TL;DR
Feed an AI assistant five things — the one-line idea, the user problem with evidence, the OKR it ladders into, hard constraints, and what is out of scope — then ask for a fixed nine-section PRD. The AI is good at the shape (it never forgets non-goals, metrics, or open questions) and bad at the content (what the feature should do). Every placeholder it writes is a decision you still owe. For a one-off, Claude Opus 4.7 or GPT-5.5 in the chat app is enough; if you write PRDs weekly, a PM-specific tool like ChatPRD (Pro $15/mo or $179/yr as of June 2026) adds custom templates, file uploads, and Notion/Linear/Slack sync. Always circulate the draft for engineering, design, and data sign-off before anyone estimates it.
The task
You have a feature idea in your head (or buried in a Slack thread) and you need it in PRD form before the next planning meeting. The PRD has to be specific enough that engineering can estimate it, narrow enough that design can mock it, and honest enough that leadership can rank it against everything else in the queue. Without a PRD the conversation circles for weeks. With a vague PRD, the wrong thing gets built — and you find out in code review.
When AI helps, and when it does not
AI is excellent at turning a paragraph of context into the right shape of a PRD. It remembers to ask for non-goals, success metrics, edge cases, and open questions even when you forget. It is poor at deciding what the feature should actually do, what the real success metric is, or which constraint is non-negotiable.
Treat the AI draft as a forcing function: every section it fills with placeholder text is a section you still owe an answer on. Do not let AI invent product requirements you did not request, especially around legal, pricing, or data handling — in those areas a confident-sounding fabrication is worse than a blank.
In a side-by-side test that PM Tom Leung ran across ChatGPT, Claude, Gemini, Grok, and ChatPRD (published December 2025), Claude ranked first on strategic depth and metric specificity — it proposed product-aware measures like “concept mastery rate” where ChatGPT defaulted to a generic “engagement rate.” ChatPRD came second on structure, Gemini third, ChatGPT fourth (“reliable but felt like an average of everything”), Grok last. The practical takeaway: the model matters less than the prompt and the evidence you feed it.
What to feed the AI
- The one-line feature idea, in plain language
- The user problem, plus any quotes, support tickets, or behavioural data that motivated it
- Business context — which OKR or strategic bet this ladders into
- Hard constraints: deadline, headcount, tech stack, compliance regime (SOC 2, GDPR, HIPAA)
- What is explicitly out of scope — anything you can rule out makes the PRD twice as useful
The single biggest quality lever is evidence. A model handed three real support tickets writes a sharper problem statement than one handed the sentence “users are confused.”
Copy-ready prompt
Paste this into Claude Opus 4.7, GPT-5.5, or Gemini 3.1 Pro. Replace every [bracket] with your own input — the brackets are placeholders, not literal text.
You are drafting a PRD for an internal product team.
Idea: [one line]
User problem and evidence: [quotes, tickets, data]
Business context / OKR: [the bet this ladders into]
Hard constraints: [deadline, stack, compliance]
Out of scope: [list]
Return a PRD with these sections, in this order:
1. Problem statement (3-5 sentences, no jargon)
2. Goals (3 max, each measurable)
3. Non-goals (at least 3 — what we are explicitly NOT doing)
4. Target users and primary use cases
5. Functional requirements (numbered, each testable in one sentence)
6. Non-functional requirements (latency, accessibility, privacy)
7. Success metrics (leading + lagging, with target numbers and a window)
8. Open questions (decisions eng or design still owes, each with an owner)
9. Rollout plan (feature flag, cohort, kill switch)
Where you do not have evidence, write "TBD — owner: [role]" instead of guessing.
Do not invent metrics, pricing, or compliance requirements I did not give you.For a sharper draft, run a second pass for adversarial review:
Now critique this PRD as a skeptical staff engineer.
Which assumptions would sink the project if wrong?
Which success metric is gameable, and how?
List the three riskiest open questions in priority order.Recommended output structure
A skim-readable PRD: one H2 per section, bullet lists for requirements, a table for metrics, and an “Open questions” list with an owner on every row. An engineer should be able to find the functional requirements without reading the strategy section.
The metrics table is the one place numbers live, so make it explicit:
| Metric | Definition | Current (baseline) | Target | Window |
|---|---|---|---|---|
| Onboarding completion | % of new users who finish setup | 61% | 75% | Q3 2026 |
| Time to first value | Median minutes to first saved doc | 9 min | under 4 min | Q3 2026 |
| Support tickets / 1k users | Tickets tagged “setup” | 14 | under 6 | 60 days post-launch |
Every row needs a baseline. A target with no baseline (“increase engagement 25%”) is the most common AI-generated tell, and stakeholders will catch it.
Which AI tool to use
| Need | Tool | Cost (June 2026) | Why |
|---|---|---|---|
| One-off PRD, you already pay for a chat app | Claude Opus 4.7 or GPT-5.5 | $20/mo Pro/Plus | Strong long-context drafting; paste tickets and docs inline |
| Long PRD with many pasted docs | Gemini 3.1 Pro | $19.99/mo (Google AI Pro) | 1M-token context handles bulky research dumps |
| You write PRDs weekly | ChatPRD | Free / Pro $15 / Teams $29 seat | PM-specific templates, file uploads, gap analysis, Notion/Linear/Slack sync |
| Draft beside the code | Claude Code | bundled in Claude Pro $20 | Reads the repo so requirements match what already exists |
ChatPRD’s Free tier caps you at three chats on a basic model; Pro ($15/mo or $179/yr as of June 2026) unlocks unlimited documents, premium models, custom templates, file and image uploads, and integrations with Notion, Linear, and Slack; Teams adds shared workspaces and admin controls at $29/mo per seat. For most people, a general assistant you already pay for is plenty — reach for a specialised tool only when PRD drafting is a recurring weekly job. See the model picks and tier details in our AI tools for product managers guide.
How to check the draft is usable
- Every goal has a number, a window, and a way to measure it
- Non-goals exist and are specific (not “we’re not building a CRM”)
- Every functional requirement is testable in one sentence
- The TBDs have owners, not just question marks
- Nothing in “Out of scope” reappears as a requirement
- The metrics table has a real baseline in every row
Common mistakes
- Letting AI invent metrics like “increase engagement by 25%” with no baseline
- Shipping a PRD with no non-goals — without them, scope quietly expands in implementation
- Mixing functional and non-functional requirements into one bullet list
- Pasting the AI draft into Notion without removing TBDs — stakeholders treat placeholders as agreed scope
- Writing it solo and shipping it. Use AI to draft, but circulate for engineering, design, and data sign-off before anyone estimates it
- Asking the model to estimate engineering effort. It will give you a confident number based on nothing
FAQ
- Should the PRD include UI mocks? No. PRDs describe behaviour, not pixels. Link the Figma file from “open questions” until design lands.
- How long should a good PRD be? One screen of bullets per section. If the PRD runs to 12 pages, you have a strategy doc, not a PRD.
- Can AI estimate engineering effort? No, and you should not ask it to. Effort estimates need someone who has touched the codebase.
- Which model writes the best PRD? In a five-tool PM head-to-head published December 2025, Claude ranked first on strategic depth and metric specificity, ChatPRD second on structure, Gemini third, ChatGPT fourth (reliable but generic), Grok last. Even so, the prompt and the evidence you provide matter more than the model.
- Is it safe to paste internal tickets and customer data into a chatbot? Only on a business plan with data-retention controls, and never paste regulated data (PII, PHI, payment details). Strip identifiers first, or use a tool your company has already cleared.
Related
- PRD draft prompts — alternative prompt phrasings and PRD section templates
- User story writing — turn PRD requirements into agile user stories
- Feature prioritisation — once you have a PRD, decide whether to build it now
- PRD outline prompts — a lighter-weight outline before the full PRD