Shopify Magic Prompts for Product Pages: 25 Copy Templates

Why this matters

Product pages are where most Shopify stores win or lose conversion. In practice, “AI copy” fails for two predictable reasons: (1) missing inputs (specs, policies, voice), and (2) missing review gates (facts, claims, tone, SEO). This prompt library fixes both by giving you bounded templates that work with Shopify Magic (or any assistant) without hallucinating guarantees.

Primary KPI: conversion rate (PDP → add-to-cart) and revenue per session.
Quality KPI: return rate / “item not as described” tickets (copy accuracy proxy).
Efficiency KPI: time-to-publish per SKU (minutes) and % of SKUs refreshed per month.

Framework / workflow

Use a simple “inputs → generate → QA → publish → measure” loop. The goal is not perfect prose; it’s consistent, factual, on-brand pages at scale.

Step 0 — Build your inputs pack (copy can’t outrun missing data)

Catalog facts: materials, dimensions, care, compatibility, what’s included, variants, constraints (e.g., “not dishwasher safe”).
Proof assets: reviews, warranty terms, certifications, lab results (only if you can substantiate).
Policies: shipping windows, returns, warranty wording (paste the exact text you want copy to match).
Brand voice: 3 example PDPs you like + “banned phrases” list.
SEO target: primary query + 3–8 supporting phrases (avoid stuffing; use naturally).

Step 1 — Generate in blocks (don’t ask for “a full page”)

Generate the page as blocks (title, bullets, description, FAQ, trust block). This makes QA faster and reduces model drift.

Step 2 — Human QA gates (non‑negotiable)

Fact check: every claim maps to a catalog field or proof asset.
Policy alignment: shipping/returns/warranty wording matches your policy text.
Claim control: no medical/legal claims; no “guaranteed results”; no invented certifications.
SEO sanity: primary keyword appears naturally in title + first paragraph; no repetitive stuffing.
UX: benefits before specs; scannable bullets; clear CTA; reduce cognitive load.

Step 3 — Publish and measure (7 days, 30 days)

7‑day checks: add-to-cart rate, scroll depth, search queries on-site, customer questions.
30‑day checks: conversion lift vs baseline, returns/tickets, organic impressions & CTR (if indexed).

Templates / prompts

Each template is intentionally strict. Replace bracketed fields with your data pack. If you don’t know a fact, instruct the model to leave it blank or ask a question.

Universal “system prompt” (paste once)

System / guardrails

Role: You are an ecommerce copy editor for a Shopify store.
Rules:
- Use ONLY provided facts. If a detail is missing, write "[NEED DATA]" and ask 1 question.
- Do NOT invent certifications, test results, guarantees, shipping times, or warranty terms.
- Avoid medical/legal claims. Keep claims soft unless proof is provided.
- Match brand voice: [voice notes] and avoid: [banned phrases].
- Output in the requested format with clear headings.

1) Titles (5)

Input: Product name, primary keyword, 2 differentiators, variant rule.
Task: Write 5 SEO-safe titles (max 70 chars). 
Constraints: include primary keyword once; no superlatives unless proof; no "best".
Output: list of 5 titles.

Task: Write 3 titles for gifting intent.
Inputs: [recipient], [occasion], [differentiator], [material], [color options].
Constraints: factual; no shipping promises.
Output: 3 titles + 1-line rationale each.

Task: Write titles for compatibility-driven products (e.g., cases/accessories).
Inputs: [compatible models], [not compatible list], [material], [feature].
Output: 5 titles + "Compatibility note" line for PDP.

Task: Create 5 collection-friendly titles (consistent structure).
Inputs: [collection naming pattern], [product name], [key attribute], [size].
Output: 5 titles that follow the pattern exactly.

Task: Rewrite my current title for clarity.
Input: Current title: "[paste]"; Facts: [paste].
Constraints: keep meaning; remove hype; keep under 70 chars.
Output: 3 rewrites + best choice (1 sentence why).

2) Benefit bullets (5)

Task: Write 6 scannable bullets (benefit → proof/fact).
Inputs: [top 3 pains], [features], [materials], [what's included].
Constraints: each bullet max 14 words; include at least 2 numeric specs if provided.
Output: 6 bullets.

Task: Write bullets for sizing/fit.
Inputs: [size chart], [model measurements], [fit notes], [care].
Output: 6 bullets + 2 "fit FAQs" questions.

Task: Write bullets for durability & maintenance (no false claims).
Inputs: [material], [care], [warranty text], [tests/proof if any].
Constraints: if no proof, use soft language ("designed to", "built for").
Output: 6 bullets + 1 maintenance tip box.

Task: Write bullets for bundles/sets.
Inputs: [bundle contents], [savings rule], [use cases], [gift option].
Constraints: do not claim savings unless math provided.
Output: 6 bullets + 1 bundle comparison line.

Task: Turn technical specs into human benefits.
Inputs: [spec list].
Output: table with 2 columns: Spec | What it means for the customer.

3) Long description (5)

Task: Write a 180–220 word PDP description.
Structure: Hook (pain) → Solution → Proof/facts → What's included → Care → CTA.
Inputs: [facts], [policy snippets], [voice notes].
Constraints: no invented guarantees; no shipping promises.
Output: one paragraph + short subheads.

Task: Write a "quick read" description for mobile (90–120 words).
Inputs: [top benefits], [2 specs], [who it's for], [compatibility].
Output: 1 paragraph + 4 bullets.

Task: Write a comparison paragraph vs common alternatives.
Inputs: [alternative types], [your differentiators], [limits/cons].
Constraints: be fair; include one limitation; no competitor naming.
Output: 120–160 words.

Task: Create an "Ingredients / materials" truth section.
Inputs: [materials], [origin], [certifications if proven], [care].
Output: bullet list + "What we don't use" line (only if true).

Task: Rewrite my existing description to remove fluff.
Input: Existing copy: "[paste]"; Facts: [paste].
Constraints: keep claims factual; keep voice; reduce by ~30%.
Output: revised copy + list of removed unsupported claims.

4) FAQ (5)

Task: Generate 7 PDP FAQs from real buyer objections.
Inputs: [top objections], [policies], [compatibility], [care], [what's included].
Constraints: answers 2–3 sentences; include policy wording exactly where relevant.
Output: Q/A list.

Task: Create return-prevention FAQs.
Inputs: [common return reasons], [fit issues], [usage mistakes], [care].
Output: 6 FAQs that reduce misunderstandings (no blame).

Task: Create shipping & delivery FAQs aligned to policy.
Inputs: [shipping policy text], [cutoff times if any], [regions].
Constraints: do NOT add timelines not in the policy.
Output: 5 FAQs.

Task: Create warranty/support FAQs aligned to policy.
Inputs: [warranty text], [support hours], [contact method].
Constraints: no new promises; match policy language.
Output: 5 FAQs.

Task: Convert messy spec notes into FAQs.
Inputs: [spec notes].
Output: 6 FAQs + "If you're unsure, choose X" guidance (factual only).

5) Merchandising blocks (5)

Task: Write an upsell block for a compatible add-on.
Inputs: [add-on product], [compatibility], [benefit], [bundle rule].
Output: 60–90 words + 3 bullets + CTA label.

Task: Write a cross-sell block for "frequently bought together".
Inputs: [3 products], [use case], [who it's for].
Constraints: no discount claims unless provided.
Output: 80–110 words + 3 bullets.

Task: Write a trust block that uses only provided proof.
Inputs: [review snippets], [certifications], [warranty text], [brand story facts].
Output: 3 short sections: Proof | Risk reversal | Brand note.

Task: Write a "How to use" block.
Inputs: [steps], [dos], [donts], [care].
Output: numbered steps (max 6) + troubleshooting mini-FAQ (3 Qs).

Task: Create a variant selector helper copy.
Inputs: [variant options], [differences], [recommended for], [not for].
Output: short guidance under selector + 3 FAQs.

Execution layer: prompt QA system

Shopify Magic prompts work best when they include the same structured inputs every time: product facts, audience, objections, proof, tone, keywords, and forbidden claims.

Save prompt versions by product type so winning patterns can be reused.
Ask for multiple variants, but publish only after checking claims against source facts.
After publishing, compare edited PDPs against a control group to see whether the copy improves conversion or only sounds more polished.

Checklist

PDP publish gate (pass/fail)

Facts: every spec and claim maps to catalog/proof. No “guaranteed” language.
Policies: shipping/returns/warranty copy matches your policy text verbatim where referenced.
SEO: primary query used naturally; meta description is clear; avoid keyword repetition.
UX: first screen communicates who it’s for + key benefit; bullets are scannable.
Trust: includes proof assets (reviews/certs) only if you have them; otherwise remove.
Internal links: include at least one link to Shopify AI and one to Getting Started.

FAQ

Can Shopify Magic write product descriptions automatically?

Yes, but quality depends on your inputs pack. If you only provide a product name, you’ll get generic copy. Provide specs, policies, and voice examples to keep output accurate and on-brand.

How do I prevent hallucinated claims?

Use the “facts only” rule, require a [NEED DATA] draft for missing fields, and run a publish gate where every claim maps to a catalog field or proof.

Should I generate the whole page at once?

No—generate in blocks (title, bullets, description, FAQ). You’ll get more consistent results and faster QA.

When should I change robots to index,follow?

When the examples are specific, policy language is verified, and the final page has been checked against source facts and internal links.

Ready to build with Shopify + AI?

Start with Shopify’s foundation, then layer AI workflows where they’re measurable and safe.

Start Shopify Free Trial See the 90-Day Plan