Prompt Library

ChatGPT Prompts for Product Managers

30 copy-paste prompts

Thirty copy-paste prompts that turn ChatGPT into your PM co-pilot — from discovery interviews and PRDs to roadmaps, user stories, experiment design, and exec updates.

In short: This page contains 30 copy-paste ready prompts, organized into 6 categories with a description and pro tip for each. The first 15 prompts are free instantly — no signup needed. Hand-curated and tested by the AI Academy team.

By Louis Corneloup · Founder, Techpresso
Last updated ·Hand-curated & tested by the AI Academy team

Discovery & User Research

5 prompts

Generate user interview questions

1/30

<context> I am a product manager planning discovery interviews for [PRODUCT]. The feature area under investigation is [FEATURE] and the target interviewee is [USER]. I want to validate the underlying problem before committing to a solution. </context> <task> 1. Write 12 open-ended, non-leading interview questions that uncover the [USER]'s current workflow, pain points, and workarounds around [FEATURE]. 2. Group the questions into three phases: warm-up/context, problem exploration, and current solutions. 3. For each question, add a one-line note on what signal it is meant to surface. 4. Flag any question that risks bias and rewrite it neutrally. 5. End with 3 follow-up probes I can use when an answer is vague. </task>

A phased, bias-checked interview guide ready to run with target users.

💡

Pro tip: Paste a transcript back into the same ChatGPT thread afterward and ask it to extract themes — it keeps the original research goals in context.

Synthesize interview notes into themes

2/30

<context> I ran discovery interviews for [PRODUCT] about [FEATURE] with several [USER] participants. I will paste my raw notes below. I need to turn scattered observations into decision-ready insight. </context> <task> 1. Read the notes I paste and cluster observations into 4-6 themes, naming each theme in plain language. 2. For each theme, give a frequency estimate (how many participants raised it) and 1-2 representative quotes. 3. Separate stated needs from underlying jobs-to-be-done. 4. Rank themes by how strongly they justify investment in [FEATURE]. 5. List the top 3 open questions the interviews did NOT answer. </task>

Structured themes, JTBD, and prioritized insights from raw interview notes.

💡

Pro tip: Turn on a fresh thread per study so old transcripts do not bleed into the synthesis; ChatGPT weights recent context heavily.

Build a discovery survey

3/30

<context> I want to validate demand for [FEATURE] in [PRODUCT] at scale before interviews. The audience is [USER]. The survey should take under 5 minutes. </context> <task> 1. Draft a 10-question survey mixing multiple-choice, Likert, and one open-text question. 2. Include a screening question to confirm the respondent matches [USER]. 3. Ensure at least one question measures current pain severity and one measures willingness to switch. 4. Avoid double-barreled and leading questions; flag any you had to fix. 5. Suggest the single metric I should track to decide go/no-go. </task>

A short, screened demand-validation survey with a go/no-go metric.

💡

Pro tip: Ask ChatGPT to output the survey as a numbered list so you can paste it straight into Typeform or Google Forms.

Map the competitive landscape

4/30

<context> I am evaluating whether [PRODUCT] should build [FEATURE]. I need a quick competitive scan to understand how [USER] solves this today. </context> <task> 1. List the main categories of alternatives a [USER] currently uses to address the [FEATURE] problem, including non-software workarounds. 2. For each, summarize the typical approach, strengths, and gaps in 2-3 sentences. 3. Identify the white-space opportunity where [PRODUCT] could differentiate. 4. Note the assumptions in your answer that I should verify with primary research. 5. Recommend 3 competitor experiences I should personally test. </task>

A competitive scan with differentiation opportunities and verification flags.

💡

Pro tip: ChatGPT can hallucinate competitor specifics — treat its output as hypotheses and confirm pricing or features on the live sites.

Write a problem statement

5/30

<context> I have early discovery signal for [PRODUCT] suggesting [USER] struggles with something around [FEATURE]. I need a crisp problem statement to align the team before we ideate. </context> <task> 1. Draft a one-sentence problem statement using the format: [USER] needs a way to [job] because [motivation], but [obstacle]. 2. Provide two alternative framings at different scopes (narrow vs broad). 3. List the evidence that supports the problem and the evidence still missing. 4. State explicitly what is in and out of scope. 5. Add a falsifiable hypothesis we could test next. </task>

A sharp, scoped problem statement with supporting evidence and a hypothesis.

💡

Pro tip: Ask for the three scope variants, then pick one — comparing breadth side by side surfaces hidden assumptions fast.

Prompts get you started. Tutorials level you up.

A growing library of 300+ hands-on AI tutorials. New tutorials added every week.

Start 7-Day Free Trial

PRDs & Specs

5 prompts

Draft a full PRD

6/30

<context> I am writing a PRD for [FEATURE] in [PRODUCT]. The primary user is [USER]. I will provide the problem, goals, and any constraints. I need a complete, review-ready document. </context> <task> 1. Produce a PRD with these sections: Summary, Problem & Context, Goals & Non-Goals, Target [USER] & Use Cases, Requirements (must/should/could), Success Metrics, Risks & Open Questions. 2. Write requirements as testable statements, not vague intentions. 3. Explicitly list Non-Goals to prevent scope creep. 4. Flag any section where I have not given you enough input and ask for it. 5. Keep it skimmable with short paragraphs and bullets. </task>

A complete, review-ready PRD with testable requirements and non-goals.

💡

Pro tip: Feed ChatGPT your problem statement and discovery themes first so the PRD inherits real context instead of generic boilerplate.

Pressure-test a spec

7/30

<context> I have drafted a spec for [FEATURE] in [PRODUCT] for [USER]. Before I share it with engineering, I want it stress-tested for gaps. </context> <task> 1. Read the spec I paste and list every ambiguous or underspecified requirement. 2. Identify edge cases and error states the spec does not cover. 3. Call out hidden assumptions about [USER] behavior or technical constraints. 4. Flag any requirement that conflicts with another. 5. Output a prioritized list of clarifications I should resolve before handoff. </task>

A gap analysis exposing ambiguities, edge cases, and conflicts in a spec.

💡

Pro tip: Have ChatGPT role-play as a skeptical staff engineer for this pass — the framing yields sharper edge-case questions.

Write acceptance edge cases

8/30

<context> I need a thorough edge-case and error-handling section for [FEATURE] in [PRODUCT]. The user is [USER]. </context> <task> 1. Enumerate edge cases across categories: empty/null inputs, max/min limits, concurrency, permissions, offline/slow network, and unusual [USER] behavior. 2. For each, describe the expected system behavior in one line. 3. Highlight which edge cases are most likely to occur for [USER] in real use. 4. Note any cases that need a product decision rather than an engineering default. 5. Format as a table I can drop into the spec. </task>

A categorized edge-case table with expected behaviors and decision flags.

💡

Pro tip: Ask for it as a Markdown table; ChatGPT pastes cleanly into Notion, Confluence, or Linear.

Create a one-page brief

9/30

<context> Leadership wants a one-pager on [FEATURE] for [PRODUCT] before approving a full spec. The user is [USER]. It must be persuasive but honest. </context> <task> 1. Write a one-page brief: the problem in one paragraph, the proposed solution, why now, expected impact, rough effort, and the ask. 2. Lead with the [USER] outcome, not the feature mechanics. 3. Quantify impact where possible and state assumptions behind any number. 4. Include a short risks section so it does not read as a sales pitch. 5. End with a clear, single decision request. </task>

A persuasive, honest one-pager that ends in a clear decision ask.

💡

Pro tip: Tell ChatGPT the exact word limit (e.g. 350 words) — one-pagers balloon otherwise.

Generate clarifying questions for stakeholders

10/30

<context> I am about to scope [FEATURE] for [PRODUCT] but several details are still fuzzy. I need to interrogate stakeholders before writing the spec. The end user is [USER]. </context> <task> 1. Generate the 10 most important clarifying questions I should ask stakeholders before specifying [FEATURE]. 2. Group them by stakeholder type (eng, design, sales, support, legal). 3. For each question, note the risk of NOT having the answer. 4. Mark which questions block writing the spec versus which can be resolved later. 5. Suggest a logical order to ask them in. </task>

A stakeholder-grouped question list ranked by blocking risk.

💡

Pro tip: Drop the answers back in the thread and ask ChatGPT to draft the spec — it threads the context automatically.

Roadmapping & Prioritization

5 prompts

Score features with RICE

11/30

<context> I am prioritizing the backlog for [PRODUCT] and need to apply RICE scoring. I will paste a list of candidate features, including [FEATURE]. The core user is [USER]. </context> <task> 1. For each feature I paste, propose Reach, Impact, Confidence, and Effort estimates with a one-line rationale each. 2. Calculate the RICE score and rank the features. 3. Flag any estimate that rests on a shaky assumption I should validate. 4. Note which features are quick wins versus big bets. 5. Output a ranked table plus a 2-sentence recommendation. </task>

A ranked RICE table with rationales, assumption flags, and a recommendation.

💡

Pro tip: Give ChatGPT real reach numbers from your analytics; without data it guesses, and the ranking is only as good as the inputs.

Build a quarterly roadmap

12/30

<context> I need to turn a prioritized backlog for [PRODUCT] into a themed quarterly roadmap. The strategic goal this quarter is to better serve [USER], and [FEATURE] is a candidate. </context> <task> 1. Organize the work I paste into 3-4 outcome-based themes, not a list of features. 2. Map candidate items including [FEATURE] to the themes and a Now / Next / Later horizon. 3. For each theme, state the [USER] outcome and the metric that proves it worked. 4. Identify dependencies and sequencing risks. 5. Note what we are deliberately NOT doing this quarter and why. </task>

An outcome-themed Now/Next/Later roadmap with metrics and explicit cuts.

💡

Pro tip: Ask ChatGPT to frame themes as outcomes — it defaults to feature lists unless you insist on the outcome framing.

Build a trade-off matrix

13/30

<context> The team disagrees on whether to build [FEATURE] now in [PRODUCT]. I need a neutral trade-off analysis to drive a decision for [USER]. </context> <task> 1. Lay out 3-4 realistic options (e.g. build now, build lite, defer, buy/integrate) for [FEATURE]. 2. Evaluate each against criteria: [USER] impact, effort, risk, strategic fit, and reversibility. 3. Present it as a scored matrix with brief justifications. 4. State which option you would recommend and the single biggest reason against it. 5. List the data point that would most change the recommendation. </task>

A neutral scored option matrix with a recommendation and counter-evidence.

💡

Pro tip: Ask for the strongest argument against the recommendation — it surfaces blind spots before a stakeholder does.

Write OKRs from a strategy

14/30

<context> I need to translate a high-level strategy for [PRODUCT] into quarterly OKRs for my product team. The strategic focus is delivering value to [USER] partly via [FEATURE]. </context> <task> 1. Propose 2-3 Objectives that are qualitative, inspiring, and tied to [USER] value. 2. Under each, write 3 measurable Key Results with baseline and target placeholders. 3. Ensure Key Results are outcomes, not feature-shipped checkboxes. 4. Flag any KR that is a vanity metric and suggest a better one. 5. Note which initiatives, including [FEATURE], support each Objective. </task>

Outcome-based OKRs with measurable key results and vanity-metric flags.

💡

Pro tip: Reject any KR phrased as "launch X" — push ChatGPT to restate it as the behavior change you expect from [USER].

Plan a feature cut for a deadline

15/30

<context> [FEATURE] in [PRODUCT] is at risk of missing its deadline. I need to identify the minimum lovable version for [USER] without shipping something broken. </context> <task> 1. Break [FEATURE] into its component capabilities from the description I paste. 2. Classify each as must-have for launch, fast-follow, or cut-entirely for [USER] value. 3. Define the minimum lovable version and what makes it lovable rather than merely minimal. 4. List the risks of each cut and how to communicate them. 5. Draft a 3-bullet message I can send to stakeholders explaining the scope change. </task>

A scoped MLP with cut rationale and a stakeholder-ready scope-change note.

💡

Pro tip: Distinguish "minimum" from "lovable" explicitly in the prompt; ChatGPT otherwise strips it to a feature stub no one wants.

User Stories & Acceptance Criteria

5 prompts

Write user stories from a feature

16/30

<context> I need to break [FEATURE] in [PRODUCT] into well-formed user stories for the backlog. The primary persona is [USER]. </context> <task> 1. Decompose [FEATURE] into 5-8 vertical-slice user stories in the format: As a [USER], I want [capability] so that [benefit]. 2. Keep each story independently shippable and small enough for one sprint. 3. Order them by delivery sequence and note dependencies. 4. Flag any story that is actually an epic and should be split further. 5. Tag each story with a rough T-shirt size estimate. </task>

A sequenced set of vertical-slice user stories with sizes and dependencies.

💡

Pro tip: Ask ChatGPT to flag epics-disguised-as-stories — it catches oversized items that bloat sprint planning.

Generate Gherkin acceptance criteria

17/30

<context> I have a user story for [FEATURE] in [PRODUCT] and need precise acceptance criteria for [USER]. The team uses Given/When/Then. </context> <task> 1. For the story I paste, write acceptance criteria in Given/When/Then format. 2. Cover the happy path plus at least 3 alternate or error scenarios. 3. Make each criterion independently testable and unambiguous. 4. Include any [USER] permission or state preconditions explicitly. 5. List any criterion that requires a product decision before it can be finalized. </task>

Testable Given/When/Then criteria covering happy and error paths.

💡

Pro tip: Specify your Gherkin dialect or tool (e.g. Cucumber) so ChatGPT matches your automated-test syntax exactly.

Define Definition of Done

18/30

<context> My team lacks a shared Definition of Done for work on [PRODUCT], which causes rework. I want a checklist that applies to stories like [FEATURE] serving [USER]. </context> <task> 1. Draft a Definition of Done checklist covering code, testing, design QA, accessibility, analytics/instrumentation, docs, and release readiness. 2. Tailor items to a [USER]-facing feature like [FEATURE]. 3. Mark which items are mandatory versus context-dependent. 4. Keep it to one page and phrased as verifiable yes/no checks. 5. Suggest where to store and enforce it in our workflow. </task>

A one-page, verifiable Definition of Done checklist for the team.

💡

Pro tip: Ask for yes/no phrasing — ambiguous DoD items ("tested adequately") defeat the purpose; binary checks do not.

Split an epic into stories

19/30

<context> [FEATURE] in [PRODUCT] is too large to estimate and is really an epic for [USER]. I need to split it sensibly. </context> <task> 1. Restate [FEATURE] as an epic-level goal in one sentence. 2. Apply splitting patterns (by workflow step, by user role, by data variation, by happy-path-then-edge) and propose the best split. 3. Output the resulting stories, each independently valuable to [USER]. 4. Sequence the stories so we can ship and learn early. 5. Identify the smallest story that still delivers a learning signal. </task>

An epic broken into independently valuable, learning-oriented stories.

💡

Pro tip: Name the splitting patterns in the prompt; left vague, ChatGPT just chops by UI screen, which creates non-shippable slices.

Write a QA test plan

20/30

<context> Before releasing [FEATURE] in [PRODUCT] to [USER], I need a lightweight test plan QA and I can run. </context> <task> 1. From the acceptance criteria I paste, derive a test plan grouped by functional area. 2. For each test, give steps, test data, and expected result. 3. Include cross-cutting checks: edge cases, performance, accessibility, and regression areas. 4. Prioritize tests as P0/P1/P2 based on [USER] impact. 5. Note which tests are good candidates for automation. </task>

A prioritized manual test plan with steps, data, and automation candidates.

💡

Pro tip: Have ChatGPT mark automation candidates so you can hand the P0 set to engineering for regression coverage.

Metrics & Experiments

5 prompts

Design an A/B test

21/30

<context> I want to A/B test a change to [FEATURE] in [PRODUCT] aimed at improving behavior for [USER]. I need a statistically sound experiment design. </context> <task> 1. State a clear hypothesis in the form: changing X will improve [metric] for [USER] because [reason]. 2. Define the primary metric, 1-2 guardrail metrics, and the unit of randomization. 3. Recommend a sample size approach and how long to run given my traffic (ask me for it if missing). 4. List threats to validity (novelty effect, contamination, seasonality) and mitigations. 5. Define the decision rule for ship / iterate / kill. </task>

A rigorous A/B test design with hypothesis, guardrails, and a decision rule.

💡

Pro tip: Give ChatGPT your weekly traffic and baseline conversion so its sample-size guidance is grounded rather than generic.

Build a metrics tree

22/30

<context> I need to connect [FEATURE] in [PRODUCT] to the business by mapping a metrics tree down from a North Star for [USER]. </context> <task> 1. Propose a North Star metric for [PRODUCT] tied to [USER] value, with rationale. 2. Decompose it into input metrics across acquisition, activation, engagement, retention, and revenue. 3. Show where [FEATURE] plugs into the tree and which input metric it should move. 4. Flag any metric that is hard to instrument or easily gamed. 5. Recommend the 3 metrics I should put on the team dashboard. </task>

A North Star metrics tree showing exactly where a feature should move the needle.

💡

Pro tip: Ask which metrics are easily gamed — ChatGPT is good at spotting proxy metrics that drift from real [USER] value.

Diagnose a metric drop

23/30

<context> A key metric for [FEATURE] in [PRODUCT] dropped unexpectedly and leadership wants answers. The affected segment is [USER]. </context> <task> 1. Generate a structured list of hypotheses for the drop across categories: tracking/data, product change, external/seasonal, segment mix, and funnel stage. 2. For each hypothesis, state how I would quickly confirm or rule it out. 3. Order the investigation from cheapest-to-check to most expensive. 4. Identify the single most likely cause given the context I paste. 5. Draft a holding update I can send leadership while I investigate. </task>

A prioritized hypothesis tree for a metric drop plus a leadership holding note.

💡

Pro tip: Paste the segment breakdown and timeline; ChatGPT reasons far better about anomalies when it sees the actual shape of the drop.

Write an experiment readout

24/30

<context> I just finished an experiment on [FEATURE] in [PRODUCT] targeting [USER] and need to write the results readout for stakeholders. </context> <task> 1. From the results I paste, write a readout: hypothesis, what we tested, primary and guardrail results with significance, and the decision. 2. Lead with the decision and the so-what, not the methodology. 3. Note any caveats, surprising segment effects, or follow-up questions. 4. Translate statistical results into plain-language [USER] impact. 5. End with a clear recommendation and next step. </task>

A decision-first experiment readout that translates stats into plain impact.

💡

Pro tip: Tell ChatGPT to lead with the decision; default outputs bury the recommendation under methodology that execs skip.

Define success metrics for a launch

25/30

<context> I am launching [FEATURE] in [PRODUCT] for [USER] and need to define what success looks like before launch, not after. </context> <task> 1. Define adoption, engagement, and outcome metrics specific to [FEATURE]. 2. For each, set a measurement window and a realistic target with the reasoning behind it. 3. Specify the leading indicator I can watch in week one versus the lagging outcome. 4. Identify counter-metrics that would signal we are harming [USER] elsewhere. 5. Recommend how to instrument these and what to put on the launch dashboard. </task>

Pre-launch success metrics with windows, targets, leading indicators, and counter-metrics.

💡

Pro tip: Force a target number per metric; ChatGPT will hedge unless you ask it to commit and justify the figure.

Go from copy-pasting to actually mastering AI.

AI Academy: 300+ hands-on tutorials on ChatGPT, Claude, Midjourney, and 50+ other tools. New tutorials added every week.

Start Your Free Trial

Stakeholder Communication

5 prompts

Write an executive update

26/30

<context> I need to send a monthly product update to executives covering progress on [FEATURE] in [PRODUCT] and its impact on [USER]. </context> <task> 1. Draft a concise update: TL;DR, key wins, metrics movement, risks/blockers, and what I need from leadership. 2. Lead with outcomes and decisions needed, not activity. 3. Quantify progress on [FEATURE] and tie it to [USER] impact. 4. Keep it scannable in under 60 seconds. 5. End with 1-2 explicit asks or decisions required. </task>

A scannable, outcome-led executive update ending in clear asks.

💡

Pro tip: Give ChatGPT your raw bullet notes and let it structure them; it is far better at shaping than at inventing your progress.

Translate technical detail for stakeholders

27/30

<context> Engineering gave me a technical explanation about [FEATURE] in [PRODUCT] that I must relay to non-technical stakeholders who care about [USER]. </context> <task> 1. Rewrite the technical content I paste in plain language for a business audience. 2. Explain the [USER] and business implications, not the implementation. 3. Use one concrete analogy if it aids understanding. 4. Preserve any caveat or risk that actually matters to the decision. 5. Keep it to under 200 words and flag anything I should double-check with eng. </task>

A plain-language translation of technical detail focused on business impact.

💡

Pro tip: Ask it to flag anything to re-verify with engineering — it prevents you from confidently relaying a simplification that is wrong.

Prepare for a tough stakeholder meeting

28/30

<context> I have a contentious meeting where I must defend a decision about [FEATURE] in [PRODUCT] that affects [USER]. Some stakeholders disagree. </context> <task> 1. Anticipate the 5 toughest objections to my decision based on the context I paste. 2. For each, draft a calm, evidence-based response. 3. Identify the underlying interest behind each objection, not just the stated position. 4. Suggest where I can offer a genuine compromise without undermining the decision. 5. Give me a one-line framing to open the meeting and align on shared goals. </task>

An objection-handling brief with interests, responses, and compromise zones.

💡

Pro tip: Have ChatGPT separate stated positions from underlying interests — that reframing defuses most product disagreements.

Write a feature announcement

29/30

<context> We are shipping [FEATURE] in [PRODUCT] to [USER] and I need internal and external announcement copy. </context> <task> 1. Write an internal announcement (Slack-length) covering what shipped, why it matters for [USER], and what teams should do. 2. Write an external-facing blurb leading with the [USER] benefit, not the feature name. 3. Include a one-sentence value proposition and a clear call to action. 4. Provide a shorter changelog-style version. 5. Suggest 2 metrics to watch to confirm the announcement landed. </task>

Internal and external launch copy plus a changelog version and watch-metrics.

💡

Pro tip: Generate the three lengths in one go, then trim — it is faster to cut ChatGPT copy than to expand a too-short draft.

Say no to a feature request

30/30

<context> A stakeholder is pushing hard for a request related to [FEATURE] in [PRODUCT] that I do not plan to prioritize for [USER]. I need to decline without damaging the relationship. </context> <task> 1. Draft a respectful response that acknowledges the request and the underlying need. 2. Explain the decision using prioritization criteria and current goals, not vague excuses. 3. Offer an alternative or a future revisit trigger so it does not feel like a dead end. 4. Keep the tone collaborative and confident, not defensive. 5. End by inviting them to share data that could change the prioritization. </task>

A respectful, criteria-based no that preserves the relationship and leaves a door open.

💡

Pro tip: Ask ChatGPT to anchor the no in your stated goals/criteria; "no because of strategy" lands far better than "no, too busy".

Frequently Asked Questions

Copy a prompt, replace the [PRODUCT], [FEATURE], and [USER] placeholders with your real details, and paste it into ChatGPT. The structured context/task format works in any model, but GPT-4-class models give the most useful, on-topic output. Paste your own notes, transcripts, or data into the same thread to ground the response in your reality rather than generic advice.
A reasoning-capable model (the most advanced one in your plan) handles multi-step PM tasks like RICE scoring, experiment design, and spec gap analysis best. Lighter models are fine for drafting announcements or interview questions. For anything involving numbers or trade-offs, give the model real data — it reasons well but cannot invent your analytics.
No. ChatGPT accelerates the scaffolding — interview guides, survey drafts, theme synthesis from notes you provide — but it cannot talk to your users. Treat its competitive scans and assumptions as hypotheses to validate, never as findings. The real signal still comes from talking to real [USER]s.
Ground every prompt in data you paste: real reach numbers, actual transcripts, your strategy doc. Ask the model to flag its assumptions and mark anything to verify. Several prompts here deliberately request assumption flags and verification lists for exactly this reason. Never ship a competitor claim or metric ChatGPT produced without checking the source.
Check your company policy first. Many teams use ChatGPT Team or Enterprise, where inputs are not used for training. Avoid pasting regulated data or secrets into consumer accounts. When in doubt, anonymize — replace real names and numbers with placeholders, run the prompt, then reintroduce specifics locally.

Prompts are the starting line. Tutorials are the finish.

A growing library of 300+ hands-on tutorials on ChatGPT, Claude, Midjourney, and 50+ AI tools. New tutorials added every week.

7-day free trial. Cancel anytime.