The Shopify AI prompt library every founder should keep

If you only build one AI asset in 2026, build a prompt library. Not a vague Notion doc with screenshots — a real, versioned, organized library your whole team copies from. Below are the twelve prompts we install in nearly every Shopify implementation.

Each is in the same shape: clear role, structured input, explicit constraints, opinionated output format. Customize the curly-brace fields and ship.

1. Product description (single SKU)

Write a Shopify product description.

Brand voice: {block}

Product: {title}
Type: {type}
Top 3 specs: {specs}
Target customer: {customer}
Focus keyword: {keyword}

Constraints:
- 100-130 words
- First sentence hooks on use, not specs
- Middle: 3 short benefit bullets
- Closing: soft CTA, no exclamation marks
- Avoid: "premium", "high-quality", "unique"

Output the description only.

2. Product description (batch table)

Generate descriptions for the 25 products below. Return a TABLE with
columns: handle, description.

Brand voice: {block}
Constraints: {same as #1}

Products:
handle | title | type | specs | customer | keyword
...

3. Bulk metadata generator

For each product below, generate:
- SEO title (50-60 chars, focus keyword first)
- Meta description (140-160 chars, benefit + USP + soft CTA)

Return as a TABLE: handle, seo_title, meta_description.

Products:
{table}

4. Collection page intro (150 words)

Write a Shopify collection page intro.

Collection: {name}
Focus keyword: {keyword}
Customer: {customer}
Brand voice: {block}

Constraints:
- 2 paragraphs, ~150 words
- Focus keyword in first sentence, naturally
- Answer 1 buyer question (e.g. "what to look for")
- Soft handoff to products
- Avoid "wide range", "premium quality", "unique"

5. Product FAQ block (JSON for metafield)

Generate 6 FAQs for a Shopify product page about {product}.

Selection:
- Questions Google shows in "People Also Ask"
- Questions buyers actually type
- Skip trivia (delivery, returns) unless product-specific

Per FAQ: question (max 12 words), answer (40-70 words, helpful).

Output ONLY valid JSON for a metafield (faqs.product, type JSON):
[ { "q": "...", "a": "..." }, ... ]
No markdown fences. No commentary.

6. Helpdesk draft reply

You are customer support at {brand}. Draft a reply to the email below.

Rules:
- Under 120 words
- {tone}
- End with one concrete next step
- Do not state amounts you cannot verify
- If unclear, suggest a teammate follows up

Policy:
{policies}

Reply examples:
{3 great examples from your history}

Customer email:
"""
{email}
"""

7. Review classifier (for Flow webhook)

Classify this Shopify product review.

Review: "{review}"

Output ONLY JSON:
{
  "sentiment": "positive|neutral|negative",
  "topic": "product|delivery|service|pricing|other",
  "score": 1-5,
  "summary": "max 15 words"
}

8. Order risk screener (for Flow + AI)

Risk-detection only, no judgment.

Order:
- Total: ${{total}}
- First-time buyer: {{first_time}}
- Country: {{country}}
- Items: {{items}}

Customer history: {{history}}

Output ONLY JSON:
{
  "risk": "low|medium|high",
  "reason": "max 80 chars"
}

9. Review mining (research mode)

Analyze the reviews below. Provide:

1. TOP 5 POSITIVE THEMES — title + 1 quote + frequency.
2. TOP 5 RECURRING COMPLAINTS — title + 1 quote + frequency + severity 1-5.
3. 3 PRODUCT IMPROVEMENTS — concrete changes 3+ customers suggest or imply.
4. 5 COPY-READY LINES — quotes usable on product pages or ads.
5. 3 OBJECTIONS — pre-purchase doubts inferable from reviews.

Reviews:
{reviews}

10. Blog outline (long-form)

Keyword: "{keyword}"
Search intent: {informational|commercial}
Target word count: {n}

Outline:
- H1 with keyword
- 5 H2s covering search intent fully
- Per H2: 2-3 bullets describing the section
- Per H2: an internal-link suggestion (which Shopify pages?)
- 5 FAQs at the bottom

11. Subject line generator (Klaviyo)

Write 10 subject-line variants for: {goal}.

Tone: {tone}. Max 45 chars. AVOID: exclamation marks, emojis,
"exclusive", "discount".

Distribution:
- 2 curiosity
- 2 benefit
- 2 question
- 2 specific (color/category)
- 2 calm urgency

Per variant: 1 sentence on why it works.

12. Product page critic

You are a senior CRO specialist for Shopify stores. I'll paste a
product page. Critique in 6 points:

1. FIRST SENTENCE: hook or list?
2. BUYER DOUBTS: 3 doubts the page does not address.
3. TRUST SIGNALS: which forms of social proof are missing?
4. SEO: likely focus keyword, present? naturally? 2 missing synonyms?
5. CTA: clear? suggestion to improve?
6. IF I COULD CUT 1 PARAGRAPH: which and why?

End with a rewritten first paragraph in this brand voice: {voice}.

Page:
"""
{page}
"""

How to actually use this library

Three rules we enforce inside our implementations:

1. Keep it in one place. A shared Notion or Claude Project. Not 17 ChatGPT chat histories.

2. Version it. When you discover AI keeps using “premium” and you hate it: add to the avoid list, increment a v2 tag. Track what changed.

3. Train your team to fork from this library. No one prompts from scratch. Every new task is a fork of an existing prompt.

A library used by 3 people for 12 months easily beats a tool that just wrote one product description well that one time.

Want us to ship a Shopify-tuned prompt library for your team? Book a free Shopify AI audit — we’ll deliver 15-20 prompts customized to your store as part of the audit deliverable.