rules need extra paperwork.
- The Orchestrator API is still in private preview, and the docs show the surface changes, so integration may need rework.
- Third-party models carry the provider's own terms, and some providers require limited retention.
- Consumer pricing is hard to verify from outside because the main site blocked our automated fetch.
Setup, support and exit with Perplexity
Pro adds Computer access, deep research and premium data sources such as PitchBook and Statista.
The company states that every answer carries inline citations.
Perplexity AI, Inc.
gives a San Francisco address, and its privacy notice lets users ask for their data in a portable, machine-readable form.
Paying for Perplexity
Pricing as recorded: Free, Pro and Max plans published; Enterprise priced separately.
Consumer plan prices could not be read because perplexity.ai returned a bot check, so we do not state them. From the API pricing page: Agent API model usage is billed per token at each provider's published rates. Tool fees are $0.0025 per standard web search, $0.001 per fast search, $0.0025 per image search, $0.0005 per URL fetch, $0.005 per people or finance search and $0.03 per sandbox session. The Search API starts at $5.00 per 1,000 requests. Orchestrator usage is billed per token with no per-request fee.
Basis: perplexity.ai, read for the Best AI Chatbots and Assistants in 2026 guide (reviewed 6 November 2026).
Perplexity: data and security
The API privacy page says Perplexity does not use customer content to train generative AI models under the API terms. Zero Data Retention is not the default and is offered to eligible customers under a written agreement.
The vendor lists a SOC 2 Type II report, a 2025 HIPAA gap assessment and a CAIQlite questionnaire, and points to a trust centre for subprocessors. Hosting region is not stated on the pages we read.
Perplexity in numbers and labels
| Product | Perplexity |
|---|---|
| Vendor website | perplexity.ai |
| Primary guide | Best AI Chatbots and Assistants in 2026 |
| Price visibility | Public on the vendor site |
| Headquarters | United States (North America) |
| What we do not hold | headcount, funding and customer count |
| Head-to-head pages | 2 |
| Rank there | #2 of 14 |
| Alternatives page | Perplexity alternatives |
| Catalogue tags | Cited answers, Comet browser, Model routing |
Perplexity integrations and ecosystem
Integrations the vendor names (10):
- OpenAI-compatible API
- Perplexity API MCP server
- Perplexity Computer MCP server
- Perplexity CLI
- AWS Marketplace
- Perplexity SDK
- OpenAI models
- Anthropic models
- Google models
- xAI models
Also ranked in
| Guide | Position |
|---|---|
| Best AI Chatbots and Assistants in 2026 | #2 of 14 |
Side-by-side pages with Perplexity
- Perplexity vs Claude: choose Perplexity if you mainly look things up and want every answer shown with inline citations, and you are happy to let the service pick the model unless you pay for a selector.
- Perplexity vs Google Gemini: pick Perplexity when you are a researcher who wants every answer shown with its sources and does not mind that a model from another lab may write it.
Alternatives to Perplexity
These are the nearest neighbours of Perplexity in the same list, picked on shared traits and price model.
A full write-up exists: Perplexity alternatives, with reasons buyers leave and what moving involves.
- Mistral Le Chat (#1 of 14): Mistral's pricing page lists Free, Pro, Team at 24.99 USD per user per month, and Enterprise. Team includes up to 30GB storage per user, data export and domain verification. Enterprise can be deployed in a private cloud or on-premises. It is a model vendor's own assistant, not a search engine built around citations. Switch if you want a European vendor with team data export and an on-premises route.
- Claude (#3 of 14): Anthropic's own assistant. Web search is available on the Free plan, Pro is 17 USD a month on annual billing (20 USD monthly), and Max starts at 100 USD. Free allows up to 5 Projects. Research is a Pro feature. It answers from one model family, where Perplexity lets you switch between several. Switch if you want one vendor's models with Projects, Artifacts and file creation. (United States)
- Microsoft Copilot (#10 of 14): Microsoft prices Copilot as part of its individual plans: Personal 9.99 USD, Family 12.99 USD, Premium 19.99 USD and Pro 99.99 USD a month on the page we read. Copilot works in Word, Excel, PowerPoint, Outlook and Teams. On Family, AI features are for the subscription owner only. Switch if your work is in Microsoft 365 and you want the assistant in those apps.
- You.com (#12 of 14): The pricing page we read is for APIs: Web Search at 5 USD per 1,000 calls, Answer at 5 USD per 1,000 calls with sources, and a free MCP tier of 100 queries a day. A Chat Platform sign-in exists, but we found no chat plan prices. It is built for developers. Switch if you want cited web answers inside your own product, billed per call.
- Strawberry (#13 of 14): Not for: developers, or anyone who wants a single chat assistant with mobile apps. A browser with built-in AI companions for non-technical work.
When Perplexity is still the right call: Perplexity remains the better fit when your main job is asking web questions and reading answers with linked sources, and when you value switching between models from several providers in one subscription. Its Enterprise tiers add SCIM, audit logs and configurable data retention for organisations, and its agreements with third-party model providers prohibit training on Perplexity data. If an alternative cannot show sources per answer in a way your reviewers accept, staying costs less than rebuilding the habit.
Perplexity: questions
9 answersWho should choose Perplexity?
Researchers who want every answer shown with its sources and links. Think twice if you are in this group: teams wanting one fixed model behind every answer.
What are the main alternatives to Perplexity?
The closest products in the same guide are Mistral Le Chat, Claude, Microsoft Copilot, You.com and Strawberry. The full write-up is on the Perplexity alternatives page.
Which country is Perplexity from?
Our records put Perplexity in the United States, a North America vendor. Country here means the contracting company, which can differ from where servers run.
What does Perplexity cost?
The pricing entry for Perplexity reads: Free, Pro and Max plans published; Enterprise priced separately. On free access, a free plan is on record.
Where does Perplexity stand among AI chatbots and assistants?
In our AI chatbots and assistants list Perplexity stands at place 2 of 14.
What is the difference between the Search API and the Agent API?
The Search API returns ranked web results as JSON with extracted page text and no generated answer, billed per request. The Agent API returns a cited answer, or calls third-party models with tools. Pick Search if your own model writes the answer.
Does Perplexity train on API data?
The vendor says no: under the API terms it does not use customer content to train generative AI models. Retention is a separate question. Zero Data Retention is not the default and needs eligible-customer status under a written agreement, and not every model supports it.
Can I connect Perplexity to my own assistant?
Yes. The docs describe a hosted MCP server at api.perplexity.ai/mcp for clients that support remote servers, using sign-in or an API key. There is also an open-source local server over stdio and a command-line tool.
What does the Perplexity API cost?
It depends on what you call. The pricing page lists $0.0025 per standard web search in the Agent API, $0.0005 per URL fetch and $5.00 per 1,000 Search API requests, plus token charges at each model's rates. Consumer subscription prices are not covered here.
Perplexity: sources
Read from the vendor’s own pages on 5 November 2026.
- docs.perplexity.ai: API product overview
- docs.perplexity.ai/docs/getting-started/pricing: Tool and request pricing, billing basis
- docs.perplexity.ai/docs/resources/privacy-security: Training, ZDR, SOC 2 Type II, HIPAA gap assessment, CAIQlite
- docs.perplexity.ai/docs/getting-started/integrations/mcp-server: Remote and local MCP server
- trust.perplexity.ai: Trust centre existence
How we rank
We test for four things rather than for features: setup time, price in practice, exit and audience. The result for Perplexity: #2, out of 14 ranked. The editorial process sets out how facts are gathered and how commercial links are handled.