Most PRDs collapse because the outline was wrong, not because the writing was bad. These 15 prompts generate the structural skeleton — the headings, the order, the missing-section checklist — for feature PRDs, MVP PRDs, platform PRDs, growth PRDs and bug-batch PRDs. Use them as the bones; fill the prose later with your draft prompts. They are written to run as-is in ChatGPT (GPT-5.5), Claude (Opus 4.7 or Sonnet 4.6), or Gemini 3.1 Pro.
TL;DR
- Generate the outline before the prose — a wrong outline costs more to fix after writing than a wrong sentence.
- Each prompt below produces sections plus a one-line description, so you can argue about structure before anyone drafts a paragraph.
- A tight PRD is 2–4 pages and is explicit about what is out of scope; over 14 top-level sections and reviewers start skimming.
- Pasting a full transcript or a draft to audit? Any of the three flagship models holds it: Claude Opus 4.7 and Sonnet 4.6 and Gemini 3.1 Pro carry a 1M-token window; ChatGPT Plus carries roughly 320 pages in-app (full 1M only on the $200 Pro tier), as of June 2026.
- Industry signal: 94% of product professionals now use AI daily or often, and 100% of surveyed teams use it at all (Productboard, 2026). The catch teams report in practice: a one-click generated PRD almost always needs a heavy structural rewrite. Outlining first — not one-shot generation — is what closes that gap.
Who this is for
PMs who keep starting PRDs from a blank page, tech leads scoping a quarter of work, founders documenting a feature for a freelancer, and design partners standardizing PRD shape across squads.
Which model to run these in (June 2026)
The outline task is short, so any current model handles it. The difference shows up on the audit and transcript prompts, where you paste a long document and ask the model to reason over all of it.
| Model | Context (standard) | Best for in this workflow | Plan to access |
|---|---|---|---|
| Claude Opus 4.7 | 1M tokens | Auditing a long draft, structural reasoning | Pro $20 / Max from $100 |
| Claude Sonnet 4.6 | 1M tokens | Fast outline + cheap iteration | Free (limited) / Pro $20 |
| Gemini 3.1 Pro | 1M tokens | Transcript-to-outline, reference-heavy work | Google AI Pro $19.99 |
| GPT-5.5 | ~320 pages in-app (Plus) | First-draft outlines, quick options | Plus $20 / Pro $100–$200 |
Pricing and context limits are as of June 2026 and change often; check the vendor page before you commit.
When not to use these prompts
Skip when the team already has a templated PRD format that works — do not import noise. Skip too for one-paragraph feature ideas where a full PRD is overkill.
Prompt anatomy / structure formula
A PRD outline prompt should always carry six elements:
- Role: who the AI plays (senior PM / solo founder / product designer / indie dev / growth lead).
- Context: stage (idea / MVP / growth / scale), team size, traffic or ARR, platform (web / iOS / Android), audience, constraints.
- Goal: one concrete deliverable — one PRD section, one user-story set, one experiment design, one launch post.
- Constraints: timeline (this sprint / this quarter), scope cuts, must-not-break (existing flows, billing, compliance).
- Output format: table, checklist, ticket-ready JSON, or labeled blocks you can paste straight into Linear / Notion / Jira.
- Examples / signal: 1-2 reference docs or competitors you like, plus 1 anti-example you want to avoid.
Best for
- New-PM onboarding template
- Standardizing PRDs across squads
- Pre-PRD sanity check on scope
- PRD review (is anything missing)
- Vendor / freelancer brief skeleton
15 copy-ready prompt templates
1. Standard feature PRD outline
The default outline. Use this when the PM has no opinion yet.
You are a senior PM. Generate a standard PRD outline for a single feature. Output as a markdown table of contents with H1/H2/H3 levels. Include: Summary, Problem, Goals + non-goals, Target users, Success metrics, User stories, UX flows, Functional requirements, Non-functional requirements, Dependencies, Risks, Open questions, Rollout plan, Out of scope. For each section, add a 1-line description of what goes there.Variables to swap: feature, audience, team
Optimization: If the output is too generic, add: “Tag every section with whether it is required for v1, or only needed for v2+. Cut anything not required for v1.”
2. MVP-grade slim outline
Generate a slim PRD outline for an MVP that ships in 4 weeks. Maximum 8 sections. Must include: thinnest slice that delivers value, what we are explicitly NOT building, success criteria, learning goals. Output as a numbered list with 1-line section descriptions.3. Platform / infrastructure PRD outline
Generate a PRD outline for an internal platform feature (not user-facing). Sections must cover: API surface, backwards compatibility, migration path, performance SLOs, observability, security model, capacity planning, sunset plan for the old thing. Mark which sections need sign-off from which team.4. Growth-feature PRD outline
Generate a PRD outline tailored for a growth experiment. Sections: hypothesis, primary metric, guardrail metrics, target lift + minimum detectable effect, control vs variant, sample size estimate, duration, rollback trigger, learning even if it fails. Output with each section + 1 example sentence.5. Bug-batch / quality PRD outline
Generate a PRD outline for a quality / bug-batch sprint. Sections: bug inventory, severity scoring, prioritization rubric, root-cause categories, prevention work, regression test plan, definition of done. Format as a checklist with sign-off owners.6. AI-feature PRD outline
For the “model choice + fallback” section, name a real primary and fallback in your context — e.g. Claude Opus 4.7 as primary and Gemini 3.1 Pro as fallback, with per-call API costs ($5/$25 per 1M in/out for Opus 4.7 vs $2/$12 for Gemini 3.1 Pro, as of June 2026) so the latency-and-cost section has numbers to anchor on.
Generate a PRD outline for a feature that uses an LLM or AI model. Must include: model choice + fallback (name the specific model and version), prompt design, eval set, hallucination handling, latency + cost budget (per-call token cost), safety / abuse mitigation, data privacy, human-in-the-loop, kill switch. Output with sections marked v1 / v2.7. Comparison: 3 PRD shapes
For this feature, generate 3 alternative PRD outlines: (a) lean 1-pager, (b) standard PRD, (c) heavy-platform PRD. For each, list sections + estimated reading time. Recommend which one to use based on this context: {team size, risk, dependencies}.
Feature: {paste}8. “What is missing” PRD audit
Below is my PRD outline. Audit it against a senior PM checklist: missing sections, sections that hide multiple decisions, sections that should be 2 documents, sections that nobody will read. Output: what to add, what to merge, what to cut.
{paste outline}9. PRD outline from a single sentence
I will give you a one-sentence feature idea. Generate the full PRD outline that idea implies. Surface 5 hidden questions the sentence dodges. Outline format: markdown headings + 1-line description + 1 open question per section.
Sentence: {paste}10. Outline from a meeting transcript
Below is a 30-minute product-discussion transcript. Extract a PRD outline from it: what was decided, what was deferred, what is still in dispute. Mark each section with [decided] / [open] / [out of scope].
{paste transcript}11. Reorder for a skeptical reader
My current outline is in the order I thought of things. Reorder it for a skeptical exec who has 5 minutes. Lead with: problem + metric + ask. Push detail to appendix. Output the new order and explain each move in 1 line.
{paste current outline}12. Outline + estimated word count per section
Take this PRD outline and add a target word count per section. Total should not exceed 1,800 words. Mark which sections should be a table, list, or diagram rather than prose.
{paste outline}13. Outline + section owners
For this PRD outline, assign a draft owner role per section (PM, design, eng lead, data, legal, GTM). Mark any section that needs more than one owner and explain why.
{paste outline}14. PRD outline diff (old vs new)
Below are two PRD outlines for the same feature, written 2 weeks apart. Diff them: what was added, removed, renamed, restructured. Infer what the team learned in those 2 weeks. End with 3 sections still missing in both.
v1: {paste}
v2: {paste}15. Out-of-scope section drafted in isolation
Draft only the "Out of scope" and "Non-goals" sections for this PRD. Format: bulleted list with 1-line rationale per item. Include at least 5 items per section. Goal: stop scope creep before the PRD is even circulated.
Feature: {paste}Common mistakes
- Treating the outline as decoration — outline IS the PRD’s argument structure.
- Skipping “out of scope” — without it scope creeps for the whole quarter.
- No success metric section — without a number the PRD is a wish list.
- Burying open questions at the bottom — they belong near the top so reviewers can answer them.
- Using the same outline for a 4-week MVP and a 6-month platform build.
- Letting AI invent sections you never need (compliance, security) when the feature does not warrant them.
- Reordering the outline after the prose is written — always lock outline first.
How to push results further
- Always generate at least 2 outlines and pick — single-shot outlines lock you in.
- Mark each section v1 / v2 / never. Most v2 sections never get filled.
- Add target word counts per section to prevent the “infinite background” trap.
- Pair the outline with a 1-line “this PRD answers Q” stated up top.
- Run the audit prompt (template 8) on every outline before any prose is written.
- Keep a /outlines folder per squad and reuse the 3-4 patterns that win.
- For platform PRDs, the outline must name the deprecation / migration story or it is incomplete.
FAQ
- How does this differ from PRD draft prompts?: These prompts only produce the structural skeleton (sections plus one-line descriptions). Use PRD draft prompts to fill the prose afterward.
- Should every PRD use the same outline?: No. Use template 7 to pick the right shape (lean / standard / platform / growth) per work item. A standard feature PRD lands around 2–4 pages; a 4-week MVP should be one page.
- When should I lock the outline?: Before drafting any prose. Reordering after writing wastes more time than getting the outline wrong upfront.
- How long is too long?: Over 14 top-level sections and reviewers skim. Push detail to appendix sections instead of adding more H1s.
- Which model should I run these in?: Any current flagship works for the short outline prompts. For the audit (template 8) and transcript (template 10) prompts where you paste a long document, prefer a 1M-token model — Claude Opus 4.7, Sonnet 4.6, or Gemini 3.1 Pro (June 2026).
- Do I still need these if I use a dedicated PRD tool?: Tools like ChatPRD (free for 3 chats, Pro $15/mo, team seats $29/mo as of June 2026) scaffold a PRD from a description and integrate with Linear, Notion and Slack, but they impose one house structure — which is exactly the part most teams end up rewriting. Use these prompts when the feature is unusual or when a templated tool feels generic, then paste the chosen outline into the tool to draft the prose.
- What page length should I aim for?: Most PRDs land at 2–10 pages depending on complexity; a single feature is comfortable at 2–4 pages and a 4-week MVP should be one page. Optimize for clarity, not page count — push depth into appendix sections rather than more top-level headings.
Related
- PRD draft prompts
- Feature prioritization prompts
- MVP scope prompts
- Roadmap planning prompts
- Product, App & Startup Prompts hub
Tags: #Prompt #Product startup #PRD