Perplexity searches the web and presents answers with source links. Use it to find starting points, compare explanations, and follow citations back to the original material.
Perplexity
Explore features, practical uses and pricing below.
Perplexity is an AI search and research assistant that turns a question into a conversational answer with source links. You can use it to find starting points, compare explanations, investigate a product requirement, or assemble a report from online and supplied material. Follow-up questions help narrow an investigation without beginning a new search from scratch.
The current official product overview also includes tools for creating work and completing tasks. Search remains a useful entry point, but Research, Projects, asset creation, and Perplexity Computer serve different needs. The practical value is being able to move from a question to source material you can inspect. Citations do not guarantee correctness, and a polished report is not a substitute for checking the evidence. The workflows below are editorial examples rather than claims of hands-on testing.
Perplexity combines retrieved information with an AI-generated explanation. For an unfamiliar technical topic, that can help establish the vocabulary and locate relevant documents. For a current product question, it can point toward a feature guide or pricing page. The follow-up is often as important as the first answer: ask which part of a claim is established, which is inferred, and what source would resolve the uncertainty.
Begin with the decision you are trying to make. “Which booking systems support a required deposit for a small training business in the UK?” gives a research assistant more useful constraints than “Best booking software.” Specify the market, customer type, must-have behavior, and any date-sensitive requirements. Ask for official documentation behind the necessary capability rather than a generic ranking.
Read citations in context. A page might confirm a feature exists without establishing that it is included in the cheapest plan. A vendor announcement might describe a rollout without proving access for your account. Keep those distinctions visible in your notes so the final comparison does not turn partial evidence into a stronger conclusion.
Pro Search is documented as a deeper search workflow that combines multiple searches, source synthesis, citations, and contextual follow-ups. Its capabilities include supported code interpretation and different kinds of sources. Eligible paid users can select advanced models, while available modes and allowances depend on the subscription and interface.
This is useful when the answer depends on more than finding a single fact. A product manager comparing data export options can ask about supported formats, scheduled export, administrator permissions, and plan restrictions. Keep each required behavior in a separate comparison row. That makes it easier to notice when an answer describes a nearby feature but does not satisfy the requirement.
The depth of the answer should be judged by the evidence it contains, not by the number of paragraphs or citations. Ask for the exact source that supports a critical difference and identify gaps explicitly. Pro Search and other tools have usage limits, so check the current plan comparison rather than relying on older reviews that describe unlimited or fixed daily allowances.
Research conducts an iterative investigation and synthesizes a report. Perplexity documents search, document reading, reasoning, and export or sharing workflows. Research selects its own model combination instead of offering the same manual selection as an ordinary search. Free users have limited access, and paid plans provide additional allowances.
The Advanced Deep Research guide describes clarifying questions, follow-ups during research, progress information, uploaded-document analysis, calculations, and editable reports. Those controls matter when the scope develops as new information appears. A broad prompt can be corrected before the assistant spends the entire investigation on an irrelevant interpretation.
For a market overview, define the geography, segment, period, and evidence standards. Ask for a separate treatment of official data, vendor claims, and the report’s analysis. Require the assumptions behind any estimate. If several sources repeat the same statistic, investigate where it originated; repeated publication does not create several independent measurements. Read the report as an organized starting point for review, not as proof of exhaustive research.
Perplexity integrates its own models and models from other providers. Its model guide explains that selection depends on the plan, mode, and organization settings, and that the model selector is the current source of truth for an account. Free users do not have the same manual access to advanced Search models as paid users.
A provider’s model inside Perplexity runs within Perplexity’s search tools, instructions, and limits. Seeing a Claude, Gemini, or OpenAI model in the selector does not include the provider’s separate application or every feature available through it.
Model Council runs a question across several models and synthesizes an answer, with access rules and consumption that depend on where it is used. Multiple models can help reveal different interpretations, but agreement between them is not independent proof. Trace a factual conclusion back to its source rather than treating repeated model agreement as verification.
Perplexity’s current file guide documents uploads of text, PDFs, code, images, audio, and video. Files can provide context for follow-up questions in a session. The guide says long files may be processed by extracting the parts most relevant to the question, so do not assume a summary represents every section.
For audio and video attachments, the documented workflow transcribes spoken content, labels speakers, and makes the transcription searchable. Visual scenes in video are not currently indexed or searchable in that workflow. A request to find a spoken statement is therefore different from asking what an unspoken diagram showed on screen.
Free, Pro, and Max uploads have rolling weekly allowances; images and files from connected sources can count toward the same allowance. Keep an original copy of the material. For a meeting recording, verify important speaker attribution and wording against the recording before quoting it. For a report, ask about a specific section and inspect the supporting passage instead of relying only on an overall summary.
The official Projects guide describes persistent workspaces for Search conversations, Computer tasks, files, instructions, and connected tools. Older descriptions may call this organization feature Spaces; the current help article uses Projects. You can keep related work together and define instructions for how the assistant should handle that project.
A suitable project might contain a product brief, an approved requirement list, several research sessions, and the checked comparison table. Give files meaningful dates or versions, and identify which document controls a disputed requirement. Instructions can ask for missing evidence to be marked explicitly and for estimates to remain separate from confirmed facts.
Projects have sharing roles and plan-dependent collaboration rules. Perplexity documents restricted access as the default and says project files are available to members with the relevant permissions. Inspect the membership and scope before uploading internal material. A convenient research hub still needs deliberate source maintenance; retaining an old file is not the same as keeping its content current.
The asset creation guide documents documents, spreadsheets, presentations, and HTML apps for eligible paid plans. You can preview the output, request changes, revisit edit history, download supported formats, and export appropriate files to connected Google Drive. The guide lists DOCX and PDF documents, XLSX spreadsheets, and PPTX presentations among the supported outputs.
This can help turn a checked investigation into material others can review. Ask for a concise decision brief first, then a presentation built from that accepted content. Keep the evidence attached to the claim that it supports, and check whether citations survive in the format you intend to share. Perplexity’s documentation specifically describes citations for documents; do not assume every visual artifact has equivalent citation behavior.
Layout requirements deserve their own check. The guide says specific slide layouts or templates are not currently directly supported. Download the file and apply the required theme or structure in compatible software when necessary. Review formulas in a spreadsheet and the actual behavior of an app rather than accepting its preview as evidence of correctness.
Perplexity Computer extends the product toward work that uses tools and creates deliverables. Its documentation describes background tasks, connected services, research, coding, documents, and scheduling. Computer has its own access and credit requirements, so its capabilities should be evaluated separately from ordinary search allowances.
For a recurring research task, describe the desired output, the approved sources, the schedule, and the condition that should trigger a notification or action. Begin with a reviewable report before adding changes to external systems. A weekly supplier update, for example, needs a definition of what counts as a meaningful change and a way to confirm the information came from the intended supplier.
Tool access makes precision more important. Retrieving a record and editing it are different requests, and a connected account determines what is accessible. Review destinations, permissions, and task status in the relevant application. A research answer should never be read as confirmation that an external action happened unless the result provides evidence of that action.
Consider a small organization choosing a customer support platform. It requires an exportable knowledge base, role controls, and a way to migrate its existing articles. A useful Perplexity workflow is to build a shortlist around evidence rather than a broad “best software” recommendation:
This example uses Perplexity to locate and organize evidence. It does not treat the generated shortlist as a purchase decision. The organization still has to verify account-specific access and complete the trial that determines whether the migration works.
Perplexity is relevant to researchers, students, analysts, product teams, and people who routinely need to understand unfamiliar online material. It can be particularly convenient when the next useful step is opening a source and asking a narrower question. For a writing task, use the research to establish facts before drafting your own argument or commissioning a separate editing pass.
Its limitations follow from the evidence available. Search can miss a source, inaccessible pages can leave gaps, and a citation can support only part of the sentence beside it. Recent information can also be wrong or provisional. Compare publication dates with the dates of the underlying events, and avoid combining incompatible populations or definitions in a numerical comparison.
For academic research, inspect the original studies and their methods; a summary does not establish the strength of the findings. You can also explore Consensus or Elicit as separate options for scholarly workflows. Broader assistant comparisons include ChatGPT, Claude, and Google Gemini. Compare them on source quality and the corrections needed for your own recurring task.
The official plan comparison lists Standard Free, Pro, Max, Education Pro, Enterprise Pro, and Enterprise Max, alongside a separate API offering. The product site lists Pro at US$20 per month or US$200 per year. Verify the checkout for your location and purchase channel, including any promotion and renewal terms.
Free provides core search with limited access to advanced features. Pro expands research, model, upload, and creation access; Max offers higher access for heavier use. The current help guides document allowances, so avoid interpreting marketing references to unlimited access as unlimited use of every feature. Developer API usage is billed separately from subscriptions.
Perplexity’s data collection guide says consumer users can turn off AI training data collection through account preferences. The opt-out applies to subsequently collected data and does not stop all service processing. Enterprise data is excluded from AI training under the documented rules. Review organization controls, file handling, project membership, and connected-service permissions before using private work material.
It combines conversational answers with search and research workflows. You can ask follow-up questions, but its source discovery and citations are central reasons to use it for online investigation.
No. Open the source and check that it supports the specific statement. Review context, definitions, and dates. Several citations can repeat one underlying claim without independently confirming it.
No. Eligible paid Search users can select available models, while Research chooses its own combination. Other tools have their own rules. The account’s current selector is more useful than an old list of model names.
No. The current file guide explicitly describes rolling weekly allowances on Free, Pro, and Max. Images and connected-source attachments can count toward the same allowance.
The documented upload workflow indexes transcribed speech and speaker information. It does not currently index visual scenes. Use the original video when an answer depends on something shown without being spoken.
No. You are using the model within Perplexity’s tools and limits. A related model does not give you the separate application, subscriptions, or all of its provider-specific features.