How to Use Perplexity AI

Overview

Perplexity AI is a web-based answer and research service that combines search with generated summaries and links to sources. A user enters a question, reviews a synthesized answer and its citations, then follows up or opens the underlying pages. Perplexity was founded in 2022 and developed around this search-and-answer approach; the service now also offers mobile apps, file analysis, longer research modes, and persistent workspaces called Projects. Reimagining search: Perplexity drives productivity with generative AI-powered answer engine | AWS Startups describes its founding and early product direction, while the What is Pro Search? | Perplexity Help Center explains how current Search sessions work.

The best way to use Perplexity is to treat its answer as a research lead, not as proof. State the question and the kind of answer needed, select a deeper mode when appropriate, inspect citations for the claims that matter, and corroborate high-stakes information with original sources. Citations make checking easier but do not guarantee that every claim is correct or fully supported.

What Perplexity does

Perplexity describes its Search process as searching relevant sources, synthesizing material into an answer, attaching source links, and allowing contextual follow-up questions. Depending on the interface, account and selected mode, it can search the web or focus on material such as academic sources, files, or finance-related content. Available models and search options change over time and may differ by subscription. (What is Pro Search? | Perplexity Help Center; What advanced AI models are included in my subscription? - Perplexity Help Center)

This differs from a conventional search-results page: instead of only returning links, the service generates a prose response based on retrieved information. That can reduce the work of locating and comparing material, but it also creates a new task—checking whether the generated summary accurately represents its sources. Search indexes, source selection, models, and product features are not fixed; verify current menus and plan access in the app or Help Center rather than relying on old screenshots or model lists.

Quick start: how to use Perplexity AI on the web

1. Open Perplexity Search | Accurate, Cited AI Answers for Every Question and start a search. Signing in may be useful for saving and revisiting sessions, but available features can vary by account and plan.
2. Type a question in ordinary language. Include the subject, relevant location or time period, and what you want returned. For example: “What changed in the latest UK rules on home insulation grants? Summarise the changes, give the effective dates, and link to official government guidance.”
3. Read the response and identify which statements answer your question. Open the citations beside the relevant claims; do not assume a source supports an entire paragraph merely because a citation appears nearby.
4. Ask a focused follow-up in the same session to clarify or test the answer. For example: “Which of those changes is stated directly in the government guidance?” or “What evidence contradicts the claim about costs?”
5. Save, share, or organize useful work using the session and project controls available in your account.

Perplexity’s current Help Center says that Search applies Pro Search automatically when it is included in a user’s plan; it is no longer necessarily a separate toggle. In the web interface, users can select Search, leave the model choice on Best, or choose another available model. Deep Research is the report-style option for a longer investigation. Interface names and access limits can change. (What is Pro Search? | Perplexity Help Center)

A simple research process

How to read the process: A useful search is iterative. If citations are weak, out of date, or not directly relevant, refine the question and inspect another source rather than repeating the answer as fact. Perplexity documents follow-up questions and source links as core Search functions. (What is Pro Search? | Perplexity Help Center)

Choosing the right feature

Perplexity describes Search as suited to everyday questions and Deep Research as intended for in-depth reports with more sources and extended analysis. The model selector and mode availability are account- and plan-dependent; the provider warns that model choices can change. (What advanced AI models are included in my subscription? - Perplexity Help Center)

Deep Research

Use Deep Research when a question needs several lines of inquiry, a broader source set, or a structured report rather than a short answer. Define what should be covered and what should not. For instance: “Compare the current public transport fare policies in these three cities, use official transit-agency sources where possible, state the date of each policy, and separate confirmed facts from proposals.” Treat the resulting report as a starting synthesis; open important sources and check dates, definitions, and exceptions yourself.

Files and images

To use a file, select the attachment control near the search bar and add the relevant document; Perplexity also documents drag-and-drop on supported desktop interfaces. Ask a concrete question such as “List the three renewal dates in this contract and quote the surrounding wording” rather than the vague “Summarise this.” The Help Center says short files can be analysed in their entirety, while long files may be reduced to their most relevant parts for the query. This means a summary is not a guarantee that every section or detail was examined. It also documents a 40 MB file size limit and rolling weekly upload allowances for individual plans, with limits varying by plan. (File Uploads - Perplexity Help Center)

For an image, the mobile app’s documented workflow is to tap the plus icon beneath the search box and choose an image or take a photo; desktop users can use Attach or drag and drop. Ask what to inspect—for example, “Read the labels and explain what this chart compares.” The official image-upload article lists JPEG, HEIF, PNG, and PDF among accepted formats and says images count toward the file-upload allowance. (Uploading Images on Perplexity - Perplexity Help Center)

Projects and persistent context

Projects are persistent workspaces for searches, files, instructions, and, where available, connected tools. They are useful when a task continues across sessions—for example, a course, a product launch, or a recurring market review. Create a Project from the Projects area, give it a clear name, add relevant context and instructions, and keep only the materials needed for that work. A Project’s access and sharing settings matter: Perplexity says personal Projects are private by default and shared only with people granted access. (What are Projects? - Perplexity Help Center)

How to use the Perplexity AI app

Perplexity offers apps for iOS and Android; its Perplexity links to the app stores. In either app, begin by typing a question into Search, review the answer and citations, and continue the same thread with follow-up questions. Current documentation says Search on iOS and Android can use Pro Search automatically when the feature is included in the user’s plan, and that Deep Research can be selected for a longer report. (What is Pro Search? | Perplexity Help Center)

The app is useful when a question starts with something in the physical world: a user can attach an image or take a picture, then ask about the visible material. Voice input is also available in the mobile product, although features and controls can differ with platform and software version. The ‎Perplexity - AI Search & Chat App - App Store describes voice queries and follow-up conversations; consult the current app listing and in-app settings for the version on your device.

Mobile convenience does not remove the need to verify answers. For a claim that will affect a purchase, assignment, health decision, or work product, open the cited page and confirm the claim directly instead of relying on a quick spoken response.

How to use Perplexity AI effectively

Write requests that can be checked

A focused prompt usually specifies the subject, scope, evidence preference, and desired output. One practical template is:

Question: [What do I need to know?]
Scope: [Place, dates, population, or document]
Evidence: [Prefer primary or official sources; identify disagreements]
Format: [Brief, comparison table, steps, or detailed explanation]
Checks: [Give dates and units; flag uncertainty and unsupported claims]

This is a prompting method, not a guarantee. It helps make the answer easier to assess because the user can see whether the response actually covers the requested scope. Avoid bundling unrelated tasks into one broad query. If the answer is vague, ask one narrower question at a time: “What is the denominator for that percentage?” or “Does the cited source give this date, or is it an inference?”

Use follow-ups to improve the evidence, not just the prose

Follow-ups work best when they direct the next research step. Useful examples include:

• “Use official sources published in the last two years, and list the publication dates.”
• “Separate the source’s direct statement from your interpretation.”
• “Compare the two sources and identify where their definitions differ.”
• “Find a primary source for the statistic and explain its population and measurement period.”
• “What evidence would change this conclusion?”

Perplexity says Search keeps conversational context so users can refine an earlier answer. If a thread has drifted, restate the key question and constraints or start a separate search so unrelated prior context does not shape the next result. (What is Pro Search? | Perplexity Help Center)

Organize repeated work

For a single question, an ordinary Search session is usually enough. If the same project requires repeated searches, files, or instructions, use a Project to keep that context in one place. Project instructions can specify a recurring response format or which kinds of sources to prioritize. Review and update uploaded materials: Perplexity notes that a file copied from a connected source does not automatically change when the original is edited. (What are Projects? - Perplexity Help Center)

Check citations and factual claims

Treat citations as links to inspect, not as a certification mark. For each consequential statement:

1. Open the cited source and confirm it exists and is relevant.
2. Find the passage that supports the claim; check whether it supports the full statement, not just a related topic.
3. Check the source’s author, publication date, definitions, and evidence. Prefer a primary document—such as a government rule, research paper, company filing, or dataset—when one is available.
4. Compare independent sources when the issue is contested or the consequence of error is substantial.
5. Check numbers for their unit, denominator, population, location, and period. Do not carry a number into another context without those qualifications.

This caution is supported by research on generative search systems. A 2023 human evaluation of Bing Chat, NeevaAI, Perplexity, and YouChat found that, pooled across the systems and sampled questions, 51.5% of generated sentences were fully supported by their citations and 74.5% of citations supported their associated sentence. These figures are an older, cross-system average—not a measurement of today’s Perplexity product or a current benchmark—but they demonstrate why visible citations should still be checked. (Evaluating Verifiability in Generative Search Engines)

Interpretation: The study tested early commercial generative-search systems using human judgments; its pooled percentages should not be used to estimate the accuracy of a current answer. Its practical implication remains relevant: verify claim-to-source support yourself. (Evaluating Verifiability in Generative Search Engines)

Common uses and limits

Perplexity can help with initial background research, explaining unfamiliar terms, comparing products or policies, summarizing uploaded documents, finding starting points for academic research, and preparing a first draft of a research brief. It is particularly useful when the user wants a concise synthesis with sources to inspect. It is less suitable as the sole basis for legal, medical, financial, or safety decisions, or when a complete audit of a long document is required without checking coverage.

When a task involves a current fact, include “as of” dates or ask Perplexity to identify the latest source dates. When the underlying source is behind a paywall, an abstract or snippet may not contain enough information to validate a technical conclusion. A generated answer can also omit a relevant disagreement or overstate a source’s conclusion; ask specifically for counterevidence and read the sources yourself.

Privacy and responsible use

Perplexity’s privacy notice says that the service processes user content including prompts, queries, uploads, generated outputs, and content such as collections, Pages, or Projects. It also describes use of data for providing and improving services, including AI models, and says voice data may be processed when voice features are used. (Perplexity Privacy Notice)

Before uploading contracts, personal records, unpublished research, client materials, or internal business information, check your organization’s rules and Perplexity’s current terms and settings. Perplexity’s plan documentation says some individual plan users can opt out of data collection in settings; the control and applicable policy may vary. An opt-out should not be assumed to mean that information is deleted, that no operational processing occurs, or that a setting applies retroactively. Avoid entering sensitive information unless you are authorized to do so and understand how it will be handled. (Which Perplexity Subscription Plan is right for you? | Perplexity Help Center)

Use AI-generated text responsibly in school, work, and publishing. Check factual claims, respect copyright and confidentiality requirements, and follow applicable rules for acknowledging assistance. Do not present an answer as a source you personally read; cite the underlying primary or authoritative material instead.

Bottom line

For quick questions, start with Search and ask in specific, ordinary language. Use Deep Research for broader investigations, files for questions grounded in your own documents, and Projects when work needs persistent context. The most effective workflow is to refine the question, inspect the citations, and verify important claims against the underlying sources. Perplexity can accelerate research and help organize information; responsibility for assessing evidence and decisions remains with the user.