Perplexity prompt patterns cheatsheet (search-first prompting)
Production-tested Perplexity prompt patterns: focus-mode selection, source filtering, the citation-anchor trick, deep-research workflow, and the patterns that get cited as the answer.
Focus modes — pick before you type
Perplexity's focus modes change the retrieval corpus, not just the answer. Picking the right one is half the result.
| Item | Description | Example |
|---|---|---|
Web | Broad web search. Default for most queries. | |
Academic | Constrained to arXiv, peer-reviewed journals, and academic sources. | Research questions, paper synthesis. |
Social | Pulls from Reddit, X, forums. | User sentiment, community opinion. |
Video | YouTube + video transcripts. | Tutorial extraction. |
Writing | No retrieval — pure generation. | When you want a writing draft, not a search. |
Source filtering
| Item | Description | Example |
|---|---|---|
site: | Restrict search to a domain. | site:anthropic.com extended thinking |
Domain include / exclude | Pro feature in the UI — pin trusted domains, hide untrusted ones. | |
Date range | Restrict to last day / week / month / year. Critical for fast-moving topics like AI. | |
Prompt patterns that get cited verbatim
If you're researching for an article, write queries that match the structural patterns AI search engines like to cite back.
| Item | Description | Example |
|---|---|---|
Question format | Ask a specific question with one clear noun. 'What is X' or 'How do I X' beats 'X overview'. | What is chain-of-thought prompting? |
Specify desired output | Tell Perplexity exactly what to return. | List the top 5 AI image generators in 2026 with one-line pros and cons. |
Citation request | Explicitly demand sources. | Cite each claim with the source URL. |
Date anchor | Anchor recency requirements explicitly. | As of 2026, what's the state of multimodal LLMs? |
Deep Research mode
Pro/Enterprise feature that fans out into 20+ queries and synthesises a long report.
| Item | Description | Example |
|---|---|---|
Use for | Decisions where 5 minutes of human reading is worth 20 minutes of Perplexity work. | Vendor comparisons, market sizing, technical deep-dives. |
Avoid for | Quick facts (overkill) and creative tasks (no retrieval needed). | |
Tip | Start with a tight scope. 'Compare X vs Y on dimensions A, B, C' beats 'Tell me about X'. | |
Workflow integration
| Item | Description | Example |
|---|---|---|
Perplexity Spaces | Persistent contexts (uploaded files, custom instructions). Like a project workspace for a research thread. | |
Pages | Publishable research reports built from a Perplexity thread. Good for shareable artefacts. | |
Comet (browser) | Inline Perplexity inside the browser — fastest way to research a page you're already on. | |
API (Sonar) | Sonar API for embedding Perplexity-grade search into your own product. | |
FAQ
Is Perplexity better than ChatGPT for research?
For factual research with citations, yes — Perplexity is built around cited live web search. ChatGPT can browse, but Perplexity's UI, source filtering, and Deep Research mode are more research-shaped.
Which Perplexity focus mode should I use for academic work?
Academic — it filters retrieval to peer-reviewed sources and arXiv, which dramatically reduces low-quality citations.
Does Perplexity have an API?
Yes — the Sonar API exposes the same retrieve-then-generate pipeline. As of 2026 it's the easiest way to add cited live search to your own product.
Last updated: 2026-06-01.