Prompt Library

Perplexity prompts for market research with sources attached

30 copy-paste prompts

Sourced market, competitor and customer research. Ask for citations. Do not invent volumes, shares or competitor lists.

In short: This page contains 30 copy-paste ready prompts, organized into 6 categories with a description and pro tip for each. The first 5 prompts are free instantly, no signup needed. Drafted with AI writing assistance and checked against our editorial policy.

Louis Corneloup
By Louis Corneloup · Founder, Techpresso
Last updated ·Drafted with AI assistance (how we make prompts)

Market Maps

5 prompts

Category map with citations

1/30

✨ What it does

Maps a category into sourced sub-segments with a definition and typical buyers. Segments without a source are omitted, and no volume or share is added.

Map the category [CATEGORY] for [AUDIENCE IN THIS MARKET] in [GEOGRAPHY]. Return: a 3 sentence definition, 5 to 8 sub-segments, and who typically buys each segment. After every factual sentence, cite a source with title, publisher, date and URL. Prefer primary sources from the last 18 months. If a segment has no source, omit it and write [VERIFY] rather than inventing a niche. Do not invent a market volume, a share, or a growth rate. End with 3 sources I should open next.

💡

Pro tip: Name the geography. Global maps hide the sources you can actually use.

Value chain from cited reports

2/30

✨ What it does

Draws a value chain from published sources, with a cited example per stage. Unnamed actors stay blank, and no take rate is guessed.

Build a value chain for [CATEGORY] in [GEOGRAPHY] using published sources only. Stages I already know: [PASTE KNOWN STAGES]. For each stage: what happens, who the public sources name as typical actors, and one cited example. Cite title, date and URL after each stage. If a stage has no named actor in sources, write [VERIFY] and leave the actor blank. Do not invent a company to fill a box. Do not invent a take rate or a volume moving through the chain.

💡

Pro tip: Paste the stages you already use internally so Perplexity maps public names onto your language.

Adjacent category fence

3/30

✨ What it does

Tables adjacent names as In, Out or Unclear with a citation per row. Disagreement stays Unclear, and size is not used as an argument.

Draw an in or out fence around [CATEGORY] versus adjacent categories for [JOB TO BE DONE]. Adjacent names I am unsure about: [PASTE ADJACENT NAMES]. For each name, say In, Out or Unclear, with a 20 word reason that cites a source. If sources disagree, mark Unclear and quote both. If you cannot find a source, write [VERIFY] and do not invent a definition to force a verdict. Do not invent a market volume to argue size. Output a table with columns Name, Verdict, Citation.

💡

Pro tip: Include the names sales already uses. The fence is for them, not for a textbook.

Geographic scope table

4/30

✨ What it does

Checks each geography you named for a local source and one non-volume fact. Global reports are not reused as local proof.

I need to know where [CATEGORY] is actually discussed in public sources for [RESEARCH GOAL]. Countries or regions I care about: [PASTE GEO LIST]. For each, give: a source that treats this geography as a real market, the date, and one concrete fact that is not a volume. If you cannot find a geography-specific source, write [VERIFY] and say not found, rather than inventing a local scene from a global report. Do not copy a US fact into another country. Cite URLs.

💡

Pro tip: Drop a country from the study if the row is not found. Do not keep it for completeness.

Category glossary with sources

5/30

✨ What it does

Defines 8 to 12 category terms from sources, showing both definitions when they clash. Seed terms with no source stay blank.

Build a glossary of 8 to 12 terms used in [CATEGORY] for [AUDIENCE]. Seed terms: [PASTE SEED TERMS]. For each term: a 20 word plain definition, who uses it according to sources, and a citation with URL and date. If two sources define it differently, show both and do not pick a winner. If a seed term has no source, write [VERIFY] and leave the definition blank. Do not invent a term to look thorough. Do not attach a volume to a definition.

💡

Pro tip: Paste the terms that appear in customer calls, even if they sound unofficial.

Competitor Scans

5 prompts

Named competitor profile

6/30

✨ What it does

Profiles one named competitor from their site, filings and dated news, with a URL after each claim. Missing facts stay blank, and extra rivals are not added.

Profile [COMPETITOR NAME] for a comparison against [OUR PRODUCT] in [GEOGRAPHY]. Cover: what they sell in their own words, who they say they serve, notable public moves in the last 18 months, and open questions. Cite a URL after every factual claim. Prefer the company site, filings, and dated news. If you cannot find a fact, write [VERIFY] and leave it blank. Do not invent a revenue figure, a customer count, or a product they do not list. Do not add rivals I did not name.

💡

Pro tip: Paste your product in one line so the open questions are about gaps, not a generic SWOT.

Feature grid from public pages

7/30

✨ What it does

Fills a feature grid from public pages, using Unclear when a page is silent. Empty columns stay empty when a site cannot be opened.

Build a feature grid for [OUR PRODUCT] versus these named rivals: [PASTE COMPETITOR NAMES]. Features I care about: [PASTE FEATURE LIST]. For each cell, Yes, No, or Unclear, plus the URL you used. If a page does not mention the feature, Unclear, not No, and write [VERIFY]. Do not invent a feature a page never listed. Do not invent a rival. If a name has no working public page, say so and leave that column empty. Add a date you opened each URL.

💡

Pro tip: Treat marketing pages as claims. Confirm Yes in a trial before you brief sales.

Positioning language scrape

8/30

✨ What it does

Copies eight positioning phrases from a rival's own pages, grouped by problem, audience and how they win. Empty groups stay empty.

Collect positioning language for [COMPETITOR NAME] from their homepage, product page and latest blog or newsroom post. Our product: [OUR PRODUCT]. Return 8 quoted phrases, each with the exact sentence, the URL and the date on the page if shown. Then group them under Problem they claim, Who they claim to serve, and How they claim to win. If a group has no quote, write [VERIFY] and leave it empty. Do not paraphrase into a slogan I did not copy. Do not invent a tagline.

💡

Pro tip: Use quotes in your brief so sales do not remix a line the rival never used.

Public moves timeline

9/30

✨ What it does

Builds a dated timeline of a competitor's public moves with a URL per row. Undated items are omitted, and disputed dates are flagged.

Build a timeline of public moves by [COMPETITOR NAME] from [START DATE] to today for [RESEARCH GOAL]. Include launches, leadership notes, funding only if a primary source names it, partnerships, and lawsuits if reported. Each row: date, what happened, source title, URL. If a date is disputed, show both and tag [VERIFY]. Do not invent a round, a valuation, or a launch. If you cannot date an item, omit it. Prefer sources from the stated window.

💡

Pro tip: Set the start date to your last review so the timeline is a delta, not a biography.

Whitespace from sourced claims

10/30

✨ What it does

Lists up to five jobs rivals do not claim, each tied to a source for the closest they come. Thin silence is not treated as a gap.

I want whitespace between [OUR PRODUCT] and these rivals: [PASTE COMPETITOR NAMES]. Our claims I can defend: [PASTE OUR CLAIMS]. From public sources, list 5 jobs rivals do not claim to do, each with a citation showing the closest they come. If you cannot find a source that shows the gap, write [VERIFY] and drop that job. Do not invent a gap from silence on a thin page. Do not invent a volume of underserved users. Cite URLs. End with 3 questions I should confirm on customer calls.

💡

Pro tip: Take the three questions to five customers before you put whitespace on a slide.

Customer and Demand

5 prompts

Jobs from published research

11/30

✨ What it does

Lists up to eight sourced jobs to be done and marks your guesses Found, Partial or Not found. Ungrounded guesses stay tagged, not dressed up.

From published research, list jobs to be done for people who buy [CATEGORY] in [GEOGRAPHY]. My current guess: [PASTE JOB GUESSES]. For each job: a 15 word statement, the source that supports it, date and URL. Mark my guesses Found, Not found or Partial. If a guess is Not found, write [VERIFY] and do not dress it up as a job. Do not invent a persona. Do not invent a demand volume. Prefer surveys, ethnographies and regulator reports over blogs. Max 8 jobs.

💡

Pro tip: Paste the jobs your sales team already recites so you can see which ones sources actually support.

Buyer roles from public org charts

12/30

✨ What it does

Maps up to six buyer roles from job posts, case studies and analyst notes, labelling thin sources Weak. No committee is invented to look complete.

Who is typically involved in buying [CATEGORY] at [COMPANY TYPE] in [GEOGRAPHY]? Use public job posts, case studies and analyst notes. For each role: title variants, what sources say they care about, and a citation with URL. If a role is only mentioned in one thin blog, write [VERIFY] and label it Weak. Do not invent a buying committee. Do not invent a budget owner. Cap at 6 roles. End with 2 questions I should ask in discovery to confirm the map.

💡

Pro tip: Save the two discovery questions in your call doc before the next first meeting.

Switching triggers with sources

13/30

✨ What it does

Lists five switching triggers with a sourced line each, and tags vendor-only marketing as weak. Single comments are not enough without a second source.

What public sources say about why buyers switch away from [INCUMBENT TYPE] toward [ALTERNATIVE] in [CATEGORY]. Return 5 triggers, each with a quoted or closely paraphrased line, the source, date and URL. If you only find vendor marketing, say so and tag the trigger [VERIFY]. Do not invent a churn rate. Do not invent a trigger from a single Reddit comment unless you also find a second independent source. Prefer reviews, case studies and news interviews.

💡

Pro tip: Keep the quoted line in the brief so product hears the buyer's words, not yours.

Review theme extract

14/30

✨ What it does

Extracts six review themes with two quoted snippets and URLs each. Blocked sources are skipped, and no star average is invented.

Extract themes from public reviews of [PRODUCT OR CATEGORY] on sources you can open. Review sources I allow: [PASTE SOURCE TYPES OR URLS]. Return 6 themes with: theme name, 2 quoted snippets, URLs, and whether the theme is praise, complaint or mixed. If a source is blocked or empty, write [VERIFY] and skip it. Do not invent a review. Do not invent a star average or a volume of reviews. If you cannot quote two snippets for a theme, drop the theme.

💡

Pro tip: Name the review sites you can actually open so Perplexity does not cite a wall you cannot check.

Demand evidence without volumes

15/30

✨ What it does

Collects up to eight non-volume demand signals with dates and URLs, ranked by recency. TAM and user counts are refused.

Gather demand evidence for [PRODUCT IDEA] serving [AUDIENCE] in [GEOGRAPHY] that is not a market volume. Allowed evidence types: job posts, waitlists that are public, regulatory filings, search interest described without a fake number, and named tenders. For each item: what it is, date, URL, and why it counts as demand. If you cannot find a URL, write [VERIFY] and omit the item. Do not invent a TAM, a search volume, or a number of users. Max 8 items, ranked by how recent the source is.

💡

Pro tip: A dated tender or a job post is stronger than a blog that says demand is exploding.

Like these prompts? There are full tutorials behind them.

Learn the workflows, not just the prompts. Easy-to-follow tutorials inside AI Academy, and the library keeps growing.

Try AI Academy Free

Source Hygiene

5 prompts

Primary source for a statistic

21/30

✨ What it does

Traces a statistic to a primary source with a quoted line, date and method, or calls it unverifiable. Close numbers are not offered as substitutes.

Trace this statistic back to a primary source: [PASTE STATISTIC]. Where I saw it: [WHERE I SAW IT]. Find the original study, filing, dataset or official statement. Quote the relevant line, give the date and method in one sentence, and link it. If you cannot find the origin, write [VERIFY] and call it unverifiable. Do not invent a number that is close. Do not cite a blog that cites another blog as if it were primary. Note if a newer figure exists, with its URL.

💡

Pro tip: Paste the statistic word for word, including the year, or the trail goes cold.

Conflicting source table

22/30

✨ What it does

Tables at least three independent sources on one question, flagging shared press releases. If a third source is missing, the table stops.

On [RESEARCH QUESTION], find at least 3 independent sources and table what they claim. Columns: Source, Date, Claim in 20 words, Method if stated, URL, Independent of the others yes or no. If they share a press release, say so. If you cannot find 3, write [VERIFY] and stop rather than inventing a third. Do not invent a volume to break a tie. End with where they agree and where they diverge, in 60 words. Question: [RESEARCH QUESTION].

💡

Pro tip: Watch for three articles that all quote the same release. That is one source, not three.

Recency check on a brief

23/30

✨ What it does

Dates every claim in a pasted brief as Current, Stale or Unclear, and points to a newer URL when one exists. Claims are not refreshed by rounding.

Check the dates on this research brief. Brief: [PASTE BRIEF]. For each factual claim, find the source date and say Current, Stale, or Unclear. Stale means the source is older than [MONTHS] months and a newer source exists. Cite the newer URL if you find one. If you cannot date a claim, write [VERIFY]. Do not invent a newer number. Do not refresh a claim by rounding. Return a table plus a 40 word note on the stalest claim I should cut.

💡

Pro tip: Set months to 12 for fast categories and 24 for slow ones, then actually open the new URL.

Open versus gated sources

24/30

✨ What it does

Lists eight sources marked Open or Gated, with what the abstract still gives you. The last three open items are the ones to read first.

For [RESEARCH QUESTION], list 8 sources you would actually use, and mark each Open or Gated. Include title, date, URL, and what I can get from the open abstract if it is gated. If you cannot tell, write [VERIFY]. Do not invent a source behind a wall you cannot name. Do not invent a volume from an abstract. Prefer official and primary sources. End with the 3 open sources I should read first if I cannot get the gated ones.

💡

Pro tip: Read the three open sources before you wait on a gated PDF from a colleague.

Unverifiable claim dump

25/30

✨ What it does

Splits a draft into sourced and unverifiable sentences, moving unsourced volumes and rival lists into the dump. No citation is invented to save a line.

Audit this draft for claims that cannot be sourced. Draft: [PASTE DRAFT]. For each sentence that states a fact, try to find a citation. Return two lists: Sourced, with URL and date, and Unverifiable, with [VERIFY] on each line. Move any volume, share or competitor list with no source into Unverifiable. Do not invent a citation to save a line. Do not rewrite the draft yet. End with the 3 unverifiable lines that would do the most damage if they stayed.

💡

Pro tip: Cut the three damaging lines before the next stakeholder read, even if the rest of the dump can wait.

Go from copy-pasting to actually mastering AI.

AI Academy: hands-on tutorials on ChatGPT, Claude, Midjourney, and many other tools. New tutorials are added regularly.

Start Your Free Trial

Decision Briefs

5 prompts

One page market brief

26/30

✨ What it does

Writes a one page sourced brief with a six row source table, and turns unsourced sections into questions. Volumes and extra rivals are refused.

Write a one page market brief on [CATEGORY] for [DECISION] in [GEOGRAPHY]. Sections: What this is, Who buys, How it is changing, What is uncertain, What I should do next. Cite a URL after every factual sentence. If a section lacks a source, write [VERIFY] and keep it to questions. Do not invent a volume, a share, or a competitor I did not ask you to check. Keep the body under 350 words plus a 6 row source table with dates. Audience: [AUDIENCE].

💡

Pro tip: Hand the source table to the stakeholder with the brief, not as an optional appendix.

Size claim caution

27/30

✨ What it does

Interrogates pasted size claims for primary source, method and date, and refuses a blended number. The closing paragraph says why a volume may stay out.

I am tempted to put a size number in a brief on [CATEGORY] for [GEOGRAPHY]. Size claims I have seen: [PASTE SIZE CLAIMS]. For each, find the primary source, method, date, and whether it is TAM, spend, or something else. If the method is missing, write [VERIFY] and tell me not to use it. Do not invent a blended number. Do not invent a bottom-up count. End with a 50 word paragraph I can paste that explains why we are not putting a volume in this brief, plus the best single sourced figure if one survives.

💡

Pro tip: If method is missing, the number does not go on the slide, even if everyone else uses it.

Go or wait memo

28/30

✨ What it does

Writes a go, wait, or cannot say memo that tests your evidence against public sources. Thin evidence stays Cannot say instead of getting a fake volume to tip it.

Write a go or wait memo on [BET] in [CATEGORY] for [AUDIENCE]. Evidence I already have: [PASTE EVIDENCE]. Use public sources to test that evidence. Structure: What would have to be true, What sources support, What sources contradict, Open questions, Recommendation as Go, Wait, or Cannot say. Cite URLs. If evidence is too thin for a recommendation, write [VERIFY] and choose Cannot say. Do not invent a volume to tip the scale. Do not invent a competitor move I did not paste or find.

💡

Pro tip: Cannot say is a valid output. Bring it back with two sourced answers, not a hunch.

Interview questions from gaps

29/30

✨ What it does

Writes ten open interview questions grouped as confirm, probe gaps, and discover, without embedding unsourced statistics. One person is never asked for a market volume.

From this research, write 10 customer interview questions about [CATEGORY] for [CUSTOMER TYPE]. Research notes: [PASTE RESEARCH NOTES]. Group them: Confirm a sourced claim, Probe a [VERIFY] gap, and Discover something sources never mention. Each question is one sentence, open, and does not lead. If a sourced claim has no citation in my notes, move it to the gap group and tag [VERIFY]. Do not invent a statistic to embed in a question. Do not ask for a market volume from one human.

💡

Pro tip: Ask the gap questions first, while the person is fresh, then confirm the sourced claims.

Cited bullets for a meeting

30/30

✨ What it does

Turns research into eight slides with conclusion headlines, short bullets and a citation line on each. Unsourced volumes become questions, not numbers.

Turn this research into 8 slides of bullets for a [MEETING TYPE] with [AUDIENCE]. Research: [PASTE RESEARCH]. Each slide: conclusion headline, 3 bullets, and a short citation line with source and date under the bullets. If a headline needs a volume I cannot source, write [VERIFY] and change the headline to a question. Do not invent a competitor list. Do not invent a share. Keep bullets under 12 words. End with 3 questions the room should still ask.

💡

Pro tip: Put the citation under the bullets so a sceptic in the room can open it during the meeting.

Free tool

Prompt Optimizer

Turn a rough idea into a structured, professional AI prompt.

Try it free →

Frequently Asked Questions

It might. These prompts tell it to skip TAM, shares and user counts unless a primary source with a method is attached, and to tag gaps as [VERIFY]. If the method is missing, leave the number out of the brief.
Ask for title, publisher, date and URL after every factual sentence, as these prompts do. Prefer primary sources from the last 18 months. Click the links. A citation the page does not support is still unsourced.
Use Deep Research for the one page brief, the conflicting source table, and the go or wait memo. Use the lighter prompts for a single competitor page, a glossary term, or a statistic you already have in front of you.
Treat Yes as a claim, not as a trial. The grid prompt marks silent pages Unclear and tags them. Confirm a Yes in a product trial before it goes to sales. Marketing pages omit limits on purpose.
That is one source. The conflicting source table asks whether items are independent. If they share a release, say so. Find a filing, a dataset or a second primary page before you call it triangulated.

Prompts are the starting line. Tutorials are the finish.

A growing library of hands-on tutorials on ChatGPT, Claude, Midjourney, and many other AI tools. New tutorials are added regularly.

Start 7-Day Free TrialCancel anytime.
Explore AI for marketers