These 13 Perplexity prompts are written for what makes Perplexity different from a chatbot: every answer is a live web search with citations. They follow Perplexity's own advice on specific wording, result caps and source handling, and they use its Pro Search focus options, Research mode and Projects. Each one is a fill-in template you can paste as is.
For academic work on papers you already have, like extracting data from a study or checking a reference list, see the AI prompts for research. For working only inside your own documents, NotebookLM prompts are the better fit. Nothing here repeats them.
How to use these prompts
Copy a prompt, replace each blank in double braces, or keep the default after the colon. {{count:5}} means five unless you change it.
Perplexity's getting started guide gives the simplest rule: "Ask in full sentences when context matters, or use two or three keywords for a quick answer." The prompts below are the full-sentence kind. Its developer prompt guide is written for the API, but four of its points matter just as much in the app:
- Specific words find specific pages. "Specificity in
inputdirectly improves retrieval." Use the vocabulary a good source would use, including product names, version numbers and years. - Cap the list. "Without an explicit cap, the model picks an arbitrary length." Ask for the top five, not "some".
- Don't ask for URLs. The guide says asking for links in the prose fights the default citation format, and the model may mistype them. The prompts here ask for named, cited sources instead.
- Give it permission to say no. Tell it to say so explicitly when results don't answer the question, and to state any mismatch, such as a different year or company, before answering.
Follow-ups stay in the same session, so you can refine an answer without starting over. Use that instead of rewriting the whole prompt.
How do you write a search question Perplexity answers well?
These work with the default search or Pro Search. They are short on purpose: each one tells Perplexity what to look for and what shape the answer should take.
1. A precise comparison with a capped result
Use it for any "X vs Y" question. Naming the criteria up front keeps the answer from drifting into generic pros and cons.
Compare {{option_a}} and {{option_b}} for {{use_case:a team of 10 developers}}.
Criteria: {{criteria:price per seat, data residency options, SSO support, API rate limits}}.
Use only information published in {{time_window:2026}}.
Answer as a table with one row per criterion and a final row for
"best fit when". Cite every cell. If a fact can't be found, write
"not published" instead of estimating.
2. The exact fact, with the mismatch rule
For a specific version, model, regulation or product where near-misses are common. The mismatch line comes straight from Perplexity's own advice on hallucinations.
What does {{exact_thing:the 2026 version of the EU AI Act guidance on general-purpose models}} say about {{question}}?
If the results are about something related but different, such as an
earlier version, a draft, a different country or a similar product,
state the mismatch explicitly before answering. If nothing answers
the question, say so explicitly rather than giving a likely answer.
3. Primary sources only
When you need the original document, not articles about it. Pair it with the Academic focus for research, or Web for official pages.
Find the primary sources for this claim: "{{claim}}".
Primary means the original study, official dataset, filing, law,
standard or the organization's own announcement. Do not count news
articles or blogs as primary, but you may use them to locate the
original.
List up to {{count:5}} primary sources, each with who published it,
when, and the exact sentence or figure that supports or contradicts
the claim.
4. Debug an error with versions
For developers. Version numbers are the vocabulary that finds the right issue thread instead of a five-year-old answer.
I get this error with {{tool_and_version:Next.js 16.0.2 on Node 22}}:
{{error_message}}
Find fixes reported for this exact error and these versions. Prefer
official docs, release notes and maintainer replies in issue threads.
List up to {{count:3}} fixes, most likely first, each with the
version it applies to and the date it was posted. Say if a fix is
for an older version than mine.
How do you prompt Perplexity's Research mode?
Research mode is Perplexity's long-form option. Its help article says it runs many searches, reads the sources, refines its plan as it learns, and writes a report you can export as a PDF or document, or turn into a Perplexity Page. It picks its models for you, and most tasks finish in under 3 minutes. Select Research in the mode selector before you paste these.
5. A landscape report with fixed sections
For getting a full picture of a market, technology or policy area. Fixed sections make reports you can compare month to month.
Write a research report on {{topic}} for {{reader:a founder deciding whether to build in this space}}.
Cover {{time_window:the last 18 months}} and {{region:North America and Europe}}.
Sections:
1. Summary in five bullets
2. Market size and growth, with the source and date of each figure
3. The main players and how they differ
4. Regulation and other constraints
5. Where sources disagree
6. Open questions nobody has good data on
Keep each section under 200 words and cite every number.
6. Company due diligence with the Finance focus
Use it before a sales call, a partnership or a job interview. In Pro Search, set the focus to Finance for filings and market data.
Do a due diligence brief on {{company}}.
Include: what they sell and to whom, revenue and growth from their
latest filings or official statements, funding or ownership, leadership
changes in the last year, major lawsuits or regulatory actions, and
the three biggest risks an outsider should know.
Date every figure and name the document it came from. Separate what
the company says about itself from what independent sources report.
7. A literature scan with the Academic focus
For a quick map of the evidence before you read anything in depth. Set the focus to Academic in Pro Search, or run it in Research mode.
Find peer-reviewed studies published since {{since_year:2020}} on
{{research_question}}.
Return up to {{count:10}} studies as a table: authors and year, study
design, sample size, main finding, and one limitation. Put systematic
reviews and meta-analyses first. Then summarize in three sentences
what the evidence suggests overall and how strong it is.
How do you check what Perplexity tells you?
Perplexity's Pro Search article advises you to "validate information by referencing the sources linked in your answer." These prompts make that part of the answer instead of a chore afterwards.
8. Does the evidence support this claim?
Paste a claim from an article, a pitch deck or a colleague. You get a verdict with the evidence on both sides.
Check this claim: "{{claim}}"
Search for evidence for and against it. Give:
1. A verdict: supported, partly supported, unsupported or false.
2. The strongest evidence for, with sources.
3. The strongest evidence against, with sources.
4. What the claim leaves out or exaggerates.
If the claim mixes several facts, check each one separately.
9. Where do the sources disagree?
Use it on any topic where you suspect the answer depends on who you ask. It surfaces the disagreement a single summary hides.
On the question of {{question}}, find where credible sources disagree.
For each disagreement, give the positions, who holds each one, the
evidence each side points to, and whether the disagreement is about
facts, methods or values. End with what kind of evidence would settle
it.
10. Drill into one cited source
Send this as a follow-up after any answer. It checks whether the source really says what the summary claimed.
For the point about {{point}} in your last answer, go back to the
source you cited for it.
Quote the passage from that source that supports the point. Say who
published it and when, whether it is a primary source, and whether
your summary overstated or simplified it. If the source does not
support the point, correct the answer.
How do you use Perplexity Projects and files?
Projects are workspaces that keep related threads, files and instructions together. Perplexity's Projects article says you can add instructions of up to 8,000 characters, attach files and links under Sources, and list web domains to prioritize.
11. Instructions for a recurring research Project
Paste this into a Project's instructions once. The rules then apply to your work in that Project, so your daily questions can stay short.
You are helping me track {{topic:the electric vehicle battery supply chain}} for {{purpose:a weekly investor note}}.
Always:
- Prefer {{preferred_sources:company filings, government data and trade associations}}.
- Date every fact and say if it is older than {{freshness:six months}}.
- Separate confirmed facts from forecasts and opinions.
- Say explicitly when a search finds nothing new.
- Keep answers under {{length:300 words}} unless I ask for more.
- Use tables for any comparison of more than two items.
12. A weekly brief inside the Project
Run it each week in the Project from prompt 11. The instructions handle tone and sources, so the prompt only sets the window.
What changed on {{topic}} in the last {{window:7 days}}?
List up to {{count:5}} developments, most important first. For each
one: what happened, why it matters for {{purpose:our investor note}},
and the date. End with one thing to watch next week.
13. Compare an uploaded file with current public information
Upload a file to the thread or Project first. Useful for checking an old report, a competitor deck or your own documentation against what is true today.
The attached file is {{file_description:our product comparison page from last year}}.
Go through its factual claims one by one and check each against
current public sources. Return a table: claim, still accurate (yes,
no, partly), what is true now, and the source. List the claims you
could not check at the end.
Keep your Perplexity prompts ready to reuse
Prompts 11 and 12 show the pattern that pays off: stable rules go in a Project's instructions, and the daily prompt stays short. Developers calling the API can go further. Perplexity's guide recommends source, date and region limits as search parameters, such as a domain filter or a recency filter, rather than prose, because filters are "applied by the search backend on every call." For how retrieval-based answers work in general, see what is RAG.
To keep the prompts themselves one shortcut away, save them in Kipt. Your prompts live as Markdown in a private GitHub repo you own, blanks like {{topic}} stay fillable, and Cmd+K finds any prompt by name or tag. Kipt is not affiliated with Perplexity.