Perplexity AI

User Guide: Search, Deep Research, Computer & Plans

Zoom

What this guide covers

Perplexity AI is an answer engine, not a traditional search engine or static chatbot. It retrieves live sources from the web, synthesizes them using large language models, and returns direct answers with numbered inline citations linking back to original sources. Every claim is grounded in retrieved content.

This guide covers all four major product areas: Perplexity Search (Quick and Pro Search, Focus modes, citations), Deep Research (autonomous multi-step research with long-form reports), Perplexity Computer (agentic workflow executor across 400+ apps), and Plans and Limits so you know what each tier actually includes.

Answer engine Live citations Focus modes Deep Research Perplexity Computer Pro and Max plans Good habits
ⓘ Important: Perplexity is a research accelerator, not a final authority. Always verify citations, especially for numbers, prices, names, and dates, before sharing output externally or basing decisions on it.
Pro and enterprise data protection: Perplexity does not use your prompts or data to train its models on paid plans. Enterprise plans include full data training opt-out by default, plus SOC 2 Type II compliance.

What Is Deep Research?

Deep Research is an autonomous, multi-step research mode that goes far beyond a standard Pro Search. Instead of retrieving sources and synthesizing one answer, it conducts a full research workflow: planning sub-questions, searching across dozens of sources, following threads, cross-referencing findings, and compiling a long-form report with 100+ citations. Most tasks complete in under 3 minutes.

📊 Deep Research vs. Pro Search

Pro Search
Breaks one query into sub-queries, searches 15-20 sources per sub-query, returns a comprehensive answer in seconds. Best for: targeted multi-angle questions on a specific topic.
Deep Research
Runs an autonomous multi-step research workflow across dozens of sources. Follows threads, cross-references, and produces a long-form report with 100+ citations. Best for: comprehensive reports, literature reviews, industry analyses. Uses Claude Opus 4.5 Thinking on Max.

How to Use Deep Research

  1. Open Perplexity and click the research mode selector (the icon left of the search bar).
  2. Select "Deep Research" from the mode options.
  3. Type your research goal in plain language. Be specific: name the topic, define the scope, and state what "good" looks like.
  4. Submit and wait. Deep Research is asynchronous on some plans: you can leave and come back.
  5. When complete, you receive a structured long-form report. Review the 100+ citations for any claim going into a deliverable.
  6. Follow up within the same thread to drill into specific sections or request a different angle.

Best for

  • Comprehensive industry or market analyses
  • Academic literature reviews
  • Competitive landscape reports
  • Policy research and regulatory summaries
  • Investment thesis background research
  • Any task that previously required a research analyst spending 2-4 hours

Limits and Plan Access

PlanDeep Research accessModel used
Free Not available (or very limited) Basic
Pro ($20/mo) 20 per month Advanced
Max ($200/mo) Unlimited Claude Opus 4.5 Thinking
Enterprise Pro 50 per seat per month Advanced
Enterprise Max Unlimited Advanced frontier

Why only 20/month on Pro? In early 2026, Perplexity switched Deep Research to use Claude Opus 4.5 Thinking, a significantly more powerful (and expensive) model. The limit dropped from 500+/day to 20/month on Pro as a result. Each run is now substantially higher quality.

Monthly reset: The Deep Research quota is monthly, not weekly. It does not refill gradually. If you use all 20 on Pro, you wait until the monthly counter resets.

Deep Research Tips

Get more value from your 20 monthly Pro runs

Spend your 20 wisely

  • Reserve Deep Research for genuinely complex tasks. Use Pro Search for targeted single-topic questions. Use Deep Research for multi-angle, multi-source comprehensive reports.
  • Write a specific brief before submitting. Vague prompts return vague reports. Specify the scope, the audience, the format, and any constraints before you send.
  • Use the follow-up thread. After the report arrives, ask follow-up questions in the same thread. This does not consume an additional Deep Research run.
  • Export or save the report immediately. Copy the full report to a document. Thread content is not permanently archived across sessions.

Example prompt structure

›
Market analysis prompt
"Write a comprehensive market analysis of [industry] in 2026. Cover: market size and growth trajectory, top 5 players and their positioning, key trends reshaping the landscape in the last 12 months, main barriers to entry, and opportunities for a new entrant. Prioritize data with citations. Return a structured report with section headers."
›
Literature review prompt
"Conduct a literature review on [topic] using Academic focus. Summarize the current state of research, major findings from the last 5 years, areas of consensus, open questions, and cite all sources. Audience: senior professional with no prior background in this area."

What Is Perplexity Computer?

Perplexity Computer (launched February 2026) is a fundamentally different product from Search. It is not a chatbot, not a search upgrade, and not a writing assistant. It is a workflow executor and AI agent system. You give it a high-level goal; it decomposes the goal into tasks, assigns each task to the best-suited AI model from a pool of 19+ frontier models, executes them across your connected tools and files, and returns a finished deliverable.

⚡ The Core Distinction

Traditional AI tools (ChatGPT, Claude)
Answer questions. You receive a text response and then do something manually with it. The AI stops when the answer ends.
Perplexity Computer
Receives a goal, decomposes it into discrete tasks, assigns tasks to specialized AI models, connects to your tools and data sources, runs sub-agents in parallel or sequentially, and returns a finished deliverable. The execution happens, not just the answer.

💡 Computer is a force multiplier, not a team replacement. Always review outputs, especially for numerical accuracy and client-facing content.

How It Works

  1. Give Computer a high-level goal in plain language. Example: "Build a competitive analysis for our product launch covering five competitors across pricing, features, and positioning."
  2. Computer automatically decomposes the goal into discrete sub-tasks.
  3. Each sub-task is assigned to the most suitable model from the pool of 19+ frontier models (Claude, GPT, Gemini, Grok, and specialized models for images, video, and long-context).
  4. Computer connects to your tools: Gmail, Slack, Notion, Salesforce, Snowflake, GitHub, and 400+ other apps via connectors.
  5. Sub-agents run in parallel or sequentially until the workflow is complete.
  6. You receive a finished deliverable, not just a response. Review it before use.

App integrations

  • 400+ app connectors including Gmail, Slack, Notion, Salesforce, Snowflake, and GitHub
  • 70+ pre-built workflow templates (shareable, schedulable, customizable, asynchronous)
  • When paired with the Comet browser, Computer can operate web-based tools without requiring direct connectors
  • Personal Computer (Mac app) extends agentic capabilities locally to your device and local files

Best Use Cases

  • Market research and competitor analysis
  • Prospect research reports for sales
  • Drafting outreach sequences from CRM data
  • Generating account summaries across tools
  • Daily briefings and summaries from multiple sources
  • Data aggregation across apps and files
  • Multi-section industry reports with charts
  • Recurring reporting workflows (schedulable)
  • Brand strategy outputs
  • Automated sales and marketing workflows
  • Financial analysis: tearsheets, PE diligence memos, public equity research (Computer for Professional Finance, launched May 2026)

What Computer is NOT suitable for

  • Production-critical workflows without human review in the loop
  • Tasks requiring precise numerical accuracy (always verify)
  • Highly specific formatting requirements (inconsistent output)
  • Simple single-step tasks (use Search instead; far fewer credits)

Credit System

Credits are consumed by agentic execution. Scope your tasks to control costs.

Credits by plan

PlanComputer accessCredits
Free Not available None
Pro ($20/mo) Pay-per-use ~$1 = 100 credits
Max ($200/mo) Included 10,000 credits/month
Enterprise Max ($325/seat) Included 15,000 credits/month per seat

Credits reset monthly. When Pro users exhaust their purchased credits, they are prompted to add more. When Max users exhaust their 10,000 included credits, additional credits can be purchased at the pay-per-use rate.

How credits are consumed

  • Open-ended prompts burn fast. "Research this market" can consume 500+ credits.
  • Scoped prompts are efficient. "Research this market, return 5 bullets, cite 3 sources, compare 2 segments, and stop" uses a fraction of the same task open-ended.
  • Long accumulated threads cost more. Context compounds. Split strategy planning from execution into separate threads.
  • Reusable workflow templates are cheapest. A scripted workflow template is cheaper than a fresh agent conversation for recurring tasks.
ROI context: A research analyst costs $2,000-$5,000/month. If Computer replaces or substantially augments even one such workflow, the $200/month Max cost is justified. Break-even is approximately 3+ complex research or content workflows per week.

Plan Comparison

All prices as of mid-2026. Annual billing saves approximately 17% on consumer plans.

Consumer Plans

Feature Free Pro ($20/mo) Max ($200/mo) Education Pro ($10/mo)
Basic (Quick) Search Unlimited Unlimited Unlimited Unlimited
Pro Search ~5/day ~300/day (rolling weekly quota) Unlimited ~300/day
Deep Research Not available 20/month Unlimited 20/day (per research notes)
Deep Research model Basic Advanced Claude Opus 4.5 Thinking Advanced
Model picker No Yes (GPT, Claude, Gemini, Sonar, more) Yes (all Pro + o3-pro, Opus 4 Thinking) Yes
Model Council No No Yes No
Computer access None Pay-per-use credits 10,000 credits/month Pay-per-use credits
Labs / Files & Apps None 50/month Unlimited 50/month
Video generation No 3/month Sora 2 Pro with audio Limited
File uploads/Space Very limited 50 files More 50 files
Comet Plus Buy $5/mo add-on Included Included Included
Sonar API credits None $5/month Included $5/month
Priority support No No Yes No
Annual price $0 $200/year (~$16.67/mo) $2,000/year Monthly only ($10/mo)

Education Pro requires verification via SheerID using a .edu email or institutional ID. The previously available 12-month free student promotion closed in 2025.

Enterprise Plans (per seat)

Feature Enterprise Pro ($40/seat/mo) Enterprise Max ($325/seat/mo)
Pro Searches/week 400 per seat 4,000 per seat
Deep Research/month 50 per seat Unlimited
Computer credits/month Not stated 15,000 per seat
Labs / Files & Apps Limited Unlimited
Comet Assistant tasks/month 80 per seat 800 per seat
File uploads/week 100 per seat 1,000 per seat
Files per Space Shared org pool (500 files) 5,000 per Space
Video generation No 15/month (Veo 3, audio, 8 sec each)
Model Council No Yes
Advanced models (o3-pro, Opus 4.1 Thinking) No Yes
SSO / SCIM provisioning Yes Yes
Audit logging Yes Yes
SOC 2 Type II Yes Yes
Data training opt-out Default (never used) Default (never used)
Role-based access control Yes Yes
Org-wide analytics No Yes
Premium security (1 seat activates org) 50-seat minimum required 1 seat activates org-wide
Email Assistant (upcoming) No Exclusive
Configurable data retention No Yes
Support Dedicated Dedicated
Minimum seats 2 seats (sales at 5+) Contact sales
Annual price per seat $400/seat/year $3,250/seat/year

Volume discounts available for 250+ seats, schools, government, and nonprofits. Education/NPO Enterprise: $30/seat/month. Contact: enterprise@perplexity.ai

How Rate Limits Work

  • Pro Search Uses a rolling weekly quota, not a daily reset. The quota refills gradually: approximately 1 query every 7-8 minutes on a standard 200-query cap. You are not cut off at midnight.
  • Deep Research Quota is monthly. Once you use all 20 on Pro, you wait until the monthly counter resets. No gradual refill.
  • Computer credits On Max, 10,000 credits reset monthly. On Pro, you pay per use with no automatic monthly reset.

What happens when you hit a limit

  • Pro Search limit: Perplexity rolls you back to Quick Search (standard) for the remainder of the quota period. You do not lose access to search; you lose the enhanced Pro Search mode.
  • Deep Research limit (Pro): You cannot run Deep Research until the monthly counter resets. Pro Search and Quick Search remain available.
  • Computer credits (Max): You are prompted to purchase additional credits at the pay-per-use rate.

Promo Code Accounts

In May 2026, Perplexity restricted some Pro accounts linked to promotional codes, citing fraud and unauthorized resale. Affected accounts saw weekly query caps drop from 200 to 100 without notice.

Official legitimate promotional code partners include PayPal, Samsung, T-Mobile, Venmo, and Xfinity. If you believe your account was affected in error, contact support@perplexity.ai or visit perplexity.ai/help.

Data privacy by tier

PlanData used for training?
Free Potentially yes (consumer terms apply)
Pro / Max No (opt-out available in settings)
Enterprise Pro / Max Never used, by default

Good Habits by Product Area

The right habit for the right tool. Using the wrong product for a task is the most expensive mistake you can make with Perplexity.

📜 When to Use Which Product

Task type Use this Why
Quick fact lookup, current events, simple question Quick Search Fast, free, no quota consumed
Multi-angle research, targeted topic with several dimensions Pro Search Sub-queries and more sources; consumes rolling weekly quota
Comprehensive reports, literature reviews, industry analyses Deep Research Autonomous multi-step workflow, 100+ citations; consumes monthly limit
Repeatable workflows, multi-tool execution, finished deliverables Computer Agentic execution across 400+ apps; consumes credits
Writing, coding, iteration, detailed editing Claude, ChatGPT, Cursor Perplexity is strongest at research and source discovery; other tools excel at generation and iteration

Search Habits

  • Match mode to task. Set the focus mode before every search. Academic for peer-reviewed sources, Reddit for opinions, News for recent events. Never leave it on "All" when a narrower mode applies.
  • Include decision context in your prompt. Generic questions get generic answers. Include: what decision you are making, what "good" looks like, and any constraints (budget, team size, existing tools).
  • Chain queries within a thread. Perplexity maintains context. Build on prior answers rather than starting fresh for each follow-up. The research compounds.
  • Always click citations for critical claims. Perplexity is especially prone to misinterpreting statistics, prices, and dates. Any figure going into a client deliverable needs to be traced to its original source.
  • Use Spaces for ongoing projects. A well-configured Space carries instructions and curated sources across all conversations in a project. Configure it once at project start.
  • Save reusable prompt templates. Develop standard prompts for recurring research types: market analysis, competitor teardown, literature review. Copy and adapt rather than starting from scratch.

Deep Research Habits

  • Reserve Deep Research for genuinely complex tasks. On Pro (20/month), each run is a limited resource. Do not use it for questions answerable by Pro Search.
  • Write a specific brief before submitting. Specify: topic and scope, audience, desired format (section headers, bullets, executive summary), and any data sources or constraints.
  • Follow up in the same thread. Post-report follow-up questions use Pro Search, not another Deep Research run. Drill into specifics without spending a second run.
  • Export the report immediately. Copy the full output to a document. Thread content is not permanently archived across sessions.
  • Verify every statistic before use. Deep Research reports cite 100+ sources. Spot-check the most important figures, especially any numbers, percentages, or pricing data, before including them in deliverables.

Computer Habits

  • Scope every task before execution. Define the deliverable, limit the number of sources or bullets, specify the output format, and explicitly tell it to stop after a specific output. Vague prompts burn credits exponentially faster.
  • Use fresh threads to reduce context cost. Long threads with accumulated context increase per-query credit cost. Split strategy (planning) from execution into separate threads.
  • Use workflow templates for recurring tasks. A scripted workflow is cheaper and more consistent than a fresh agent conversation for tasks you run weekly. The 70+ pre-built templates are a starting point.
  • Require human review on client-facing content. Computer can produce confident-sounding errors. Fact-check numerical claims, validate data from less-known sources, and review tone before delivery.
  • Leverage the 70+ pre-built templates. Share, schedule, and customize them across your team rather than building from scratch.

Prompt Best Practices

  • Give Perplexity a role. Start with "You are a senior analyst..." or "Act as a competitive intelligence researcher..." Performance improves with specific role framing.
  • Specify format. "Return 5 bullets, each under 25 words" or "Write a structured report with H2 headers" produces directly usable output.
  • Include constraints. Tell Perplexity what NOT to include as much as what to include. "Do not include general background. Focus only on the last 12 months."
  • When output misses the mark, add one constraint. Audience, format, length, or a "do not include" instruction usually fixes it without starting over.
  • Iterate, do not regenerate. If the first answer is 70% right, ask for a targeted revision of the specific part that is wrong. Do not ask it to redo the whole thing.

Common mistakes to avoid

  • Using Deep Research for simple questions (wastes your monthly 20 on Pro)
  • Sending open-ended prompts to Computer (burns credits at 10x the rate)
  • Not verifying citations before including figures in deliverables
  • Starting a new thread for every follow-up (loses research context)
  • Using Perplexity for writing and editing when Claude or ChatGPT are better tools for that

Universal Tips

Right tool for the right job
Perplexity excels at research, source discovery, and fact-checking. For writing, coding, and iteration: Claude, ChatGPT, and Cursor often perform better.
Build a prompt library
Save prompts that work well for your recurring research types. A good prompt template is worth dozens of hours over a year.
Use real-time advantage
Ask "What has [competitor] announced in the last 30 days?" Perplexity's live web access is its biggest differentiator from static AI tools.
Click every citation
For any claim going into a deliverable, trace it to the original source. Statistics and prices are where Perplexity is most prone to error.
Configure Spaces early
Every new thread is a cold start. A Space with curated sources and standing instructions carries context across an entire project lifecycle.
Always verify before sharing
Perplexity is a research accelerator, not a final authority. Review every output for accuracy and appropriateness before sharing externally.
🎓 Perplexity Search & Deep Research Resources

Tutorials and official docs for Perplexity Search, Pro Search, Focus modes, and Deep Research.

Perplexity AI – Official 📄 Getting Started Guide · Free Getting Started with Perplexity Official Perplexity guide covering search basics, Focus modes, Pro Search, Citations, and using Spaces for ongoing research projects. Read Guide → Perplexity AI – Official 📄 Help Center · Free Perplexity Help Center Official support documentation covering all Perplexity features: search modes, Focus filters, file uploads, Spaces, and account settings. Open Help Center → Perplexity AI – Official Blog 📄 Blog Post · Free Introducing Perplexity Deep Research Official Perplexity blog post introducing Deep Research: how it works, what it produces, how to export reports, and when to use it over Pro Search. Read Post → YouTube · 2024 📹 Video · Beginner · Full Tutorial How to Use Perplexity AI for Beginners: Complete Tutorial Step-by-step beginner walkthrough covering search modes, Focus filters, Citations, follow-up threads, Collections, and file attachments in Perplexity. Watch on YouTube → Perplexity Developer Docs – Official 📄 Technical Docs · Free Perplexity API Overview Official developer documentation covering the Perplexity API, Sonar models, search tools, and how to integrate Perplexity search into workflows and applications. View Docs → YouTube · 2026 📹 Video · Step-by-Step · Intermediate How to Control Search Results in Perplexity AI Practical walkthrough on adjusting Focus modes, using Deep Research, refining AI responses for accuracy, and controlling result relevance in Perplexity. Watch on YouTube →
⚡ Perplexity Computer Resources

Learn how to use Perplexity Computer for agentic workflows, multi-model execution, and 400+ app integrations.

YouTube · Apr 2026 📹 Video · Beginner · Overview Perplexity Computer for Beginners: Everything You Need to Know Comprehensive beginner guide to Perplexity Computer: what it is, how the 19-model orchestration works, and real demo prompts to try yourself. Watch on YouTube → YouTube · Mar 2026 📹 Video · Intermediate · Use Cases Perplexity Computer Agent Tutorial: Sales, Marketing & Workflow Automation Hands-on tutorial building a sales and marketing agent with Perplexity Computer: lead research, outreach automation, expense auditing, and multi-step agentic workflows. Watch on YouTube → Perplexity AI – Official 📄 Product Page · Free Perplexity Computer: Official Overview & Getting Started Official Perplexity product page for Computer: what it does, how to connect your apps, workflow templates, and how to get started with agentic task execution. Read Official Docs → YouTube · Nov 2025 📹 Video · Intermediate · Integration Build an AI Research Agent with Perplexity AI (n8n Tutorial) Step-by-step guide to building a Perplexity-powered AI research agent in n8n -- covers API connection, real-time web search, and CRM lead enrichment automation. Watch on YouTube →

Resources are provided for informational purposes. Always verify content is current before sharing.

This guide is for training and orientation only. Plans, features, limits, and pricing may change as Perplexity updates its products. Information reflects the mid-2026 product state. Always check perplexity.ai for current details.