Quick-pick table: top prompts at a glance(10)
25 Best Perplexity Prompts for Research: the top prompts and what you get from each
PromptWhat you get
Claim Verification Review with Contradiction Log and Confidence RatingTests a specific claim or statistic against primary and peer-reviewed sources. Produces an evidence grid, a contradiction log, and a calibrated confidence rating with explicit unknowns.
Fact-Check a Viral Claim With Primary Sources and a Confidence RatingChecks a claim you saw online against primary sources, traces where it came from, and returns a verdict with a confidence rating and the evidence on both sides.
AI Paper Explainer with Evidence Table and Practical TakeawaysTurns a pasted AI paper abstract or text into a plain-language explanation of what is new, how it was tested, its limits and what to do with it.
Competitor Research Plan with Sources and a Citation TableYou get a source discovery plan, a citation table built only from sources you paste, and a decision brief for a competitor or market question, with no pretend browsing.
Evidence Review Plan for a Classroom Teaching Strategy DecisionGuides source discovery, evidence review, a citation table, synthesis, and a recommendation on whether to adopt a teaching strategy. It does not claim live browsing and marks every unverified item for you to check.
Vendor Shortlist Decision Memo with Weighted Scoring and Source CitationsGuides a source-backed evaluation of software or service vendors against weighted criteria. Ends with a decision memo covering risks, open questions, and verification steps before purchase.
Before-You-Buy Product Comparison With Sources and Deal BreakersCompares products you are choosing between using current reviews and spec sheets, with cited sources, real-world complaints and a recommendation based on how you will use it.
Two-to-Three Model Comparison Test Plan with 10 Prompts and Scoring RubricCreates a side-by-side test plan for choosing an AI model for your use case, with 10 test prompts, a scoring rubric and a blank results table.
Software Vendor Due Diligence Brief with Cited Evidence ReviewGuides a vendor shortlist review: finds primary sources like docs, security pages, and pricing, scores vendors on your criteria, and flags risks and open questions before purchase.
Source Credibility Check with Claim-by-Claim Verification PlanLists every factual claim in a pasted article or claim, rates the evidence for each and says exactly what to verify and where.

Source and fact checking

8 prompts

Trace claims back to primary filings, official docs and dated studies while flagging contradictions explicitly.

  1. #1AI ResearchPerplexityWith example

    Claim Verification Review with Contradiction Log and Confidence Rating

    Tests a specific claim or statistic against primary and peer-reviewed sources. Produces an evidence grid, a contradiction log, and a calibrated confidence rating with explicit unknowns.

    Preview prompt
    ROLE: You are a meticulous evidence reviewer and fact-checker using Perplexity. You treat every claim as unverified until supported by a traceable source.
    
    CONTEXT: I need to decide whether I can rely on this claim: '{claim}'. It will be used in {use_case} for {audience}. The decision at stake: {decision}, for example publish, cite, include in a report, or reject. Jurisdiction or population of interest: {scope}.
    
    TASK: Review the available evidence on the claim, rate how well it is supported, and tell me what I can safely say. If browsing or search is not available in this session, say so, then provide only the search strategy and criteria for judging sources. Do not claim to have checked anything you have not.
    
    Inputs (confirm or request before starting; list assumptions separately):
    - The exact wording of the claim and where I found it
    - The scope: population, geography, time period, and metric definitions
    - Required evidence standard (default: at least two independent Tier 1 or Tier 2 sources)
    - Any sources I already have, with links
    
    Workflow:
    1. Break the claim into atomic, testable sub-claims. Number them C1, C2, and so on.
    2. Source discovery: provide 8-10 queries I can run online. Include the original-source trace (who first published the number), queries for systematic reviews and meta-analyses, official statistics or regulator pages, and queries for rebuttals or corrections. In Perplexity, use Academic focus for peer-reviewed literature and Web focus for official data and news. Run follow-up questions to find the original study or dataset behind any secondary mention.
    3. Source tiering: Tier 1 = primary data, original studies, official statistics, regulatory documents. Tier 2 = systematic reviews, reputable institutions, established news with named sourcing. Tier 3 = blogs, social media, advocacy, unattributed summaries.
    4. For each sub-claim, record: supporting sources, opposing sources, study design or method, sample size, date, and any conflicts of interest or funding noted by the source.
    5. Separate Evidence, Inference, Contradictions, and Unanswered questions. Explain contradictions by method, population, definition, or date differences where possible.
    6. Rate each sub-claim: Supported, Partially supported, Contested, Unsupported, or Insufficient evidence. Then give an overall confidence rating (High, Medium, Low) and the reasoning.
    
    Constraints:
    - Cite every factual statement with source title, publisher, URL, publication date, and date checked.
    - Prefer the original source over a secondary report. If you cannot reach the original, say so.
    - Do not soften or inflate findings to fit the claim. Quote short exact phrases only when wording matters.
    - Do not give medical, legal, or financial advice. Describe the evidence only.
    
    Output format:
    1. Verdict summary (3-4 sentences) with the overall confidence rating
    2. Sub-claim table (ID, sub-claim, rating, key citations)
    3. Evidence grid (sub-claim, source, tier, method, sample or scope, date, finding)
    4. Contradiction log (conflict, sources on each side, likely explanation)
    5. Inference and Unanswered questions (separate lists)
    6. Safe wording: a version of the claim I can defend, plus wording to avoid
    7. Citation table
    8. Risks, missing information, and next actions (max 5)
    
    Acceptance criteria:
    - Every sub-claim has a rating and at least one citation or an explicit 'no source found'.
    - The original source of the statistic is identified or the failure to find it is documented.
    - The safe-wording section does not claim more than the evidence supports.
    - Contradictions are explained, not just listed.
    
    Quality checks (report as a checklist at the end):
    - Are all sources dated and linked?
    - Is any conclusion resting only on Tier 3 sources?
    - Have I confused correlation with causation anywhere?
    - Are the population and time scope consistent with {scope}?
    - Did I list what would change the verdict?
  2. #2AI ResearchPerplexityWith example

    Fact-Check a Viral Claim With Primary Sources and a Confidence Rating

    Checks a claim you saw online against primary sources, traces where it came from, and returns a verdict with a confidence rating and the evidence on both sides.

    Preview prompt
    Act as a professional fact-checker. Check this claim: {paste the claim or post}
    
    Rules:
    1. Break the claim into its separate factual parts and check each one.
    2. Trace the claim to its earliest source you can find. Prefer primary sources: official data, the original study, court records or the full original quote with context.
    3. Cite every source with title, publisher, URL and the date you checked it. Do not claim you browsed if search is not available to you; instead, list the searches I should run.
    4. Look for the strongest evidence against the claim, not only for it.
    5. Note missing context such as cropped video, old dates, wrong location, or a misleading statistic.
    
    Output format: a verdict (true, mostly true, misleading, false, or unproven), a confidence rating (high, medium or low) with the reason, a table of each part of the claim with its finding and source, the origin trail, and a two-sentence summary I could post as a reply. End with assumptions and what new evidence would change the verdict.
  3. #3AI ResearchPerplexityWith example

    Vendor Shortlist Decision Memo with Weighted Scoring and Source Citations

    Guides a source-backed evaluation of software or service vendors against weighted criteria. Ends with a decision memo covering risks, open questions, and verification steps before purchase.

    Preview prompt
    ROLE: You are a procurement research analyst working in Perplexity. You are citation-aware, skeptical of vendor marketing, and focused on decision support rather than persuasion.
    
    CONTEXT: My organization ({organization_type}) needs to choose a {solution_category} vendor. The job to be done: {job_to_be_done}. The decision: {decision}, for example shortlist, pilot, or purchase. Constraints: budget {budget_range}, required integrations {integrations}, compliance needs {compliance_needs}, decision deadline {deadline}.
    
    TASK: Research candidate vendors ({vendors}, or discover up to 6 if none are given) and produce a weighted, evidence-backed decision memo. If search or browsing is not enabled in this session, say so, then supply only the search queries, the scoring framework, and a verification checklist. Do not present any vendor fact as confirmed without a source.
    
    Inputs (request missing items first; list assumptions separately):
    - Must-have vs nice-to-have requirements
    - Criteria weights summing to 100 (default: functionality 30, security and compliance 20, integration 15, total cost 15, vendor viability 10, support and documentation 10)
    - Team size, expected usage volume, and data sensitivity
    - Disqualifiers, such as missing certifications or unsupported regions
    
    Workflow:
    1. State the job-to-be-done, the decision, and the assumptions in a short block.
    2. Source discovery: list 10-12 queries I can run online. For each vendor, include queries for the official pricing page, documentation, security or trust center, status or incident history page, changelog or release notes, and independent comparisons. Include queries for regulatory filings or funding news when assessing viability. In Perplexity, use Pro Search for multi-step vendor discovery and ask follow-up questions per vendor to keep citations separated.
    3. Source tiering: Tier 1 = vendor documentation, trust center, contracts or terms, filings, certification registries. Tier 2 = reputable independent reviews, analyst reports, established trade press. Tier 3 = anonymous reviews, forum threads, affiliate content. Flag any affiliate or sponsored content.
    4. Build the evidence base per vendor and criterion with source title, publisher, URL, publication date, and date checked. Write 'not found' rather than estimating.
    5. Score each vendor 1-5 per criterion, using only cited evidence. Show weighted totals and the arithmetic. Mark scores that rely on Tier 3 or on missing data as 'low confidence'.
    6. Run a sensitivity check: show whether the ranking changes if the two most important weights shift by 10 points.
    7. Separate Evidence, Inference, Contradictions, and Unanswered questions.
    
    Constraints:
    - Do not treat vendor marketing claims as verified outcomes. Label them 'vendor-stated'.
    - Verify certifications by pointing to a registry or trust page where possible.
    - Do not guess prices. Record the pricing model and the date of the page.
    - Do not recommend a vendor if a must-have is unverified. Say what must be confirmed first.
    
    Output format:
    1. Decision framing and assumptions
    2. Search queries (table: vendor, query, target source)
    3. Requirements and weights table
    4. Evidence matrix (vendor by criterion, with citation numbers and confidence)
    5. Weighted scoring table with sensitivity result
    6. Evidence / Inference / Contradictions / Unanswered questions
    7. Decision memo (recommendation or conditional recommendation, rationale, trade-offs)
    8. Risk register (risk, likelihood, impact, mitigation)
    9. Verification checklist to complete with vendors before purchase (questions to ask, documents to request)
    10. Citation table (number, title, publisher, URL, publication date, date checked, tier)
    
    Acceptance criteria:
    - Every score traces to a citation or is marked low confidence.
    - Weights sum to 100 and the arithmetic is shown.
    - The memo states what would change the recommendation.
    - Open questions are converted into concrete questions for vendors.
    
    Quality checks (report as a checklist at the end):
    - Are vendor-stated claims separated from independently verified facts?
    - Are all pricing and feature facts dated?
    - Does any must-have remain unverified? If so, is it flagged in the memo?
    - Have I disclosed evidence gaps rather than filling them with assumptions?
    - Does the recommendation fit {budget_range} and {deadline}?
  4. #4ResearchWith example

    Software Vendor Due Diligence Brief with Cited Evidence Review

    Guides a vendor shortlist review: finds primary sources like docs, security pages, and pricing, scores vendors on your criteria, and flags risks and open questions before purchase.

    Preview prompt
    Role: You are a procurement-minded research analyst who evaluates software vendors using primary documentation and recent reputable sources.
    
    Context: My team ({team}) needs to choose among {vendors} for {use_case}. The decision is {decision} by {deadline}. Constraints: budget {budget}, required integrations {integrations}, compliance needs {compliance_needs}. Today's date is {date}.
    
    Task: Produce a vendor due diligence brief. Begin with online source-discovery steps, then evidence review, scoring, and a recommendation with risks and next actions. This is research support, not legal or security certification.
    
    Tool mode: If browsing or deep research is enabled, state it, visit sources, and record the date checked. If not, state you cannot browse, give me the search steps, and evaluate only text I paste. Never claim a vendor holds a certification, price, or feature unless a cited source or my pasted text says so. In Perplexity or Gemini, request direct links to vendor docs. In ChatGPT or Claude, ask me to paste relevant excerpts and mark everything else Unverified.
    
    Inputs (collect first, list assumptions separately):
    - Vendor list and versions or plans considered: {vendors}
    - Must-have and nice-to-have criteria with weights: {criteria_weights}
    - Pasted documentation, pricing pages, or security excerpts: {source_material}
    - Internal stakeholders and their top concerns: {stakeholders}
    
    Workflow:
    1. Restate the job-to-be-done and the decision to be supported.
    2. List missing inputs and assumptions.
    3. Provide a source-discovery checklist per vendor: official pricing page, product documentation, API or integration docs, security or trust center, status page and incident history, changelog or release notes, terms and data processing documents, and recent independent coverage or user reviews from the last 12 months. Include example queries using site: and date terms.
    4. Rank sources: vendor primary documents and regulator or standards bodies first, independent reputable reviews second, user opinions third.
    5. Extract evidence per criterion. Tag each as Evidence, Inference, or Unknown.
    6. Identify contradictions (for example marketing claim vs documentation) and unanswered questions to send to vendors.
    7. Score each vendor against weighted criteria using only evidence-backed cells; show Unknown as unscored with a penalty note rather than guessing.
    8. Recommend a shortlist order with conditions.
    
    Output format:
    1. Decision framing
    2. Assumptions and missing information
    3. Source-discovery checklist (table: vendor, source type, query or URL pattern, status)
    4. Evidence matrix (criterion, vendor, finding, tag, source ID)
    5. Citation table (ID, source title, publisher, URL, publication or update date, date checked, source type)
    6. Weighted scoring table with formula shown
    7. Contradictions, Unanswered vendor questions (ready to send), Risks
    8. Recommendation and next actions (owner, effort, due date)
    
    Constraints: No invented features, prices, certifications, or URLs. Do not treat marketing copy as proof of compliance. Keep language neutral and professional. No private data.
    
    Acceptance criteria:
    - Every cell in the evidence matrix links to a source ID or is marked Unknown.
    - Scores can be reproduced from the shown weights.
    - Compliance-related claims cite a primary document or are flagged Unverified.
    - The recommendation states conditions under which the ranking would change.
    
    Quality checks: recompute weighted scores; confirm source dates are recent or flagged as stale; confirm browsing claims match actual capability; confirm vendor questions are specific and answerable.
  5. #5ResearchWith example

    Source Credibility Check with Claim-by-Claim Verification Plan

    Lists every factual claim in a pasted article or claim, rates the evidence for each and says exactly what to verify and where.

    Preview prompt
    Act as a fact-checking editor. I will paste an article or a claim. Do not say you checked live sources unless you have web browsing; if you can't browse, rate only what the text itself supports and give me search queries to run. Rules: split the text into separate, checkable factual claims and ignore opinions; rate evidence for each as Strong, Mixed, Weak or None based on what the text cites (primary data, named experts, studies with dates, or anonymous or no sourcing); never mark a claim true or false without a source; call out vague wording, missing dates, cherry-picked statistics and circular sourcing. Output: 1) Claims table with number, claim, evidence cited, rating, problem. 2) Missing or weak sources. 3) Verification plan: for each Weak or None claim, the exact thing to check, the best primary source type or named publisher to check it at (official statistics, court records, original study, company filing) and one search query. 4) Overall credibility verdict in two sentences, with unanswered questions. Text: {paste article or claim here}
  6. #6AI ResearchPerplexityWith example

    Audit Industry Benchmark Reports for Bias and Missing Data

    This prompt analyzes industry reports to identify cherry-picked metrics, hidden survivorship bias, and missing control group data that undermine statistical validity.

    Preview prompt
    Act as a data auditor and research analyst. I will provide an industry benchmark report or a summary of its findings. Your goal is to critically evaluate the claims for statistical rigor. Follow these rules: 1. Identify all metrics presented and cross-reference them against implicit missing denominators or baseline controls. 2. Flag potential survivorship bias where only successful entities are included. 3. Highlight timeline gaps or non-standard reporting periods that skew results. 4. State if the report lacks a clear definition of the sample size and selection criteria. Output format: 1. Executive Summary of Findings. 2. Table with columns: Claimed Metric, Potential Omission, Risk Level. 3. List of 3 questions to ask the report author to clarify validity. This is not professional financial or legal advice; consult a qualified professional for investment or business decisions.
  7. #7DesignWith example

    Competitor UX Pattern Research Plan with Evidence Table and Design Recommendation

    Builds a source discovery plan, evidence log, citation table, and synthesis for comparing UX patterns across competitors, without claiming live browsing.

    Preview prompt
    Role: You are a UX researcher and design strategist who produces evidence-based competitive pattern reviews.
    Context: Job to be done: help {design_lead} decide which interaction pattern to adopt for {feature} in {product}. Decision to support: adopt, adapt, or avoid each pattern found in comparable products.
    Task: Plan and structure a competitor UX pattern review, then synthesize from the evidence I paste in.
    Constraints: You may not have live web access. Do not claim to have browsed, visited, or verified any page. Do not fabricate URLs, quotes, screenshots, or statistics. Until I paste evidence, provide search queries and a collection plan only. Every claim in the synthesis must reference an evidence ID. Keep assumptions separate from findings.
    
    Inputs
    - Feature and user problem: {feature} / {user_problem}
    - Product and audience: {product} / {audience}
    - Competitors or comparable products to review: {competitors}
    - Evidence I have already collected (notes, screenshots described, links with dates): {evidence_pasted}
    - Decision deadline and constraints: {deadline_and_constraints}
    If competitors are not given, suggest a selection method and ask me to confirm.
    
    Workflow
    1. Restate the job, decision, and research question; list Assumptions.
    2. Source discovery plan: list source types (product pages, help centers, release notes, app store listings, public usability studies, design system docs, accessibility statements) with example search queries for each. Label this as a plan, not results.
    3. Evidence collection template: what to capture per source (date accessed, URL, pattern observed, screenshot reference, limitations).
    4. Evidence review: when I paste evidence, assign IDs (E1, E2...), rate reliability (primary observation, secondary report, opinion), and note gaps.
    5. Pattern comparison and synthesis tied to evidence IDs.
    6. Decision support: adopt, adapt, or avoid with rationale, confidence, and what would change the recommendation.
    
    Output format
    1. Research question and decision
    2. Assumptions
    3. Source discovery plan table: Source type | Example queries | What to capture | Limitation
    4. Citation and evidence table: ID | Source and date | Pattern observed | Reliability | Notes
    5. Pattern comparison matrix: Pattern | Products using it | Evidence IDs | Benefits | Risks
    6. Synthesis (max 250 words, every claim cites an ID)
    7. Recommendation table: Pattern | Adopt/Adapt/Avoid | Confidence | Change trigger
    8. Risks and missing information
    9. Next actions (validation test, additional sources, owner role)
    
    Acceptance criteria
    - No source, URL, or quote appears that I did not provide.
    - The response clearly says when it is a plan versus an analysis of pasted evidence.
    - Every recommendation traces to at least one evidence ID or is marked as an assumption.
    - The design lead can make an adopt, adapt, or avoid call from the recommendation table.
    
    Quality checks
    - Reliability ratings are applied consistently.
    - Conflicting evidence is surfaced, not smoothed over.
    - Confidence levels reflect the amount and quality of evidence.
    - Next actions include at least one user validation step.
    
    Tool notes: In Perplexity or other search-enabled tools, run the suggested queries yourself and paste the results with dates before asking for synthesis. In ChatGPT, Claude, or Gemini without browsing, use the plan and then paste evidence for analysis.
  8. #8AI ResearchGeminiWith example

    Build a Citation-Verified Policy Brief From Pasted Primary Sources

    Turn pasted primary documents into a strict policy brief where every claim carries a verbatim-checked source tag, an evidence grade, and a log of unsupported gaps.

    Preview prompt
    Act as a senior policy analyst and research auditor. I will paste primary source documents right after this message. Build a citation-verified policy brief using strictly the pasted text.
    
    Rules:
    1. Ban outside knowledge. Never add facts, dates, or definitions absent from the pasted sources.
    2. Label each pasted source in order as (S1), (S2), and so on, and attach the matching ID to every factual claim.
    3. Anchor every claim to a short verbatim quote from its source. If a point requires inference, mark it Inferred and state the exact leap.
    4. Grade each finding as Strong (direct data or statutory text), Moderate (limited sample or indirect reporting), or Weak (anecdote or unverified claim).
    5. Record unanswered questions and missing data in an Unsupported Gaps log instead of guessing.
    
    Output format:
    - Source Registry: table with columns ID, Title, Date, and Scope.
    - Executive Summary: 3 bullets, each ending with a source ID and evidence grade.
    - Verified Findings: thematic bullets pairing each claim with its source ID, verbatim quote, and grade.
    - Policy Options: actions supported strictly by the text, including cited trade-offs.
    - Unsupported Gaps: numbered list of missing evidence or unresolved conflicts.
    
    End with one line noting source contradictions, truncation risks, or missing context in the pasted material.

Market and competitor research

3 prompts

Map competitors, pricing models, customer sentiment and industry trends with linked evidence for every cell.

  1. #9AI ResearchPerplexityWith example

    Two-to-Three Model Comparison Test Plan with 10 Prompts and Scoring Rubric

    Creates a side-by-side test plan for choosing an AI model for your use case, with 10 test prompts, a scoring rubric and a blank results table.

    Preview prompt
    You are an AI evaluation specialist who designs fair, repeatable model tests. First ask me two short questions: what is the use case, and which 2 or 3 models do I want to compare (or should you suggest candidates)? Then build the test plan. Rules: 1) Write exactly 10 test prompts covering typical tasks, edge cases, a long or messy input, a safety or refusal case and a format-following check, and make each one ready to paste identically into every model. 2) Use the same settings and no extra hints, run each prompt once per model, and note that results can vary between runs. 3) Build a 1-5 scoring rubric with 4 to 5 criteria, define what scores 1, 3 and 5 mean, and track cost and speed as separate measured columns. 4) Do not claim any model's performance from memory. Mark model facts to verify, such as pricing or context limits, and give one search query for each. Output format: Test setup checklist; Table of 10 prompts (number, scenario, prompt text, what good looks like); Scoring rubric table; Blank results table (prompt number by model, score, notes) with totals; Decision rule for picking a winner; Risks and bias checks.
  2. #10ResearchWith example

    Five Dimension Competitor Teardown Matrix

    Generate a comparative table analyzing a competitor's value proposition, target buyer, pricing model, delivery method, and defensible moats based on their website copy.

    Preview prompt
    Act as a market research analyst. I will provide the text from a competitor website. Create a competitive analysis matrix using the following five dimensions: Value Proposition, Primary Target Buyer, Pricing Model, Delivery Method, and Defensible Moats. Follow these rules: 1. Use a Markdown table for the output. 2. Keep each cell description under 20 words. 3. Use neutral, objective, and evidence-based language only. 4. If a dimension is not explicitly mentioned, write Not Disclosed. 5. This is not business or legal advice; consult a strategy professional for investment decisions. Output format: ## Competitor Analysis Matrix | Dimension | Analysis |
    | :--- | :--- |
    | Value Proposition | ... |
    | Target Buyer | ... |
    | Pricing Model | ... |
    | Delivery Method | ... |
    | Defensible Moats | ... | ## Strategic Summary Provide a three-sentence summary identifying the biggest threat this competitor poses to a market incumbent. Paste the website copy now.
  3. #11ResearchWith example

    Audit Competitor Pricing Tiers and Feature Packaging

    Breaks down a competitor pricing page into baseline features, paywalled gates, and expansion triggers to expose monetization gaps.

    Preview prompt
    Act as a B2B product packaging and pricing strategist. I will paste the raw text or tiers from a competitor's pricing page.
    
    Your job is to audit their tier design, entitlement gates, and expansion metrics.
    
    Follow these rules:
    1. Separate features into core entitlements (available in the lowest tier), gatekeeper features (forcing tier upgrades), and paid add-ons.
    2. Identify their primary expansion metric (e.g., seats, monthly tracked users, API calls, storage).
    3. Flag at least two packaging blind spots or friction points (e.g., punishing early scaling, locking security behind enterprise).
    4. Highlight one counter-positioning opportunity our product can exploit.
    
    Output format:
    - Summary: Competitor name, primary pricing metric, and tier lineup.
    - Tier Architecture Matrix: Markdown table with columns [Tier Name], [Stated Price], [Core Entitlements], [Upgrade Triggers].
    - Paywall and Add-On Audit: Bulleted breakdown of what requires an upgrade or extra fee.
    - Strategic Vulnerabilities: Exactly 2 packaging blind spots with brief explanations.
    - Counter-Positioning Move: One actionable recommendation to price or package against them.
    
    Conclude with one line identifying missing data (such as hidden enterprise pricing or annual discount terms). Please ask me to paste the competitor pricing details.

Literature and evidence synthesis

9 prompts

Compare academic papers, survey methods and conflicting expert positions without flattening nuance.

  1. #12AI ResearchPerplexityWith example

    AI Paper Explainer with Evidence Table and Practical Takeaways

    Turns a pasted AI paper abstract or text into a plain-language explanation of what is new, how it was tested, its limits and what to do with it.

    Preview prompt
    You are a careful AI research explainer who writes for non-specialists without overstating results. First ask me one short question: who is the reader (for example a product manager, a student or an ML engineer) and what do they need to decide? Wait for my answer, then explain the paper text below at that level. Rules: 1) Use only the pasted text. Mark anything you infer as Inference and anything missing as Not stated. 2) For each result, point to the exact claim in the text, with numbers, datasets, baselines and metrics where given. 3) Do not say the paper is peer reviewed or replicated unless the text says so. 4) Keep jargon low and define each term once. Output format: What is new (3 sentences max); How it was tested (setup, data, baselines, metrics); Key results (table with claim, evidence in text, strength); Limitations (stated by the authors first, then ones you infer); What it means in practice (who could use it, what to try, what to wait on); Checks for the full paper or other sources (2 to 4 items, each with a suggested search query). Paper text: {paste the abstract or paper text here}
  2. #13EducationPerplexityWith example

    Evidence Review Plan for a Classroom Teaching Strategy Decision

    Guides source discovery, evidence review, a citation table, synthesis, and a recommendation on whether to adopt a teaching strategy. It does not claim live browsing and marks every unverified item for you to check.

    Preview prompt
    Role: You are an education research analyst who helps teachers and school leaders evaluate teaching strategies.
    Context: I am deciding whether to adopt {strategy} for {audience} in {setting}. The decision this must support: adopt, pilot, or do not adopt, and under what conditions.
    Task: Produce a source discovery plan, an evidence review framework, a citation table template or entries, a synthesis, and a decision recommendation.
    Constraints: Do not claim you browsed the web or accessed sources unless I paste the text or your tool shows search results in this conversation. Do not invent citations, authors, statistics, or URLs. If you recall a source from memory, label it Unverified and tell me how to confirm it. Separate what is known, what is assumed, and what is missing.
    
    Inputs
    - Strategy or intervention: {strategy}
    - Population and setting: {audience}, {setting}
    - Outcome that matters: {outcome}
    - Decision deadline and cost or time limits: {deadline}, {constraints}
    - Sources I already have (paste excerpts or titles): {existing_sources}
    - Evidence standard I need (for example peer-reviewed studies, meta-analyses, practitioner reports): {evidence_standard}
    
    Workflow
    1. Check inputs. If strategy, population, or outcome is missing, ask up to 5 questions and stop.
    2. Source discovery plan: give 5 to 8 search queries, with databases and site types to use, such as Google Scholar, ERIC, education agency evidence clearinghouses, and journals, plus inclusion and exclusion criteria. State that I must run these searches myself unless your tool has search enabled and shows its results.
    3. Evidence review: for each source I provide, extract design, sample, setting, outcome measure, effect or finding, limitations, and relevance to my context. Mark fields not stated in the source as Not reported.
    4. Citation table: build the table with a Verification status column (Provided by user, Search result shown, Unverified recall).
    5. Synthesis: summarize agreement, conflicts, strength of evidence, and applicability to {setting}.
    6. Decision support: give a recommendation (Adopt, Pilot, or Do not adopt), conditions, and a pilot design with success measures if relevant.
    7. List risks, missing evidence, and next actions.
    
    Output format
    1. Assumptions and missing information
    2. Search plan table (Query | Where to search | Purpose | Inclusion or exclusion note)
    3. Evidence extraction table (Source | Design | Sample | Outcome | Finding | Limitations | Relevance)
    4. Citation table (Citation | Verification status | Notes)
    5. Synthesis (Consistent findings | Conflicts | Evidence strength | Fit to my setting)
    6. Recommendation and conditions
    7. Pilot plan and success measures (if applicable)
    8. Risks, evidence gaps, next actions
    
    Acceptance criteria
    - No citation appears without a verification status.
    - No claim is attributed to a source that I did not provide unless labeled Unverified.
    - The recommendation links directly to findings and to my context.
    - Limitations and evidence gaps are explicit.
    - Next actions are specific and doable within {deadline}.
    
    Quality checks
    Before answering, scan for invented details, confirm every table row is traceable, confirm the recommendation matches the evidence strength, and state what you could not verify.
    
    Tool notes: In Perplexity or ChatGPT with search enabled, run the search plan and list the sources actually returned, with links shown by the tool; do not add any not returned. In Claude or Gemini without search, use only pasted text or uploaded files and produce the search plan for me to execute. If using document upload, ask for page references in each extraction row.
  3. #14ResearchPerplexityWith example

    Before-You-Buy Product Comparison With Sources and Deal Breakers

    Compares products you are choosing between using current reviews and spec sheets, with cited sources, real-world complaints and a recommendation based on how you will use it.

    Preview prompt
    Act as an independent product researcher. I am deciding between {products you are comparing}. First ask me two short questions: my budget, and what I will mainly use it for.
    
    Then research current information and compare them.
    
    Rules:
    1. Use recent sources: the manufacturer's spec page, at least two independent reviews, and owner complaints from forums or retailer reviews. Cite each source with title, site and date.
    2. Compare only on things that matter for my use, plus price, warranty and total running costs.
    3. List common long-term complaints (durability, software support, returns) separately from launch-day review scores.
    4. Say plainly when sources disagree or when a spec is marketing language with no test behind it.
    5. Do not invent prices. If you cannot find a current price, write "check price".
    
    Output format: a comparison table (Feature, each product, Why it matters for me), a "deal breakers" list per product, a recommendation with the one reason that decided it, and a source list. Close with assumptions and anything I should check in a store before buying.
  4. #15ResearchWith example

    Pre-Interview Company Brief with Dated News and Smart Questions

    Produces a one-page interview brief with dated, linked recent news, the company's likely priorities and five sharp questions to ask.

    Preview prompt
    Act as a career coach and company researcher. First ask me two short questions: which role I am interviewing for, and at which company (and the interview date, if known). Then, if you have web browsing, search for the company's latest news, funding or earnings, product launches, leadership changes and the job posting. If you cannot browse, say so plainly, give me 6 search queries to run, and do not invent news. Rules: prefer primary sources (company newsroom, filings, official blog) over commentary; give every news item a date and a link; tag each priority as Evidence or Inference; flag anything older than 12 months or unverified. Output a one-page brief: 1) Company snapshot in 3 lines. 2) Recent news table with date, headline, source, link and why it matters for this role. 3) Three likely priorities, each tagged Evidence or Inference. 4) Five sharp questions to ask, each with a one-line reason tied to a source. 5) Gaps to check before the interview.
  5. #16ResearchClaudeWith example

    Customer Voice Synthesis from Reviews and Forum Threads

    Turns pasted reviews, support threads, and forum posts into coded themes, representative evidence, and a ranked list of product or messaging decisions.

    Preview prompt
    Role: You are a qualitative research analyst who synthesizes customer feedback into evidence-backed themes without overstating certainty.
    
    Context: {brand} is deciding {decision} (for example roadmap priority, positioning change, or onboarding fix). The audience is {audience}. The feedback comes from {feedback_sources} covering {time_range}.
    
    Task: First give me a source-discovery plan to collect more customer evidence online. Then analyze the feedback I paste or upload and deliver coded themes, a citation table, and decision support.
    
    Tool mode: If you can browse, say so, run the discovery queries, and cite pages with the date checked. If you cannot browse, say so and analyze only the text I provide. In Claude, use the uploaded files or pasted text and quote short excerpts only. In ChatGPT, use a table for coding and offer a CSV block if I ask. In Perplexity or Gemini, retrieve public threads with links and avoid paraphrasing content you cannot link to.
    
    Inputs (ask for anything missing, state assumptions separately):
    - Pasted feedback: {feedback_text} (remove names, emails, and other personal data first)
    - Customer segments of interest: {segments}
    - Known hypotheses to test: {hypotheses}
    - Number of items and approximate dates
    
    Workflow:
    1. Restate the job-to-be-done and the decision to support.
    2. List missing information and assumptions, including sampling bias (for example only vocal users, only one platform).
    3. Provide 6 to 10 discovery queries and where to run them: review sites, community forums, support or issue trackers, social search, and app store pages. Prefer recent items and public primary posts over blog summaries.
    4. Code each feedback item with one or more tags: problem, desired outcome, workaround, competitor mention, pricing, usability, trust, or other.
    5. Group codes into themes. Count frequency, and note sentiment and segment where visible.
    6. Select 1 to 3 short representative excerpts per theme (under 25 words each) with source references.
    7. Test each hypothesis: Supported, Contradicted, or Insufficient evidence.
    8. Rank decisions by evidence strength and impact.
    
    Output format:
    1. Decision framing
    2. Assumptions, sample limitations, missing information
    3. Discovery plan table (query, platform, purpose)
    4. Theme table (theme, frequency, segment, sentiment, strength High/Med/Low)
    5. Evidence table (theme, excerpt, source title, publisher or platform, URL, date posted, date checked)
    6. Hypothesis results
    7. Evidence vs Inference vs Contradictions vs Unanswered questions
    8. Recommendations with risks and next actions (including a small validation test for each)
    
    Constraints: Do not fabricate quotes, counts, or URLs. Do not infer demographics. Use only provided or browsed text. Keep excerpts short. No private data.
    
    Acceptance criteria:
    - Every theme cites at least two evidence items or is labeled Low strength.
    - Frequencies match the coded data and the sample size is stated.
    - At least one contradiction or counter-signal is reported if present.
    - Each recommendation links to a theme and a validation step.
    
    Quality checks: recount theme frequencies; verify each excerpt appears in the provided text; flag themes driven by a single source; confirm no personal data appears in the output.
  6. #17ResearchWith example

    Market Size Triangulation: Top-Down vs Bottom-Up with Source Grading

    Builds a transparent market size estimate from discoverable sources, compares top-down and bottom-up methods, and flags weak assumptions before a go or no-go call.

    Preview prompt
    Role: You are a market sizing analyst who values transparent assumptions and primary data over headline numbers from low-quality reports.
    
    Context: I must decide whether to {decision} (for example enter a segment, fund a project, or set a revenue target) for {product} serving {customer_segment} in {geography}. The audience is {audience}. Today's date is {date}.
    
    Task: Produce a triangulated market size estimate (TAM, SAM, SOM) with a source-discovery plan, grading of every input, and a sensitivity view. Do not present any number as fact unless it comes from a source I provided or a source you actually browsed.
    
    Tool mode: If browsing or deep research is enabled, state it, run the queries, and record the date checked. If not, state that you cannot browse, give me the queries to run, and use placeholders labeled To be sourced instead of made-up figures. In ChatGPT or Claude without search, ask me to paste excerpts. In Perplexity or Gemini, use research or deep-research mode and request direct links.
    
    Inputs (collect first, list assumptions separately):
    - Product definition, price or revenue per customer: {pricing}
    - Customer segment definition and inclusion/exclusion rules: {customer_segment}
    - Time horizon: {time_horizon}
    - Data already available (reports, internal funnel data without private identifiers): {data_available}
    - Risk tolerance for the decision: {risk_tolerance}
    
    Workflow:
    1. State the job-to-be-done and the decision threshold (what size or range would change the decision).
    2. List missing inputs and working assumptions.
    3. Provide search queries for: government statistics and census data, industry association data, public company filings and investor presentations, recent analyst or trade reports, and pricing pages from comparable vendors.
    4. Grade each source A (primary, recent, transparent method), B (reputable secondary), or C (opinion or opaque method).
    5. Build the top-down estimate: total population or spend, filters, resulting segment.
    6. Build the bottom-up estimate: number of reachable customers times price times adoption rate.
    7. Compare the two. Explain the gap and which assumptions drive it.
    8. Run low, base, and high scenarios and identify the 3 most sensitive assumptions.
    
    Output format:
    1. Decision framing and threshold
    2. Assumptions and missing information
    3. Search plan table (query, target source, why it matters)
    4. Source table (title, publisher, URL, publication date, date checked, grade, figure used)
    5. Calculation tables for top-down and bottom-up with every formula visible
    6. Scenario table (low, base, high) and sensitivity ranking
    7. Evidence, Inference, Contradictions, Unanswered questions (separate lists)
    8. Recommendation, risks, and next actions to reduce uncertainty
    
    Constraints: Show formulas. Never blend sourced and assumed numbers without labeling. Use ranges instead of false precision. No guaranteed outcomes. No private or personal data.
    
    Acceptance criteria:
    - Each number is tagged Sourced, Derived, or Assumed.
    - Top-down and bottom-up results are both shown and the gap is explained.
    - The three most sensitive assumptions are named with a way to validate each.
    - The recommendation references the decision threshold from step 1.
    
    Quality checks: recompute arithmetic once; confirm units and currency are consistent; confirm no C-grade source is the sole basis for a key figure; confirm browsing claims match the tool's real capability.
  7. #18ResearchWith example

    Calculate Rapid Market Size Estimates with Sensitivity Ranges

    Generate a structured market sizing model that calculates conservative, base, and aggressive cases with explicit formulas and variable assumptions.

    Preview prompt
    Act as a market strategy consultant. Your job is to create a Fermi market sizing model based on the information provided by the user. First, ask these 3 questions: 1. What is the specific product or service and the target customer segment? 2. What is the geographic scope of the market? 3. Do you have any known data points like current total industry revenue or estimated unit price? Wait for my answer. Once I reply, provide a 3-tier model. Rules: 1. Clearly state the top-down or bottom-up logic used. 2. Every calculation step must be visible in a table. 3. Explicitly list all assumptions for each variable. 4. Provide three scenarios: Conservative, Base, and Aggressive. Output format: # Market Sizing Estimate for Product Name. ## Logic Summary. ## Sensitivity Table (Columns: Variable, Conservative, Base, Aggressive). ## Calculation Steps (Numbered list of formulas). ## Key Assumptions. This is not professional financial or investment advice; consult a qualified financial advisor before making business decisions based on these estimates.
  8. #19ResearchWith example

    Create a Board-Ready Executive Research Brief

    Transforms raw research notes into a structured, high-impact one-page executive brief using the Minto Pyramid Principle.

    Preview prompt
    Act as a strategy consultant. Your job is to draft a one-page board-ready executive brief based on the research notes I will provide. Follow these rules: 1. Start with a governing thought that summarizes the primary conclusion. 2. Group supporting evidence into 3 distinct logical pillars using the Minto Pyramid Principle. 3. Every bullet point must state a specific fact or data point. 4. Include a dedicated section for unverified assumptions or risks. 5. Keep the tone professional, objective, and concise. This is not professional financial or legal advice; consult a qualified professional before making business decisions based on this summary. Output format: Executive Summary (The governing thought), Key Findings (3 pillars with 3 bullets each), Risks and Assumptions (Numbered list), and Recommended Next Steps (Actionable list). Please ask me for the notes now.
  9. #20ResearchWith example

    Turn Qualitative Interview Notes into an Opportunity Matrix

    Transforms unstructured user interview findings into a standardized opportunity matrix scored by severity, frequency, and willingness to pay.

    Preview prompt
    Act as a principal user researcher. Your job is to convert raw qualitative interview notes and customer feedback into a prioritized opportunity matrix.
    
    Rules:
    1. Consolidate overlapping interview themes into distinct user pain points phrased from the customer perspective.
    2. Score each pain point on a 1-5 scale across three explicit dimensions: Severity (impact when it occurs), Frequency (how often it occurs), and Willingness to Pay (urgency to resolve financially or operationally).
    3. Calculate an overall Opportunity Score by multiplying Severity x Frequency x Willingness to Pay (max score 125).
    4. Provide one concrete observation quote or evidence snippet from the notes for each theme.
    
    Output format:
    - A Markdown table with columns: Pain Point, Evidence / Quote, Severity (1-5), Frequency (1-5), Willingness to Pay (1-5), Opportunity Score (1-125), and Priority Tier (High, Medium, Low).
    - 2-3 bullet points under the table highlighting critical assumptions and missing participant segments.
    
    To begin, please ask me 3 brief questions about the user audience, the interview sample size, and then tell me to paste the raw interview notes.

Citation tables and executive briefs

5 prompts

Turn multi-source web searches into scannable decision briefs with dated link tables and open questions.

  1. #21BusinessPerplexityWith example

    Competitor Research Plan with Sources and a Citation Table

    You get a source discovery plan, a citation table built only from sources you paste, and a decision brief for a competitor or market question, with no pretend browsing.

    Preview prompt
    ROLE: You are a business research analyst who builds evidence-based competitive assessments and is strict about source quality.
    
    TASK: Help me make a decision about competitors or a market. Produce a source discovery plan, and for any sources I paste, an evidence review, citation table, synthesis and decision summary.
    
    START: Before anything else, ask me these questions in one numbered message and wait for my answers:
    1. What decision are you making, and what are the options (for example launch, hold or change price; go, no-go or conditional go)?
    2. Which competitors or market segment are in scope?
    3. Who is the decision for, and by what date?
    4. Which geography, and how far back should evidence go (time window)?
    5. What are the research questions that must be answered (up to 5)?
    6. Which sources have you already collected? Paste excerpts with publisher and date, or say "none yet".
    7. Anything to exclude, or biases I should know about?
    If I skip a question, continue and list the gap under Assumptions.
    
    CONSTRAINTS: Do not claim you browsed the web, opened links or verified live data unless this tool actually has browsing enabled and you name the sources it retrieved. If you cannot browse, give search queries, source types and evaluation criteria for me to run, then analyze only material I supply. Never fabricate citations, URLs, statistics, quotes or dates. Mark anything unverified as Unverified.
    
    Workflow
    1. Restate the decision and turn each research question into a testable claim.
    2. Source discovery plan: for each question, 3 to 5 search queries, the best source types (filings, company pages, pricing pages, regulator data, customer reviews, trade publications, analyst reports) and why each is credible or biased.
    3. Evidence review: for each source I provide, record publisher, date, source type, claim and a reliability rating (High, Medium, Low) with a one-line reason. Flag conflicts between sources.
    4. Citation table: only from sources I provided or you actually retrieved.
    5. Synthesis: answer each question with evidence for and against, a confidence level and what is still unknown.
    6. Decision support: map findings to the options, say which option the evidence favors and what evidence would change that.
    7. Risks, evidence gaps and next actions.
    
    Output format
    1. Decision and research questions (table)
    2. Source discovery plan (table: question, queries, source types, credibility notes)
    3. Evidence log and citation table (columns: ID, publisher, date, source type, key claim, reliability, notes)
    4. Conflicts and gaps
    5. Synthesis by question with confidence level and evidence IDs
    6. Decision support summary (favored option, conditions, what would change it)
    7. Next actions checklist with owner and timing
    
    Acceptance criteria
    - Every finding cites at least one evidence ID from the citation table.
    - No source appears that I did not provide or the tool did not actually retrieve.
    - Confidence levels are justified by source quality and agreement.
    - The decision summary names the evidence gaps that could change the answer.
    
    Quality checks
    - Every evidence ID in the synthesis exists in the table.
    - Flag any claim older than the time window.
    - Keep facts, interpretations and assumptions visibly separate.
    - State at the top whether live browsing was used.
    
    Tool notes: In Perplexity or ChatGPT with search on, ask it to list retrieved sources with titles and dates before the analysis, and treat that list as the citation base. In Claude or Gemini without browsing, run the discovery queries yourself and paste excerpts back in labeled blocks.
  2. #22AI ResearchPerplexityWith example

    Weekly AI Tools Brief for a Team Lead with Dated Sources and One Experiment

    Produces a dated, linked summary of the past week's AI model and tool releases that matter to your team, plus one experiment to run.

    Preview prompt
    You are a research analyst writing a weekly AI tools brief for a team lead. First ask me two short questions: what is the team's function (for example marketing, customer support or engineering), and which tools do they already use? Then use web search to find notable AI model and tool releases from the past 7 days. If you cannot browse live, say so plainly and give me the search queries to run instead of inventing releases. Rules: 1) Prefer primary sources such as official blogs, release notes, changelogs and docs, and use news only as secondary confirmation. 2) Give every item a publication date and a working link, and leave out anything you cannot date to the past week. 3) Label each point as Confirmed (stated by the vendor), Reported (secondary source only) or Inference. 4) Skip hype and minor updates that do not affect this team. Output format: Search queries used; Top 5 releases in a table (name, date, what changed, relevance to the team, source link, confidence); What to ignore this week (one line each); One suggested experiment (goal, steps, time needed, success measure, risk); Open questions; Date checked.
  3. #23ProductivityPerplexityWith example

    Productivity Tool Selection: Source Discovery, Evidence Review, and Citation Table

    A research workflow to choose between candidate productivity or automation tools. It plans online source discovery, structures evidence review in a citation table, and ends in a decision memo without claiming live browsing took place.

    Preview prompt
    Role: You are a careful research analyst supporting a software selection decision.
    
    Context: I must decide which tool to adopt for {use_case}. Candidates: {candidate_tools}. Decision criteria: {criteria}. The job to be done is to make a defensible adoption decision (adopt, pilot, or reject) that I can justify to {decision_audience}. Budget and constraints: {constraints}.
    
    Task: Produce a source discovery plan, an evidence review framework, a citation table, a synthesis, and a decision recommendation.
    
    Important honesty rule: You may not have live web access. Do not claim to have browsed, and do not invent URLs, quotes, prices, version numbers, or statistics. If you cannot verify a fact, label it Unverified and tell me exactly how to verify it. If I paste source excerpts, use only those for evidence and cite them by the label I give.
    
    Inputs (ask for anything missing before starting):
    - Use case and team size: {use_case}
    - Candidate tools: {candidate_tools}
    - Weighted criteria (for example integration, cost, security, learning curve): {criteria}
    - Must-have requirements and deal-breakers: {constraints}
    - Pasted source excerpts, if any, each with a label like S1, S2: {source_excerpts}
    
    Workflow:
    1. Assumptions: list them separately from facts I gave.
    2. Source discovery plan: for each candidate, list the source types to find (official documentation, pricing page, security or compliance page, changelog, independent reviews, user community threads) and give 2-3 search queries per type. Mark which sources are primary versus secondary.
    3. Evidence review: for each criterion and tool, state what evidence is needed and how to rate its reliability (High: official or primary; Medium: independent; Low: anecdotal).
    4. Citation table: build a table from the excerpts I provide. If none were provided, build an empty table with the columns ready and a checklist of what to collect.
    5. Synthesis: score each tool against weighted criteria only where evidence exists; show Unknown otherwise.
    6. Decision support: recommend adopt, pilot, or reject for each tool, with the conditions under which the recommendation changes.
    
    Constraints: Concise and professional. No guaranteed outcomes. Distinguish evidence from opinion.
    
    Output format:
    - Assumptions and open questions
    - Source discovery plan (table: Tool | Source type | Search query | Primary/Secondary)
    - Citation table (Source label | Tool | Claim | Evidence type | Reliability | Verified? Y/N)
    - Weighted scoring matrix with Unknown allowed
    - Synthesis (max 200 words)
    - Recommendation with decision conditions
    - Risks, missing information, and contradictions
    - Next actions (max 5) including a 2-week pilot design if pilot is recommended
    
    Acceptance criteria:
    - No fabricated URLs, quotes, or numbers.
    - Every scored claim maps to a source label or is marked Unknown.
    - Deal-breakers are checked explicitly per tool.
    - The recommendation states what evidence would change it.
    
    Quality checks (report pass/fail): browsing not claimed; unverified items flagged; weights sum to 100 percent; citations match the labels I supplied; risks are specific to the candidates.
    
    Tool-specific instructions: In Perplexity or ChatGPT with search enabled, run the queries from the discovery plan and list real sources with links as you find them, marking anything you could not open. In Claude or Gemini without search, treat the plan as a checklist for me to run, and ask me to paste excerpts.
  4. #24AI MarketingWith example

    Audience Research Source Discovery and Evidence Synthesis for Segment Prioritization

    Guides you to plan source discovery, review the evidence you collect, build a citation table, and synthesize findings into a segment prioritization decision, without pretending the model browsed the web.

    Preview prompt
    ROLE: You are a marketing research analyst who designs audience research plans and synthesizes evidence supplied by the user for {brand}. CONTEXT: The team must decide which audience segment to prioritize for {campaign_or_product} in {market}. The job-to-be-done is to move from open questions to a defensible segment choice with documented sources and known gaps. TASK: Create a source discovery plan the user can run, then review and synthesize the evidence the user pastes back, and produce a segment prioritization recommendation. CONSTRAINTS: Do not claim to have browsed the internet, accessed live data, or verified that a source exists. Only cite sources the user provides, and label any source you suggest searching for as a lead to verify. Do not invent statistics, quotes, or market sizes. Separate facts from interpretations and assumptions. Do not request or use private personal data. TOOL INSTRUCTIONS FOR CHATGPT: Work in two phases. In Phase A, produce the research plan and search strings, then stop and ask the user to paste source excerpts with URL, title, publisher, date, and the relevant passage. In Phase B, after the user pastes evidence, perform the review and synthesis. If browsing or file tools are enabled in the user's account, say the user must confirm that outputs came from those tools and should list the sources explicitly. Use markdown tables for the citation table and scoring. Inputs: 1) Business question and decision to support: {decision_question}. 2) Candidate segments or hypotheses: {candidate_segments}. 3) Product or offer summary: {offer_summary}. 4) Geography, language, and time window that matter: {market}. 5) Existing first-party insights such as survey results, CRM summaries, or support themes, with no personal data: {first_party_insights}. 6) Decision criteria and weights such as size, need intensity, reachability, margin fit, and competition: {decision_criteria}. 7) Pasted source excerpts in Phase B: {source_excerpts}. Workflow: Step 1, Question framing: restate the decision, list 5 to 8 research sub-questions, and define what evidence would change the decision. Step 2, Source discovery plan: list source types to look for such as industry reports, public company filings, review sites, community forums, job postings, search trend tools, analyst blogs, and government statistics, with a search string set for each and a note on typical bias. Step 3, Source intake template: provide a table the user can fill with ID, URL, publisher, date, source type, passage, and relevance. Step 4, Evidence review: after excerpts are supplied, grade each source for recency, relevance, method transparency, and potential bias, and mark it high, medium, or low reliability with a reason. Step 5, Citation table: list each claim extracted, the source ID, whether it is a direct fact or interpretation, and the confidence level. Step 6, Synthesis: group findings by segment, identify agreements and conflicts across sources, and state what remains unknown. Step 7, Scoring: apply the decision criteria and weights to produce a segment scoring table with rationale referencing source IDs. Step 8, Decision support: recommend a primary segment and a secondary segment, define the messaging hypotheses to test, and list the research needed to raise confidence. Output format: Phase A output uses the headings Question Framing, Source Discovery Plan, Search Strings, Source Intake Template. Phase B output uses the headings Source Reliability Review, Citation Table, Findings by Segment, Conflicts and Gaps, Segment Scoring Table, Recommendation, Risks and Missing Information, Next 5 Actions. Acceptance criteria: Every factual claim in the synthesis maps to a user-supplied source ID or is labeled as an assumption. The recommendation follows the scoring table and the stated weights. Conflicts between sources are shown and not hidden. Unverified leads are clearly separated from supplied evidence. The next actions include owner role, effort estimate, and the decision they unlock. Quality checks: Confirm there are no unsupported numbers, that dates are noted for time-sensitive claims, that small or biased samples are flagged, that the plan does not imply live browsing occurred, and that the final confidence rating of low, medium, or high is justified in two sentences.
  5. #25AI SEOPerplexityWith example

    Write an Answer-Engine FAQ for Conversational Follow-Up Prompts

    Get a ready-to-publish FAQ that pairs likely second-turn AI follow-up questions with 40-to-60-word direct answers and concrete facts for Perplexity, Gemini, ChatGPT, and Claude.

    Preview prompt
    You are a technical SEO and Generative Engine Optimization (GEO) editor. Your job is to write an answer-engine FAQ targeting the conversational follow-up questions users ask after an initial query in Perplexity, Gemini, ChatGPT, or Claude.
    
    Before writing, ask me these 3 short questions in one numbered message and wait for my reply:
    1. What is the product, service, or topic of this page, and who is the target reader?
    2. What is the primary initial question users ask before they reach this page?
    3. Paste 3 to 6 verified facts, specs, pricing details, or policies from your documentation that I must use.
    
    Rules:
    - Write 4 FAQ entries phrased as natural second-turn follow-up questions a user would type into an AI chat.
    - Write each answer as a 40-to-60-word direct-answer lead that resolves the question in the first sentence.
    - Embed at least one concrete, verified fact from my input in every answer; never invent statistics, benchmarks, or citations.
    - Strip all promotional phrasing, vague superlatives, and filler intros that answer engines ignore.
    - Keep sentences under 22 words using plain, declarative prose.
    
    Output format:
    - Heading: Page Topic and Primary Query
    - Table: Follow-Up Intent Map (Columns: Question, Follow-Up Intent, Verified Fact Used)
    - Section: Ready-to-Publish FAQ (H3 per question followed by the 40-to-60-word answer)
    - Final note: List any missing facts or policy gaps that could cause an AI engine to skip or misquote this page.

Playbook

How to get better results

Which AI model to pick

Perplexity
Fast live web searches with clickable inline citations and Research mode comparisons across many sources.
Gemini
Long-running Deep Research reports you want to export straight into Google Docs.
Claude
Deep synthesis once you have downloaded the primary PDFs or transcripts and want strict quote-grounded analysis.

Follow-up prompts to refine any output

  1. Re-check every figure above against primary sources only and remove any claim backed solely by a secondary blog.
  2. Add a short Gaps and Conflicting Evidence section showing where credible sources disagree or data is outdated.
  3. Convert the findings into a clean comparison table with columns for claim, publication date, source type and direct link.

FAQ

About this list

Unlike an offline chat window, Perplexity searches current sources while it writes, which makes prompt structure matter even more. Each prompt tells the engine which source types to prioritise, asks for publication dates beside every claim and separates verified primary facts from analyst commentary or open disagreement. Every prompt shows an example output so you can inspect the citation format before running it.

You will find prompts for due diligence, competitive positioning grids, literature reviews, policy or regulatory lookups and executive research briefs that highlight missing data instead of papered-over guesses. Open in Perplexity starts the search with your filled prompt straight away, and every prompt also works in ChatGPT, Claude or Gemini with web search enabled.

When should I use Perplexity Research mode versus standard search?

Use standard search for quick fact checks, single-company lookups and verifying a specific statistic. Switch to Research mode when a prompt asks to compare multiple competitors, synthesise conflicting studies or produce a cited brief across dozens of sources, since it runs follow-up searches automatically before summarising.

How do I verify that Perplexity citations actually support the claim?

Click through the linked primary sources for every number, quote or date you plan to publish or present. Ask the prompt to include the publication date and exact source type next to each finding, because an answer engine can still cite a secondary blog post or misread a table.

How can I restrict Perplexity to higher-quality sources?

Name the evidence standard right inside the prompt, such as official documentation, filings, government statistics or peer-reviewed journals, and tell it to ignore undated listicles or vendor press releases. You can also toggle Perplexity's source focus controls beside the search box before sending the prompt.

Do these research prompts also work in ChatGPT, Claude or Gemini?

Yes. Every prompt uses plain research criteria rather than Perplexity-only commands, so you can run them in ChatGPT with web search, Gemini Deep Research or Claude when you upload source PDFs. Perplexity is often fastest when you want compact inline links next to every sentence in a summary.