The short answer
Perplexity, the AI answer engine from Perplexity AI, Inc., gives the best research results when your prompt names the source types you trust, a date range and the output format, because every answer is already cited. Search works on the free plan, which Perplexity's pricing page describes as good for limited daily usage. Deep Research needs Pro, Automations and Perplexity Computer need a paid plan and use credits, and Projects replaced Spaces on 4 August 2026. These 14 prompts use each feature.
What makes Perplexity different for research
Perplexity, made by Perplexity AI, Inc. in San Francisco, searches the web before it answers and puts numbered citations in every reply. That design changes how you should prompt it. You do not need to ask for sources, because you always get them; you need to tell it which sources count. A prompt that names the source types you trust, the date range and the format you want will beat a clever question every time, because Perplexity follows those constraints closely.
It also means Perplexity prompts differ from the general prompts in our research prompt library. The 14 prompts below use Perplexity's own features as they stand in November 2026: Search, Deep Research, Projects, Automations, Perplexity Computer, @Microsoft 365 and Model Council. Each note says which plan a feature needs, because Perplexity's limits differ a lot between Free, Pro and Max. If you are new to the tool, start with our guide to using Perplexity.
Perplexity prompts for quick cited answers
Get a sourced answer with dates, Brief me on an organisation and Explain a technical topic for a non-expert all work on the free plan, which Perplexity's pricing page describes as good for limited daily usage. Write them as full sentences rather than keywords: "best project tool" returns a listicle, while a question that names your team size and constraint returns an answer. The line asking Perplexity to say when it cannot confirm a point is not decoration. It noticeably reduces filler, and it tells you where your own checking should start.
Perplexity prompts for comparisons and tables
Perplexity is at its best when the facts live on public pages and you need them in one place. Compare options in a cited table and Benchmark against competitors both ask it to use each provider's own website for that provider's row, which keeps third-party opinion out of a vendor comparison. Ask for "not stated" in empty cells; otherwise the model may fill a gap with something from a review site. Before you share the table, click the citation behind any price or contract term, because a citation proves where a claim came from, not that it was summarised correctly.
Perplexity prompts for tracking what changes
Build a dated changelog is the fastest way to catch up on a law, a standard or a product you have not followed for months. Schedule a weekly market watch uses Automations, added to Perplexity Computer on 5 October 2026, which run recurring work on a schedule or when something happens in a connected app. Automations are a Computer feature, so they need a paid plan and draw on Computer credits; Perplexity prices credits at 100 to one US dollar. Keep the brief tight and the output short so each run stays cheap.
Perplexity prompts for Deep Research and Computer
Use Run a Deep Research report when a normal answer is too thin. Deep Research is part of the Pro plan; the free plan includes only a small allowance of Research. Name the audience and the decision, because the report will be long and you want it to be long about the right things. Hand Computer a finished deliverable is for when the output is a file. Perplexity Computer, launched on 27 February 2026 and now on every paid plan, runs multi-step work in a cloud sandbox and can produce documents and presentations. Effort Mode lets you choose Light, Standard, High or Ultra in the Computer selector; set it to Light or Standard before you send the prompt, and have Computer show you an outline for approval before it builds the file, so you can stop a weak job before it spends more credits. Find data you can actually cite works in Search or Deep Research and pays off in any report where numbers matter. Pro also adds premium data sources such as Statista and PitchBook, and since October Perplexity can search and cite Wiley journals and books on all plans.
Perplexity prompts for your own files and Projects
Projects, which replaced Spaces on 4 August 2026, are shared workspaces that keep files, memory and connectors together. Paste Set up a research Project into a Project's instructions once and every question asked there follows the same source rules, which is how a team gets consistent research rather than five people prompting five ways. Older guides that mention Spaces are describing the previous version of this feature. Question your own work files uses @Microsoft 365, added in September, which lets Pro, Max, Enterprise Pro and Enterprise Max users with a work or school account search SharePoint and OneDrive and read Excel workbooks. It is the best prompt on this page for answering "is our internal document still right?"
Perplexity prompts for checking claims
Fact-check a paragraph before you send it works on every plan and catches outdated numbers before your reader does. Ask Model Council for a second and third opinion runs one question through several models and shows where they agree and disagree. In Search you pick Model Council from the Search menu under the box, and it always runs three models; Perplexity's help centre lists it there for Max and Enterprise Max, while its pricing page also lists it under Pro, so check inside the app before you rely on it. Inside Computer, any paid plan can type "Run a Model Council on..." or pick it from the dropdown, choose two to eight models and pay in credits.
How to adapt these Perplexity prompts
- Swap the source rules for your field. Replace "official sources, regulators and reputable press" with what counts in your work: case law, clinical guidelines, company filings or platform documentation.
- Always keep a date range. Perplexity can surface old pages that still rank well. "Since [month year]" plus "give the publication date" fixes most of it.
- Use Search before Computer. A normal search does not use credits. Move to Computer only when the result is a file or a multi-step job.
- Follow up in the same thread. "Now only US sources" or "drop anything older than 2026" refines the answer without starting again.
- Switch off training if the topic is sensitive. On Free, Pro and Max, AI data retention is on by default. Turn it off in Account settings under Preferences; Enterprise data is never used for training.
When Perplexity is the right tool, and when it is not
Choose Perplexity when the job starts with "find out": briefings, comparisons, fact checks and market scans. Choose ChatGPT or Claude when the job starts with "write" or "analyse this file", and Google's Gemini Notebook when the answer must come only from documents you supply. Many professionals pair Perplexity for the facts with one assistant for the draft. See ChatGPT vs Perplexity, Claude vs Perplexity, Gemini Notebook vs Perplexity, our best AI for research pick and Perplexity alternatives. Free lessons are at /learn?tool=perplexity, and current prices are on Perplexity's pricing page.
Prompts to copy
Get a sourced answer with dates
[Your question, written as a full sentence]. Use official sources, regulators and reputable press published since [month year]. Give the publication date next to each source, keep the answer under 200 words, and say clearly if a point could not be confirmed.
You get: A short, dated answer where every claim has a numbered source
Brief me on an organisation
Brief me on [organisation] for a [meeting type] with their [role] on [date]. Cover what they do and for whom, how they make money, leadership changes and announcements in the last 12 months with dates, and any public controversies or legal issues. Cite each point and finish with three informed questions I could ask.
You get: A one-screen briefing with dated, clickable sources
Explain a technical topic for a non-expert
Explain [technical topic] to a [role] who needs to make a decision about it, not build it. Use sources from the maker's documentation, standards bodies or reputable press. Cover what it is, why it matters for [our industry], what it costs in time or money if we ignore it, and the two terms I must understand to follow the discussion.
You get: A plain-English explainer grounded in authoritative pages
Compare options in a cited table
Compare [option A], [option B] and [option C] for [use case] in a table. Rows: [criterion 1], [criterion 2], [criterion 3], [criterion 4], free plan or trial, and where the data is stored. Use each provider's own website for its row and write 'not stated' where the site is silent. Give the date of the page each detail came from, and flag any row where the provider's own page and a third-party source disagree.
You get: A comparison table where each cell can be traced to the provider's own page
Benchmark against competitors
Find how [company A], [company B] and [company C] position [product category] on their own websites today: their headline claim, the audience they name, published pricing model and the three features they lead with. Quote their wording where possible, cite each page, and end with the one gap none of them claims.
You get: A positioning snapshot built from competitors' own words
Build a dated changelog
List every official change to [law, standard, product or policy] between [start date] and today, newest first. For each entry: the date, what changed in one sentence, and a link to the official announcement or document. Exclude commentary, opinion pieces and reposts.
You get: A dated timeline built only from primary sources
Schedule a weekly market watch
Every [weekday] at [time], search for news from the past seven days about [market, competitor or regulation]. Send me up to six bullet points, each with one source link and its date. Put anything about pricing, funding, leadership changes or new regulation at the top, and say 'nothing new' rather than repeating old stories.
You get: A recurring digest via Perplexity Automations (paid plans, uses Computer credits)
Run a Deep Research report
Research [topic] for [audience] who need to decide [decision]. Cover: current state with figures and their sources, the main players or options, what changed in the last 12 months, risks and open questions. Use official data, filings, regulators and reputable press. Mark any estimate as an estimate and say where sources disagree.
You get: A long, sourced research report from Deep Research (Pro and above)
Hand Computer a finished deliverable
Create a [number]-slide presentation on [topic] for [audience]. Slide one gives the answer to [question] in two sentences. Each following slide covers one finding with a chart or table where the data allows, and a source line at the bottom. Finish with a slide of open questions and a full source list. Before you build any slides, show me the slide-by-slide outline and the sources you plan to use, and wait for my approval.
You get: A sourced slide deck built by Perplexity Computer, with an outline check before credits are spent on the build
Find data you can actually cite
Find the most recent published figures for [metric] in [market or country]. For each figure give the value, the year it refers to, the publisher, whether it is official statistics, a company estimate or analyst research, and the link. If figures from different publishers disagree, show them side by side and explain the difference in method.
You get: Comparable figures with their type, year and origin made clear
Set up a research Project
This Project is for researching [topic] for [team or client]. Instructions for every answer here: use sources published since [month year], prefer [preferred source types], ignore [sources to exclude], cite every claim, give dates, and keep answers under [number] words unless I ask for more. Our working definitions are: [define key terms].
You get: Standing instructions so every question in a Perplexity Project follows the same rules
Question your own work files
@Microsoft 365 Find the latest [document type, e.g. pricing sheet or board pack] about [topic] in SharePoint or OneDrive. Tell me what it says about [specific question], quote the relevant lines with the file name, and then check the web for anything published since that file was last updated that changes the answer.
You get: An answer that combines your company files with current web sources (Pro, Max or Enterprise with a work or school account)
Fact-check a paragraph before you send it
Check every factual statement in this paragraph: [paste paragraph]. Number each claim, mark it confirmed, outdated, wrong or unconfirmed, and give the source that settles it. Suggest a corrected sentence for anything that is wrong or outdated.
You get: A claim-by-claim check with corrections you can paste back in
Ask Model Council for a second and third opinion
Run a Model Council on this question. Based on current public evidence, is [claim or forecast] likely to hold over the next [time period]? Show where the models agree, where they disagree, what each one raised that the others did not, and which sources each relied on.
You get: Several models' views side by side (in Search, select Model Council from the menu; the help centre lists it for Max and Enterprise Max; in Computer, paid plans use credits)