The short answer (July 2026): Perplexity is an AI-powered search tool that retrieves live web content and returns a synthesized, inline-cited answer rather than a list of links. You can start free at perplexity.ai, which includes basic search and a small daily allotment of Deep Research runs. Pro ($20/month or $200/year) adds unlimited Pro Search, higher Deep Research volume, file uploads, and persistent Spaces workspaces. Max ($200/month or $2,000/year) adds Model Council, which routes your query through three frontier models simultaneously and synthesizes one consensus-divergence report.
Last verified: July 23, 2026, against official docs and pricing pages.
What Perplexity Is and How It Works
Where a search engine returns ranked links and a chatbot returns a freeform reply, Perplexity sits between them. It runs your query through a retrieval layer that fetches and reads live web pages, then generates a cited answer with numbered footnotes you can click to verify.
The underlying model varies by plan and query type. Perplexity's own Sonar model handles fast, cost-efficient queries by default. On paid plans you can route to Sonar Pro for more complex synthesis, Sonar Reasoning Pro for chain-of-thought reasoning on multi-step problems, or Sonar Deep Research for exhaustive multi-session research reports. You can also select models from Anthropic, OpenAI, and Google when you want a specific model's strengths.
Because every answer is grounded in retrieved content, Perplexity is less likely than a standalone chatbot to hallucinate facts and more likely to surface recent information. It still makes errors. For anything high-stakes, read the linked source directly.
Plans and Pricing
| Plan | Price | What it adds |
|---|---|---|
| Free | $0 | Quick Search, limited Pro Search queries per rolling period, limited Deep Research per day |
| Pro | $20/month or $200/year | Unlimited Pro Search, higher Deep Research volume, file uploads, Spaces, frontier model access, Labs |
| Education Pro | $10/month | Same as Pro; student verification required via SheerID |
| Max | $200/month or $2,000/year | Everything in Pro plus Model Council, expanded Deep Research limits, early access to new products including Comet browser agent |
| Enterprise Pro | $40/seat/month | Admin controls, SSO, data isolation, team Spaces |
Annual billing is available on the web app only. The Education Pro plan halves the monthly price and is verified through SheerID; it carries the same features as Pro.
The Three Search Modes
Quick Search delivers fast answers from a small source set. Use it for simple factual lookups where speed is the priority.
Pro Search is the workhorse for most research. It runs a multi-step query, asks clarifying questions where your query is ambiguous, and synthesizes from a broader, higher-quality source set with richer inline citations. Free accounts get a limited number of Pro Search queries per rolling period; Pro subscribers get a substantially larger daily allotment.
Deep Research is Perplexity's autonomous research agent. Submit a query and Perplexity iterates: it searches, reads what it finds, revises its research plan based on what is missing, searches again, and compiles a structured report with an executive summary, key findings, and source references. The process takes 2-4 minutes per query. You can export the output as a PDF, convert it to a Perplexity Page, or share it directly.
At Deep Research's launch, Perplexity reported a 21.1% accuracy score on Humanity's Last Exam, above competing research-mode models at that time. Free users get a small daily allotment of Deep Research queries. Pro users get substantially more. Max users get the highest limits.
Focus Modes: Filtering by Source Type
By default, Perplexity searches the open web. Focus modes restrict retrieval to a specific domain, which raises citation specificity and changes what gets surfaced.
Click the Focus button in the search bar before submitting. Available modes:
- Academic: Peer-reviewed and scholarly publications. Use when you need formally citeable sources.
- News: Recent news outlets. Use for current events where publication date matters.
- YouTube: Summarizes video transcripts. Useful for extracting insights from conference talks or technical demos without watching in full.
- Reddit: Community threads. Useful for real-world experience reports and niche technical discussions.
- Wolfram Alpha: Routes math and computation queries to Wolfram's engine for precise numeric output.
- Finance: Structured financial data on markets, companies, and economic indicators.
You can combine a Focus mode with natural-language date or geography constraints in the same query. "Focus: Academic. Meta-analyses on sleep deprivation published after 2023" produces more targeted results than using either filter alone.
One limitation: Academic mode does not guarantee peer-review exclusivity. Some secondary sources still appear. Verify the source type manually before citing in formal work.
Spaces: Persistent Research Workspaces
A Space is a named project container. It holds conversation threads, uploaded files, and a custom system prompt that applies to every query you run inside it. Upload your source documents once and query them repeatedly across sessions.
To create a Space: click Spaces in the left sidebar, name it, write a system prompt that sets your research context (subject domain, required output format, audience level), and upload source files. Supported formats include PDF, Word documents, plain text, and images. Files stored in a Space persist until you manually delete them.
Files attached directly to a one-off conversation thread are temporary. If you need a document available across multiple sessions, add it to a Space rather than attaching it per query.
Spaces support shared access: invite collaborators as viewers or contributors. Enterprise Pro adds admin controls and data isolation for organizational use.
Pages: Turning Research Into a Shareable Document
After a research thread or a Deep Research session, you can convert the output to a Perplexity Page: a formatted, publicly accessible web document with sections, headers, and inline citations.
To generate one, look for the Share or Create Page option at the end of a thread. Choose a target audience level (beginner, advanced, or expert), edit sections, and publish. The result is a standalone web page with a public URL that anyone can read without a Perplexity account.
Pages are most useful when the deliverable is the research output itself: a technical brief, a literature summary, or a project overview you need to share outside of Perplexity.
Where Perplexity Falls Short
Paywalled content is invisible to Deep Research. If the primary literature on your topic sits behind journal paywalls, Perplexity surfaces abstracts and secondary sources but skips full papers. Academic mode makes this gap more apparent but does not close it.
Synthesis can introduce errors. Perplexity is a retrieval and synthesis layer, and that process can omit nuance, misrepresent the strength of evidence, or conflate claims from different sources. Click through every citation on any fact you intend to use or publish.
Free-tier limits are tight enough to make Deep Research hard to evaluate. You cannot form a genuine opinion on whether Deep Research fits your workflow from a handful of free runs. Perplexity offers a free trial for Pro; that is the reasonable way to assess whether the upgrade justifies the $20/month.
FAQ
Can I use Perplexity for academic research and formal citations?
Yes, but cite the original source Perplexity references, not Perplexity itself. Click each numbered citation, confirm the claim exists in the original document, and format it in your required citation style (APA, MLA, Chicago). Academic focus mode increases the odds that sources are peer-reviewed; manual verification before citation is still necessary.
What is Model Council and do I need it?
Model Council, launched February 5, 2026, runs one query through three frontier AI models simultaneously and produces a synthesized output that flags where models converge and where they diverge. It is available on Max ($200/month) and the web only. Most researchers will not need it. It is most useful when you want to stress-test a conclusion across different reasoning systems or when surfacing disagreement between models is itself the signal you are after.
Does Perplexity have an API for developers?
Yes. The Perplexity API provides access to Sonar, Sonar Pro, Sonar Reasoning Pro, and Sonar Deep Research via REST endpoints. Pricing combines per-token costs with per-request fees. The API recently migrated from a Chat Completions format to the Agent API, which supports a broader range of models and tools including web search, finance search, and file attachments.
How is Perplexity different from ChatGPT or Claude with web search enabled?
All three can search the web and return cited answers. Perplexity is optimized as a search-first tool: the interface centers on cited retrieval threads and every response defaults to grounded output. ChatGPT and Claude are stronger for open-ended writing, code generation, and extended document reasoning. For a full side-by-side breakdown see the AI Weekly ChatGPT vs. Claude vs. Gemini guide.
Is the free plan enough for occasional research?
For quick factual lookups and light research, often yes. Quick Search is unlimited on the free plan and Pro Search plus Deep Research each carry a small daily allotment. The free plan becomes a constraint when you need multiple Deep Research sessions in a day, persistent Spaces for project organization, or the ability to choose which underlying model handles a query.
Stay current on tools like this one: get AI Weekly free, 3 issues a week, read by 40,000+ practitioners.