Skip to content
explore
Perplexity logo

Perplexity

Visit Perplexity

An answer engine that searches the web and cites its sources.

WritingResearchAgents

How it works / How to use

An answer engine that searches the web and cites its sources. Ask a question in a Thread; Perplexity searches the live web, synthesizes an answer, and attaches numbered citations to sources. Search gives fast answers; Research goes deeper across many sources. Pro Search and model choice depend on your plan.

  1. Open Perplexity and start a new Thread with your question or task.
  2. Pick the mode that fits: Search for quick factual answers, Research for deeper multi-source reports.
  3. Read the answer and click citation numbers to verify sources.
  4. Follow up in the same Thread to narrow scope, ask for format changes, or dig into one source.

Quick web search and Q&A

Lead with the question, add context and constraints, and say how detailed the answer should be. Perplexity searches the web in real time and returns cited summaries.

What to provide

  • A clear question or decision you need help with
  • Any scope limits (region, time period, product tier)
  • The output shape you want (short answer, bullets, comparison)

Details that improve the result

  • Search mode scans the web for direct, up-to-date answers with sources
  • Each answer includes numbered citations linking to original pages
  • Straightforward and complex questions both work—scope still matters

Example prompt

What changed in the EU Digital Markets Act enforcement timeline in the last 12 months? Answer in 5 bullets with dates. Focus on gatekeeper obligations, not speculation.

If the first output is not good

  • “Expand only bullet 3 with primary sources.”
  • “Compare EU and UK approaches in a two-column table.”
  • “What would you need to answer this for one specific company?”

Common mistakes

  • Asking several unrelated questions in one Thread without structure
  • Treating summaries as verified without opening citations
  • Leaving out time or geography when they change the answer

Research mode for deep reports

Use Research when you need comprehensive reports. Perplexity analyzes many sources, reasons through the topic, and delivers longer cited insights.

What to provide

  • A broad topic that needs multi-source analysis
  • Depth expected (overview vs strategic report)
  • Sections or angles you want covered

Details that improve the result

  • Research performs extensive searching across sources before synthesizing
  • Good for learning new subjects or strategic planning
  • Still verify critical claims through the linked citations

Example prompt

Research: market landscape for open-source observability tools in 2025–2026. Sections: major vendors, pricing models, open-core vs fully open source, risks for small teams. Cite primary sources.

If the first output is not good

  • “Add a one-page executive summary at the top.”
  • “Remove vendor marketing blogs; prefer docs and analyst reports.”
  • “Turn section 4 into a decision matrix for a 10-person SaaS team.”

Common mistakes

  • Using Research for a single fact that Search would answer faster
  • Not specifying audience, so the report tone misses the mark
  • Skipping citation checks on numbers and dates

Summarizing articles and webpages

Ask Perplexity to summarize content and highlight what matters for your query. Useful for long articles, docs, or pages you paste or link.

What to provide

  • URL, pasted article text, or uploaded document when supported
  • What you need from the summary (decision, risks, quotes)
  • Length and format constraints

Details that improve the result

  • Perplexity can distill key points from lengthy material
  • Say whether you want neutral summary or answer to a specific question about the content
  • Request citations to the original when working from linked pages

Example prompt

Summarize this article for a product manager: main claims, evidence cited, and open questions. Keep under 200 words.

[paste URL or text]

If the first output is not good

  • “Pull direct quotes for the three strongest claims.”
  • “Rewrite for a non-technical executive.”
  • “Compare this article’s conclusion to [second URL].”

Common mistakes

  • Pasting paywalled content without the actual text
  • Asking for summary and full analysis in one message without priority
  • Assuming Perplexity fetched a page you did not link or paste

Pro Search and model selection

On eligible plans, use Pro Search for more comprehensive answers. Select a model when offered—Best picks automatically, or choose specific models such as OpenAI GPT-5 and Anthropic Claude 4.6 Sonnet per Perplexity Help Center.

What to provide

  • A question that benefits from deeper multi-step search
  • Preferred model if your plan exposes model choice
  • Output format and verification needs

Details that improve the result

  • Pro Search conducts thorough research across multiple sources
  • Pro Search model picker includes Best, GPT-5, Claude 4.6 Sonnet, and others on eligible Pro plans (Perplexity Help Center)
  • Match model choice to task: reasoning depth vs speed

Example prompt

Pro Search: compare public pricing and context limits for major coding assistants updated in the last 6 months. Table: Product, Free tier, Paid entry price, Notable limits. Cite official pricing pages only.

If the first output is not good

  • “Re-run with Best model for a shorter answer.”
  • “Add a column for team or enterprise tiers.”
  • “Flag rows where pricing could not be verified.”

Common mistakes

  • Assuming Pro Search or model picker is on your plan without checking settings
  • Choosing a model without changing an unclear prompt
  • Ignoring citations when comparing fast-changing pricing pages

Follow-up Threads and refinement

Continue in the same Thread for context. Reference the previous answer and change one thing at a time instead of restarting.

What to provide

  • The prior answer or the part that needs correction
  • One dimension to change (shorter, deeper, different angle)
  • New constraints or sources to prefer

Details that improve the result

  • Threads keep conversational context for follow-ups
  • Targeted follow-ups are cheaper than rewriting the entire question
  • Ask Perplexity to list what it still could not verify

Example prompt

From your last answer: expand only the security section, prefer official docs and CVE references, and add a short risk summary for a startup without a dedicated security team.

If the first output is not good

  • “Shorten to half length; keep citations.”
  • “Convert the timeline to a markdown table.”
  • “What sources conflicted on this point?”

Common mistakes

  • Starting a new Thread when the old one already has useful context
  • Asking for ten changes at once without priority
  • Forgetting to say which part of the answer to keep unchanged

Structured and formatted output

Perplexity’s help center recommends including instruction, context, input, keywords, and output format. Define structure explicitly in the prompt.

What to provide

  • Raw question or pasted notes
  • Target format (table, JSON, outline, checklist)
  • Column names, ordering, and empty-value rules

Details that improve the result

  • Name sections or columns up front
  • Say whether extra fields are allowed
  • Request markdown tables for human-readable research briefs

Example prompt

Instruction: Compare three email providers for a 20-person team.
Output: markdown table with columns Provider, SPF/DKIM ease, Typical cost, Source URL.
Context: US-based, Google Workspace incumbent, need DMARC reporting.

If the first output is not good

  • “Same table as JSON with keys provider, spfDkimEase, cost, sourceUrl.”
  • “Add a Recommended column with one sentence rationale each.”
  • “Remove providers without official pricing cited.”

Common mistakes

  • Ambiguous column names
  • Not specifying citation requirements per row
  • Asking for JSON without defining required keys

Fact-checking and source verification

Ask Perplexity to answer with citations, then open the numbered sources. Use follow-ups to resolve conflicts between sources.

What to provide

  • Claims you need to validate
  • Preferred source types (official docs, journals, government sites)
  • What counts as sufficient evidence

Details that improve the result

  • Answers include numbered citations mapped to source URLs
  • Use citations to verify numbers, dates, and quotes before acting
  • Ask for conflicting sources explicitly when decisions are high stakes

Example prompt

Verify: does CapCut Pro subscription work across mobile and desktop according to official CapCut help articles? Answer yes/no with quote snippets and links only from capcut.com help pages.

If the first output is not good

  • “List sources that disagree and why.”
  • “When was each source last updated if visible?”
  • “What could not be confirmed from cited pages?”

Common mistakes

  • Treating AI summary text as a primary source
  • Ignoring citations on time-sensitive facts
  • Not specifying official-only sources when that matters

API and developer integrations

Developers use Perplexity’s documented APIs—Search, Agent, Embeddings, and Router—for programmatic web research. Follow docs.perplexity.ai for presets, citations, and streaming behavior.

What to provide

  • Use case: search, grounded answers, or embeddings
  • Query, filters, and response format needs
  • API key and preset or model selection per docs

Details that improve the result

  • Search API returns ranked web results with filtering
  • Agent API provides web-grounded answers with citations in one call
  • Streaming responses map [N] markers to search result ids—collect all batches

Example prompt

Agent API task: given user question about today's AI product launches, return JSON list of {name, announcement, sourceUrl} using web_search tool and json_schema response_format. Cite official blogs or press releases only.

If the first output is not good

  • “Add search_domain_filter for arxiv.org only.”
  • “Use pro-search preset for deeper multi-step retrieval.”
  • “Handle multiple search_results batches when mapping citations.”

Common mistakes

  • Reading only the first search_results batch in multi-step presets
  • Assuming OpenAI SDK types include Perplexity citation fields
  • Skipping citation mapping in streamed responses

How to prompt

Perplexity’s help center recommends clear prompts with instruction, context, input, keywords, and output format. Because answers are web-grounded, scope and citation needs matter as much as the question itself.

Instruction: Explain recent changes to SPF alignment requirements.
Context: Small SaaS sending product email via SendGrid.
Output: 5 bullets + 3 verification links from official sources.
Keywords: SPF, alignment, DMARC, SendGrid.

Include instruction and output format

Say what Perplexity should do and how the answer should look—bullets, table, short paragraph, or JSON.

Example

Instruction: List pros and cons of server vs client rendering for our marketing site. Output: two-column markdown table, max 6 rows each side.

Bound time, place, and sources

Limit geography, date range, or domain when facts change quickly or official sources matter.

Example

What are CapCut auto-caption steps on desktop as of 2026? Use capcut.com help articles only. Last 12 months if dates are shown.

Choose Search vs Research deliberately

Search for fast cited answers; Research when the topic needs broader multi-source analysis.

Example

Research mode: competitive landscape for AI code editors—features, pricing entry points, and team plans. Cite official pricing pages.

Follow up in the Thread

Refine one dimension at a time: shorter, deeper, different section, or stricter sources.

Example

Keep bullets 1–3. Rewrite bullet 4 with only primary academic sources and add DOI links.

Verify through citations

Treat numbered citations as the audit trail. Open sources for numbers, dates, and quotes before decisions.

Example

Answer with citations, then: Which of your cited sources disagree on the pricing figure? Quote the conflicting lines.

Best output tips

Lead with a clear goal

Perplexity performs best when your prompt states the task, background, and how detailed the answer should be—not just a keyword dump.

Include the five prompt elements

Official tips list instruction, context, input, keywords, and output format. Covering each reduces vague or overly broad answers.

Pick Search for speed

Search mode scans the live web and returns direct cited answers—ideal for facts, comparisons, and quick how-to questions.

Use Research for depth

Research analyzes many sources and produces longer reports. Use it when learning a new domain or preparing strategic summaries.

Read the citations

Numbered citations link to original sources. Open them to confirm numbers, dates, and quotes—especially for pricing, policy, and health topics.

Scope time and place

Add date ranges, regions, or product tiers when they change the answer. Web results shift quickly on news and pricing.

Prefer official sources when needed

Say “official docs only” or name domains when marketing pages or forums would mislead. Follow up if citations are weak.

Refine in Threads

Same-Thread follow-ups preserve context. Change one thing at a time: shorter, deeper, different section, or stricter sources.

Use Pro Search on hard questions

Pro Search runs deeper multi-source research on eligible plans. Pair it with explicit output structure for complex comparisons.

Select models intentionally

When your plan exposes model choice, Best picks automatically or you can select listed models such as GPT-5 or Claude 4.6 Sonnet for Pro Search per Perplexity Help Center.

Summarize with a purpose

When summarizing articles or pages, say whether you want neutral overview or answers to a specific question about the material.

Ask about conflicts and gaps

Request sources that disagree or a list of unverified points when preparing decisions from Perplexity output.

Format for reuse

Tables, outlines, and JSON schemas reduce cleanup when pasting research into docs, slides, or tickets.

Integrate via documented APIs

Search, Agent, Embeddings, and Router APIs are documented separately from the consumer app. Use official presets and handle all citation batches when streaming.

Do not skip human verification

Perplexity grounds answers in web sources, but synthesis can still miss nuance. Treat output as a researched draft, not an infallible authority.

  • Start with a clear question, context, and desired output format.
  • Use Search for quick facts; switch to Research for multi-source deep dives.
  • Click citation numbers to verify claims before acting on them.
  • Bound time range and geography when answers are time-sensitive.
  • Continue in the same Thread with targeted follow-ups instead of restarting.
  • On eligible plans, use Pro Search and model selection for harder questions.
  • Ask what could not be verified when decisions are high stakes.
  • For API use, read docs.perplexity.ai for presets, filters, and citation mapping.

Try this AI

Try Perplexity

Product Details

Pricing, features, limits and latest updates

Perplexity

An answer engine that searches the web and cites its sources.

Free / Paid · Free

Pricing Plans

Free

Free

Pro

$20/ Monthly

Max

$200/ Monthly

Enterprise

Custom

Key Features

Chat

Conversational follow-ups on cited search threads, with model choice on paid plans.

API

Sonar API is metered separately from Pro and Max subscriptions.

Agents

Perplexity Computer and Comet run multi-step research and browser tasks; limits depend on plan.

Cited Search

Answers use live web search with numbered citations; Pro Search and Research modes add deeper retrieval on paid plans.

Limits

  • Free Pro Searches: 5 per day. Signed-in free accounts receive five Pro Searches per day (Perplexity Help Center).
  • Free file uploads: 3 per day. Signed-in free accounts receive three file uploads per day (Perplexity Help Center).
  • Research mode model pick: Research mode automatically selects models; manual model selection is not available (Perplexity Help Center).

Ideas / Prompt experiences

Share a prompt that worked for you. Username and email are shown with your submission. External links are not allowed.

Example prompt

Research with citations

Prompt

Compare DMARC, SPF, and DKIM for a developer who has never configured DNS. Output: five bullets, one common misconfiguration, and numbered citations. Scope: official standards and major provider docs only.

Short explanation

Perplexity help recommends instruction, context, keywords, and output format for stronger research answers.

Example prompt

Pro Search comparison table

Prompt

Pro Search: Compare official free-tier limits for Perplexity, ChatGPT, and Claude as documented on each vendor's help or pricing pages. Return a table with Product, Free tier limits, Paid entry price, Source URL.

Variation

Add a short 'What is still plan-dependent' section for anything not explicitly documented.

Short explanation

Use Pro Search for multi-step questions that need broader source coverage than a quick answer.

Share your experience

Required fields are marked. Variation, result, and explanation are optional.

You may also like

ChatGPT logo

ChatGPT

Explore

Conversational AI for writing, coding, research, files, images, and automation. Capabilities and limits vary by plan.

WritingFree / Paid

Claude logo

Claude

Explore

AI assistant for writing, analysis, coding, and collaborative problem-solving.

WritingFree / Paid

Notion AI logo

Notion AI

Explore

AI assistant in Notion for agents, enterprise search, meeting notes, and workspace writing.

WritingFree / Paid

Back to Explore