Perplexity AI Guide 2025: Search, Pro Features & Tips
Perplexity gives cited, real-time answers from the web. Free tier for everyday research. Pro ($20/mo) adds Claude 3.5 Sonnet / GPT-4o / Gemini model choice, 300 Pro searches/day, file uploads, Collections, and $5/mo in API credits.
1. Perplexity vs Google vs ChatGPT
Perplexity occupies a unique position between Google (returns links, no answer) and ChatGPT (gives answers, may not cite sources):
| Feature | ChatGPT | Perplexity | |
|---|---|---|---|
| Direct answer | Snippets only | ✓ (no cite) | ✓ + cited |
| Real-time web | ✓ | Plus only | ✓ always |
| Source citations | Links list | ✗ | ✓ inline [1][2] |
| File upload | ✗ | Plus only | Pro only |
| Free tier | ✓ | ✓ limited | ✓ unlimited basic |
2. Focus modes
Focus modes restrict which sources Perplexity searches. Switch Focus before typing your query via the icon next to the search bar (or Alt+F).
Web (default)
Searches the open internet. Best for general questions, news, product research, and anything that doesn't need a specialized database.
Academic
Searches Semantic Scholar, arXiv, PubMed, and other academic databases. Essential for research papers — citations point to actual papers, not blog summaries. Best for literature review, methodology research, and scientific questions.
YouTube
Searches YouTube video transcripts. Ask "how to set up Supabase RLS" — Perplexity returns answers cited from actual tutorial videos. Good for finding specific instructions from video content without watching 40-minute tutorials.
Searches Reddit threads. Best for "what do real users think about X" — software reviews, tool comparisons, and community opinions that don't appear in SEO-optimized blog posts.
Writing
Disables web search and uses the model's internal knowledge only — like ChatGPT mode. Best for creative writing, brainstorming, or when you don't need real-time sources.
Wolfram
Routes queries to Wolfram Alpha for computation — math, unit conversions, chemical equations, statistics. More reliable than LLM math for precise calculations.
3. Pro Search (Pro tier)
Pro Search uses advanced models (Claude 3.5 Sonnet, GPT-4o, Gemini 1.5 Pro) instead of Perplexity's own Sonar model. It performs multi-step reasoning — generating sub-queries, searching more sources, and synthesizing more complex answers.
How Pro Search works
1. Question decomposition: Pro Search breaks your question into sub-queries. Ask "What are the best databases for a SaaS application in 2025?" and it generates separate searches for relational databases, vector databases, and recent benchmarks.
2. Multi-source synthesis: It searches more sources (typically 15–30 vs 6–10 for Basic) and synthesizes them with the advanced model's reasoning capabilities.
3. Model selection: Switch between Claude 3.5 Sonnet (best reasoning), GPT-4o (versatile), or Gemini 1.5 Pro (long context) in the Pro Search panel.
4. Collections — organizing research
Collections are Perplexity's answer to research organization — like folders for your search threads. Save related searches together, add notes, and share them with team members.
Creating a Collection
- Click Collections in the left sidebar
- Click "New Collection" — give it a name and description
- Save any search thread to a Collection via the bookmark icon
- Threads in a Collection share context — Perplexity can reference earlier searches
Use Collections for multi-session research
Create a Collection for a project (e.g. "Competitor research Q2 2025") and save relevant searches there. Come back days later and Perplexity shows all related threads in one place. You can also share a Collection with team members via a public link.
5. Perplexity API (pplx-api)
The Perplexity API gives programmatic access to Perplexity's search-augmented models via an OpenAI-compatible interface. Pro plan includes $5/mo in API credits automatically.
API basics
Base URL: https://api.perplexity.ai
Auth: Authorization: Bearer pplx-YOUR_KEY
Endpoint: POST /chat/completions (OpenAI-compatible)
Model: sonar-pro — the online model that includes real-time web search + citations in API responses
OpenAI SDK: Use with base_url="https://api.perplexity.ai" and your pplx key as api_key
| Model | Web search | Context | Price |
|---|---|---|---|
| sonar-pro | ✓ | 127k | $3/1M in, $15/1M out |
| sonar | ✓ | 127k | $1/1M in, $1/1M out |
| sonar-reasoning-pro | ✓ | 127k | $2/1M in, $8/1M out |
citations array with URLs for each source used in the answer. Parse this array to display inline citations in your application.
6. Perplexity Pro vs free
| Feature | Free | Pro ($20/mo) |
|---|---|---|
| Basic searches | Unlimited | Unlimited |
| Pro searches (Claude/GPT-4o) | 5/day | 300/day |
| File uploads (PDF, CSV) | ✗ | ✓ |
| Image generation | ✗ | ✓ |
| API credits included | ✗ | $5/mo |
| Collections | Limited | Unlimited |
7. Tips & prompts for deep research
"Summarize the last 5 years of research on {topic}. Use Academic focus. List the 5 most cited papers with their key findings."
"What are users saying about {product} vs {competitor} in 2025? Use Reddit focus. Highlight the most common complaints and praised features."
"Show me how to {specific technical task} in 2025. Use YouTube focus. Include the most relevant tutorials and step-by-step instructions."
"Compare {option A} vs {option B} for {use case} in 2025. Consider: pricing, accuracy, API quality, community size, and recent updates. Give a recommendation for my situation: {brief context}."
Monitor Perplexity status
Perplexity outages mean your research pipeline stops and API calls fail. Track Perplexity uptime at prismix.dev and get free email alerts so you know immediately.
FAQ
What is Perplexity AI?
Perplexity is an AI search engine that gives direct answers with cited web sources. Unlike Google (returns links) or ChatGPT (answers without citations), Perplexity gives synthesized answers with inline source numbers [1][2] you can verify.
Is Perplexity free?
Yes — unlimited basic searches are free. Pro ($20/mo) adds 300 Pro searches/day using Claude 3.5 Sonnet, GPT-4o, or Gemini, plus file uploads, image generation, and $5/mo in API credits.
How is Perplexity different from ChatGPT?
Perplexity always cites sources from real-time web search. ChatGPT (without browsing) answers from training data without citations. Use Perplexity for research needing current, verifiable sources. Use ChatGPT for creative writing, code, and complex generation tasks.
What is Perplexity Pro and is it worth it?
Perplexity Pro ($20/mo) adds: 300 Pro searches/day with Claude/GPT-4o/Gemini, file upload (PDF/CSV), image generation, and $5/mo API credits. For daily research use it's worth it. Casual users are well-served by the free unlimited basic tier.