Join our Newsletter — 33% off our NHI Course
Home› Guides› Perplexity AI Explained: Plans, Sources, Comet and Enterprise…
Guide AI Platforms

Perplexity AI Explained: Plans, Sources, Comet and Enterprise Controls

← All guides
By NHI Mgmt Group, edited by Lalit Choda Updated 15 November 2026 14 min read
On this page

Perplexity answers questions with a written summary and numbered citations instead of a page of links, and its plans, models and Comet browser now shape how many people and organisations search the web. Perplexity is an AI-powered answer engine from Perplexity AI that searches the web in real time and returns a conversational answer with numbered citations linking to the original sources. This page sets out how Perplexity retrieves and cites sources, which models its plans include, what each plan costs and allows, what the Comet browser does, what publishers can control, and which enterprise and privacy controls exist. It is written for publishers, marketers, IT and privacy teams and everyday users who want facts from Perplexity's own help centre and blog rather than commentary.

Last checked: 2026-11-15.

This is an independent reference page from NHI Mgmt Group based on public provider documentation; it is not sponsored or endorsed by any provider.

Key takeaways

  • Answers come with numbered citations. Perplexity's help centre describes four steps: interpret the question, search the web in real time, summarise the findings and attach numbered citations to the sources.
  • Six consumer and business plans. Standard (Free), Pro, Education Pro, Max, Enterprise Pro and Enterprise Max, plus a separate pay-as-you-go API Platform with its own billing.
  • Many models, one automatic default. The "Best" option picks a model per query; as of 17 October 2026 paid plans could also choose models from Anthropic, Google, Moonshot AI, NVIDIA, OpenAI, xAI and Z.ai.
  • Comet is a Chromium-based browser. It launched on 9 July 2025 for Max subscribers and became a free download worldwide on 2 October 2025; Comet Assistant queries are not included on the Free plan.
  • Two crawlers, two rule sets. PerplexityBot follows robots.txt and is not used to crawl content for foundation models; Perplexity-User fetches pages on a user's request and generally ignores robots.txt.

What is Perplexity?

Perplexity's help centre calls it "an AI-powered search engine that transforms how you discover and interact with information". A user types a question, Perplexity searches the web, and the reply is a written answer with numbered source links, followed by the option to ask follow-up questions. It differs from a classic search results page in three ways: the answer is composed by a language model, sources are attached to the answer as citations, and the user can choose a deeper mode (Pro Search or Research) when a question needs more work.

FieldValue (as of 2026-11-15)
ProviderPerplexity AI
Product typeAI answer engine with numbered citations; Comet AI browser; API Platform
StatusActive
Default model"Best", an automatic option that selects a suitable model for each query; no single named default
Search modesStandard search, Pro Search, Research (also called Deep Research)
PlansStandard (Free), Pro, Education Pro, Max, Enterprise Pro, Enterprise Max; API Platform billed separately
Entry business priceEnterprise Pro from $40 per seat per month or $400 per seat per year
BrowserComet, Chromium-based, free download since 2 October 2025
CrawlersPerplexityBot (search indexing) and Perplexity-User (user-requested fetches)
Training on enterprise dataPerplexity states Enterprise and API data are never used for training

How do you access Perplexity?

Perplexity runs on its website and apps. Every query is a search; users can choose Pro Search for more thorough answers across more sources, or Research for a longer report. Comet, Perplexity's browser, is a further way in.

What is Comet?

Comet is a Chromium-based browser with Perplexity's AI built in. Perplexity's help centre lists standard browser features (bookmarks, page translation, support for most Chrome extensions) alongside AI features.

  • Search: Perplexity is the primary search engine.
  • Personal Search with Comet Intelligence: analyses browsing history to retrieve relevant information.
  • Browser Commands: natural language commands for tab management.
  • Gmail connector: email and calendar questions.
  • Ask and Summarize buttons: context on, or a summary of, the open page.
  • Built-in AdBlock.
  • Comet Assistant: the July 2025 launch post said the assistant can run entire browsing sessions and handle tasks such as booking meetings and sending emails.

On installation Comet requests keychain access to store website passwords, payment details for autofill, tokens for connected services and encrypted Comet Intelligence data; the help centre says this information is stored locally on the device with encryption. Extensions must be installed fresh in Comet and reconfigured.

DateComet milestone
9 July 2025Launched ("Browse at the speed of thought") for Perplexity Max subscribers, with invite-only access for the waitlist over the summer
2 October 2025Free download for everyone at perplexity.ai/comet; desktop only at the time, with a mobile app announced as planned
As of 2026-11-15Comet for Enterprise available for macOS and Windows; Comet Assistant queries on paid plans only

Security teams assessing AI browsers will find deployment risks and controls in the Agentic Browsers Security Guide.

How does Perplexity find and cite sources?

The help centre article "How does Perplexity work?" describes the same four steps for every answer.

  1. Interpret: language models work out what the user is asking, which Perplexity says "goes beyond simple keyword searches".
  2. Search: Perplexity searches the internet in real time and gathers material from articles, websites and journals.
  3. Summarise: the most relevant findings are combined into one conversational answer.
  4. Cite: each answer includes numbered citations linking to the original sources, so the reader can check a claim or read further.

The depth of the search depends on the mode the user picks.

ModeWhat Perplexity says it doesWho can use it
Standard searchQuick answer with citations using the automatic "Best" model choiceAll plans
Pro SearchMore detailed answers, with the option to choose the AI model3 per day on Free; weekly limits on paid plans
Research (Deep Research)Performs dozens of searches automatically, reads hundreds of sources and produces a report in about 2 to 4 minutes1 per month on Free; monthly limits on paid plans
Model CouncilRuns a question across several modelsMax and Enterprise Max, on web

As of 2026-11-15. Perplexity's help centre does not explain how sources are ranked or selected for an answer. Readers should open the numbered sources for anything important, since a summary can misread or drop context from the page it cites.

Which models power Perplexity?

Perplexity's help article on advanced models lists the Search models available by plan as of 17 October 2026 and says the in-product model selector is the source of truth. The Free plan uses "Best" and a limited number of Pro Searches but cannot choose an advanced model manually.

Model (provider, alphabetical by provider)ProMaxEnterprise ProEnterprise Max
Claude Fable 5.1 (Anthropic)NoYesYesYes
Claude Opus 5.5 (Anthropic)NoYesYesYes
Claude Sonnet 5.5 (Anthropic)YesYesYesYes
Gemini 3.8 Flash (Google)YesYesYesYes
Gemini 4 Argon (Google)NoYesYesYes
Kimi K3 (Moonshot AI)YesYesYesYes
Nemotron 3.5 Super VL (NVIDIA)YesYesYesYes
GPT-6 Astra (OpenAI)NoYesYesYes
GPT-6.1 Sol (OpenAI)YesYesYesYes
Grok 4.7 (xAI)YesYesYesYes
GLM 5.3 (Z.ai)YesYesYesYes

As of 17 October 2026, per Perplexity's help centre. Further notes from the same article:

  • Usage limits: manual model selection on paid plans is subject to usage limits, published on the plan comparison page.
  • Enterprise restrictions: administrators can restrict providers or individual models, and regional or data retention settings can shorten the list.
  • Thinking: a setting that some models support, not a separate model.
  • Comet and Computer: each has its own model selection; Computer tasks use credits.
  • API: the API has a separate model catalogue and billing and is not included in any subscription.

How much does Perplexity cost?

Perplexity's plan comparison article publishes prices for some plans and fixed limits for the Free and Enterprise plans. For Pro, Education Pro and Max it describes limits qualitatively ("average use" or "advanced use") rather than as numbers.

PlanPrice stated by PerplexityPro SearchesResearch queriesComet Assistant queriesFile uploads
Standard (Free)Free3 per day1 per monthNot includedLimited
ProNot stated on the plan articleWeekly limit, average useMonthly limit, average useMonthly limit, average useWeekly limit, average use
Education Pro$10 per month with SheerID verificationWeekly limit, average useMonthly limit, average useMonthly limit, average useWeekly limit, average use
MaxNot stated on the plan article; annual billing only on perplexity.aiWeekly limit, advanced useMonthly limit, advanced useMonthly limit, advanced useWeekly limit, advanced use
Enterprise ProFrom $40 per seat per month or $400 per seat per year400 per week50 per month80 per month100 sessions per week
Enterprise MaxNot stated on the plan article4,000 per week500 per month800 per month1,000 sessions per week
API PlatformPay-as-you-go creditsSeparate product: no access to the web app's advanced models or Pro features, no complimentary credits

As of 2026-11-15, prices in US dollars. Other plan details from the same article:

  • File and app creation: not on Free; 50 per month on Enterprise Pro and 500 per month on Enterprise Max.
  • Image and video generation: not on Free; Enterprise Max includes 15 videos per month, each 8 seconds with audio in 16:9.
  • Projects: Pro allows up to 50 file uploads per project; Enterprise Max's personal repository (My Files) holds up to 10,000 files.
  • Education Pro extras: includes Perplexity Computer and Learn Mode.
  • Support: basic help centre support on Free, priority support on Pro and Education Pro, dedicated support on Enterprise plans.
  • Retirement: the Organization File Repository was retired on 1 October 2026, and Perplexity said files left after that date would be deleted.

How do ads work in Perplexity?

Perplexity's July 2024 Publishers Program post described the advertising model: brands can pay to place specific related follow-up questions in the answer engine interface and on Pages, with ads to arrive through the related questions feature "in the coming months". The plan comparison article does not list ads as a feature or limit of any plan. Market-by-market rollouts across engines are tracked in the Ads in AI Answers Tracker.

What can publishers control?

Perplexity documents two user agents with different rules.

User agentPurpose stated by Perplexityrobots.txtIP list
PerplexityBotSurfaces and links websites in Perplexity search results; not used to crawl content for AI foundation modelsRespected; Perplexity recommends allowing PerplexityBot to appear in resultsperplexity.ai/perplexitybot.json
Perplexity-UserVisits a page when a user asks a question, to help answer and link to it; not used for crawling or collecting training content"Generally ignores robots.txt rules", because a user requested the fetchperplexity.ai/perplexity-user.json
  • Crawl rules: a robots.txt rule for PerplexityBot controls whether a site is indexed for Perplexity search results.
  • User fetches: robots.txt does not reliably stop Perplexity-User; publishers who want to block it need network-level controls, using the published IP list to identify the traffic.
  • Verification: both IP lists let site operators check that traffic claiming to be Perplexity really comes from Perplexity.
  • Publishers Program: announced on 30 July 2024 with launch partners TIME, Der Spiegel, Fortune, Entrepreneur, The Texas Tribune and WordPress.com. It promised a revenue share when Perplexity earns advertising revenue from an interaction that references a partner's content, free access to Perplexity's Online LLM APIs, and one year of Enterprise Pro for partner employees. Comet Plus named its first participating publishers on 2 October 2025.

Rules for every major engine's crawlers sit side by side in the AI Crawler and Grounding Controls Tracker. For what optimisation for answer engines means, see What Is Answer Engine Optimisation (AEO)? Definition, Signals and Measurement.

What privacy controls exist?

AreaConsumer plansEnterprise plans
TrainingPro, Education Pro and Max users can opt out of data collectionEnterprise data "is never used to train or fine-tune Perplexity's models"; the API is also never used for training
Third-party model providersPerplexity says its agreements with providers such as Anthropic and OpenAI prohibit using Perplexity data to train their models
RetentionPersonal information kept while the account is active; removed within 30 days of account deletionUploaded files deleted after seven days; custom retention periods for organisations with 50 or more seats or at least one Enterprise Max seat; Enforce Incognito Mode under Data and privacy in Organization Settings
IdentityPersonal accountSSO through the organisation's identity provider; SCIM on Enterprise Max
OversightNot applicableAudit Logs, configurable Data Retention and an Insights dashboard on Enterprise Max; audit logs for Comet with 50 or more Enterprise Pro seats or one Enterprise Max seat
AssurancePerplexity states it is SOC 2 Type II certified

As of 2026-11-15. Comet for Enterprise adds deployment and policy controls:

  • Deployment: MDM deployment for macOS and Windows, silent and offline installation, and a pilot option ("Enable for specific members" under Enable Comet Access in Organization Settings).
  • Policy: more than 500 Chromium-based browser policies, plus Agent Permission Controls to restrict websites and require approval for assistant actions.
  • Data: Perplexity states enterprise Comet data is not logged or used for training, third-party AI partners contractually cannot retain data sent through Comet, and corporate DNS blocking rules still apply.
  • Telemetry: admins can review Comet usage and performance data.

Admin settings, connector risks and Comet hardening are covered step by step in the Perplexity Enterprise Security guide.

When was Perplexity released, and how has it changed?

DateMilestone
30 July 2024Publishers Program announced: ad revenue share through sponsored related questions, free Online LLM API access, Enterprise Pro for partner staff
9 July 2025Comet browser launched for Max subscribers
2 October 2025Comet made free worldwide; first Comet Plus publishers named
1 October 2026Organization File Repository retired
17 October 2026Model-by-plan list updated in the help centre

As of 2026-11-15.

What are Perplexity's limitations?

  • Summaries can be wrong: numbered citations make checking easier, but the summary is still model-written. Open the source for legal, medical, financial or security decisions.
  • Source selection is opaque: the help centre does not explain how sources are ranked or chosen.
  • Limits vary by plan: Pro and Max limits are described qualitatively, so heavy users should check the in-product counter rather than assume a figure.
  • Model availability moves: the model list is dated and the selector is the source of truth; enterprise admins and regional settings can shorten it.
  • User fetches bypass robots.txt: publishers cannot rely on robots.txt alone to stop Perplexity-User.

Common questions about the service

Is Perplexity free?

Yes, the Standard plan is free, with 3 Pro Searches a day, 1 Research query a month and limited file uploads. Choosing advanced models manually, image generation, file and app creation, and Comet Assistant queries need a paid plan.

Is Perplexity data used to train AI models?

Perplexity states that Enterprise and API data are never used for training, that Pro, Education Pro and Max users can opt out of data collection, and that its agreements with third-party model providers prohibit them from training on Perplexity data.

Where identity and security fit

Comet stores passwords, payment details and tokens for connected services such as Gmail, and its assistant can act inside signed-in sessions, so the browser becomes a holder of user credentials and delegated access. Organisations should decide which connectors and assistant actions are allowed, enforce SSO and, on Enterprise Max, SCIM so leavers lose access promptly. Staff using personal accounts for work should follow the Safe Use of AI Chatbots Guide.

Practitioner checklist

  • Decide whether PerplexityBot should index your site and set robots.txt to match.
  • Use the published Perplexity-User IP list if you need to identify or limit user-requested fetches, since robots.txt does not reliably apply.
  • For work use, move staff from personal plans to an Enterprise plan with SSO, and turn on SCIM and Audit Logs where the plan includes them.
  • Set a data retention period and consider Enforce Incognito Mode under Data and privacy in Organization Settings.
  • Restrict models or providers that do not meet your data residency or policy needs.
  • Pilot Comet with "Enable for specific members", deploy by MDM, and set Agent Permission Controls before wide rollout.
  • Recheck prices, limits and the model list on Perplexity's help centre before decisions; they change frequently.

Standards and references

Related NHI Mgmt Group resources: AI Answer Engines Hub · Perplexity Enterprise Security · Agentic Browsers Security Guide · AI Crawler and Grounding Controls Tracker

Free weekly newsletter

Subscribe to the NHI & AI Identity Journal

The latest on NHI and Agentic AI security – articles, research, breaches, news and events every week.

Bonus 33% off our NHI Course when you subscribe.

NHIMG Editorial Note
Researched and written by the NHI Mgmt Group editorial team using AI tools, under the editorial direction of Lalit Choda. Facts are drawn from the sources listed on this page and checked by our automated quality process. Last updated 15 November 2026. Spotted an error? Use our contact page.
NHI Mgmt Group, the independent authority on Non-Human Identity, IAM and Agentic AI security. nhimg.org