By Volkier Bentinck, co-founder of Stellar Projects. Published on May 26, 2026.
AI prompts for Shopify are not all created equal. An amateur prompt generates generic text, while a well-constructed prompt yields publishable assets. The difference between the two is measured in conversion rates, hours saved, and brand consistency. This catalog brings together 30 operational prompts categorized by e-commerce use case, ready to paste into ChatGPT, Claude, Shopify Magic, or Sidekick. RCORCF method, 6 categories (product pages, SEO, emails, marketing, customer service, strategy), customizable variables, and expert advice for each prompt. No lead magnet, no email required: everything is in this article, copyable from the first read.
Table of Contents
- Why a catalog of prompts for Shopify
- RCORCF Method, the anatomy of a good e-commerce prompt
- 5 prompts for your Shopify product pages
- 5 SEO and GEO prompts for Shopify
- 5 prompts for email and Klaviyo loyalty
- 5 prompts for marketing and acquisition
- 5 prompts for customer service and operations
- 5 prompts for strategy and analytics
- Connecting these prompts to Shopify Magic and Sidekick
- Limitations and best practices
- Frequently asked questions about Shopify AI prompts
- The STELLAR Method book to go further
Why a catalog of prompts for Shopify
Generative AI has become a commodity; the prompt has become the differentiator. In 2026, every Shopify merchant has access to Shopify Magic, Sidekick, ChatGPT, and Claude. What distinguishes a brand that produces publishable content from one that produces noise is no longer the tool. It's the quality of the instruction given to the tool.
A study cited by Shopify in 2025 shows that users trained in prompt engineering achieve results 27% superior to those who improvise. The difference does not come from the model used, it comes from the framework the user provides. The new Brand Voice Cloning feature of Shopify Magic in 2026 illustrates this shift: Magic now learns your brand's voice from your existing articles, product pages, and emails, and automatically applies it to future generations. But this only works if your initial content is solid itself, and therefore derived from precise prompts.
This catalog addresses a concrete need: to stop reinventing the wheel with each task. If you write 50 product pages a month, you shouldn't have to formulate a new prompt every time. If you manage 10 Klaviyo email sequences, you need a reusable prompt framework, not one-off inspiration. The 30 prompts that follow are the result of real-world testing in production on Shopify stores for DTC brands, B2B, and traditional e-commerce. They are based on Stellar's comprehensive Shopify AI methodology for 2026 published elsewhere.
RCORCF Method, the anatomy of a good e-commerce prompt
Before the prompts themselves, there's the framework. A prompt that converts follows six elements in the same order, identifiable with the acronym RCORCF: Role, Context, Objective, Expected Result, Constraints, Format. Each element plays a precise role in the quality of the output.
Role. You specify to the AI who it embodies for this task. "You are an e-commerce copywriter specializing in women's fashion" is not the same mission as "You are an e-commerce analyst." The role conditions the vocabulary, references, and level of depth. Without a role, the AI produces average content, valid everywhere, distinctive nowhere.
Context. You provide the AI with information it cannot guess: the brand, the target, the tone, sector-specific constraints. The richer the context, the more usable the output without alteration. Context is the fuel of the prompt. Don't neglect it: if you save 30 seconds on the prompt by skipping the context, you'll lose 15 minutes rewriting.
Objective. You clearly state what you want to achieve. Not "Write a product page" but "Write a product page that convinces a target audience of women aged 30-45 to choose this product over a cheaper competitor." The objective is measurable, specific, and contextualized. This is where real utility comes into play.
Expected Result. You describe the output: length, structure, tone. Specifying "100 to 150 words with 3 benefit bullet points of 5 to 8 words each" prevents the AI from improvising. You gain consistency and direct publishability.
Constraints. You list what is forbidden. This is the most neglected and most profitable part. Prohibiting empty superlatives ("best, perfect, incredible"), anglicisms, typical ChatGPT phrases ("in the ever-evolving landscape") doubles the quality. List 3 to 6 constraints per prompt.
Format. You indicate the expected technical output: markdown, table, JSON, HTML, plain prose. This avoids manual conversions and allows direct pasting into Shopify, Klaviyo, or Notion. It's the least glamorous but most practical element for everyday use.
All the following prompts adhere to this structure. You can use them as is by customizing the {{variables between double brackets}}, or adapt them to your brand while maintaining the RCORCF structure.
5 prompts for your Shopify product pages
Prompt N°1, write a product description that converts
Objective: Generate a publishable product page in one go. Recommended model: Shopify Magic, Claude, or ChatGPT.
You are an e-commerce copywriter specializing in {{sector}}. Write a product description for {{product name}} intended for {{specific customer persona}}. Structure: 1 emotional hook sentence that addresses the main need, then 3 benefit bullet points of 5 to 8 words each, then 1 reassurance sentence (quality, warranty, return). Brand tone: {{3 adjectives}}. Constraints: 100 to 150 words total, prohibition of using the words "best, perfect, incredible, revolutionary," 1 action verb per bullet. Here are the technical characteristics: {{list characteristics}}.
Expert tip: never skip the customer persona. "Woman 35-45 in high socioeconomic status, Parisian, sensitive to sustainable design" yields a radically different page than "woman." This is the most profitable lever for 2 seconds of work.
Prompt N°2, 5 A/B variations to test the angle
Objective: Produce 5 different angles for the same product page, to be split tested on Shopify. Recommended model: Claude (best for creative diversity).
You are an e-commerce conversion optimization tester. Based on the current description of {{product}} below, propose 5 description variations of 100 to 150 words each targeting the same persona but exploring 5 different angles: 1) emotional and aspirational, 2) technical and expert, 3) social proof and testimonials, 4) urgency and scarcity, 5) practical daily benefit. Each variation keeps the same product characteristics but completely changes the seduction angle. Current description: {{paste current page}}.
Expert tip: test the 5 variations on Shopify with Rollouts (new in Winter '26) for at least 14 days. The winning angle is never the one you would have guessed.
Prompt N°3, product FAQ from customer reviews
Objective: Generate a Schema-ready FAQ from real customer questions. Recommended model: Claude (handles long review contexts well).
You are an e-commerce analyst. From the last 50 customer reviews of {{product}} below, identify the 6 to 8 most frequent questions buyers ask before purchasing. For each question, write an answer of 40 to 70 words that reassures AND provides concrete information (size, material, durability, compatibility). Output format: H3 question, then paragraph answer. Avoid promotional answers, stay factual. Reviews: {{paste reviews}}.
Expert tip: this FAQ feeds your Schema FAQPage (see prompt #8) and improves your chances of appearing in Google's People Also Ask and in ChatGPT responses.
Prompt N°4, compare two catalog products
Objective: Create a comparative table that helps decide between two similar products. Recommended model: ChatGPT or Claude.
You are an e-commerce advisor for the brand {{brand}}. Build a clear comparative table between {{product A}} and {{product B}} that helps a hesitant customer choose. Columns: criterion, product A, product B, who it's better for. Rows: price, main material, durability, main use, ideal target audience, 2 strengths, 1 point to note. Be honest about the differences. End with a summary sentence "Choose A if..., choose B if...". Characteristics A: {{list A}}. Characteristics B: {{list B}}.
Expert tip: insert this table on the more expensive product page to reduce doubt. This is an underutilized conversion lever on Shopify.
Prompt N°5, bulk rewrite short catalog pages
Objective: Bulk process pages under 50 words that hurt SEO and conversion. Recommended model: Shopify Magic (bulk action in admin) or Claude.
You are a Shopify catalog editor for {{brand}}. Here is a list of 10 currently short product descriptions (under 50 words): {{paste list with name and current description}}. For each product, rewrite the description in 100 to 150 words respecting the brand tone {{3 adjectives}}. Structure of each description: 1 hook sentence, 2 to 3 concrete benefits in continuous prose (no bullets), 1 reassurance sentence. Constraints: prohibition of superlatives, unnecessary anglicisms, and empty phrases like "perfect for you" or "look no further." Output: product name in bold + new description.
Expert tip: process in batches of 10 maximum to maintain quality. Beyond that, the AI fatigues and produces increasingly generic copies.
5 SEO and GEO prompts for Shopify
Prompt N°6, 5 A/B meta-titles for a product page
Objective: Quickly test multiple title angles to maximize CTR. Recommended model: ChatGPT.
You are a Shopify SEO copywriter. For the product page {{product name}}, propose 5 variations of meta-titles of 50 to 60 characters each. Constraints: the main keyword {{KW}} appears in the first 30 characters, each variation explores a different angle (promise, key figure, benefit, question, exclusivity), mandatory format "... | {{brand}}." Also provide the character count for each variation and flag any exceeding 60.
Expert tip: after publication, monitor CTR in Search Console and roll out the one that performs best to pages in the same category.
Prompt N°7, CTR-optimized meta-description
Objective: Write 3 meta-descriptions from three angles for A/B testing. Recommended model: ChatGPT.
You are an SEO copywriter. Write 3 variations of meta-descriptions of 145 to 160 characters each for the page {{URL or page title}} targeting the main keyword {{KW}}. Each variation includes the KW in the first half, contains a clear and differentiated promise, and ends without a period so as not to be cut off by Google. Variant A factual, variant B catchy, variant C benefit-oriented. Display the character count for each variant.
Expert tip: never reuse the same meta-description on 2 pages. Google penalizes even partial duplication.
Prompt N°8, FAQ Schema markup from a page
Objective: Produce a ready-to-paste JSON-LD FAQPage. Recommended model: Claude (best for structured JSON).
You are an SEO developer. From the content of the following page: {{paste content or URL}}, extract 6 to 8 questions visitors ask (based on actual content, not invented) and write answers of 50 to 100 words each. Then generate the complete JSON-LD compliant with the schema.org FAQPage schema, ready to paste into a script like application/ld+json. Verify that the JSON syntax is valid and that quotes are properly escaped.
Expert tip: systematically test JSON-LD on search.google.com/test/rich-results before publishing. A comma error breaks the entire rich snippet.
Prompt N°9, restructure an article for LLMs
Objective: Convert existing content to GEO-ready mode. Recommended model: Claude.
You are a GEO writer specializing in visibility in ChatGPT, Perplexity, Claude, and Gemini. Restructure the following article so that it is better cited by LLMs. Rules: 1) 3-sentence intro directly addressing the topic, 2) short sections of 200 to 300 words starting with the answer, then the context, 3) include 3 to 5 explicitly sourced key figures, 4) add a "key facts" bullet block every 3 H2s, 5) final FAQ with extractable questions. Article: {{paste article}}.
Expert tip: supplement this work with the complete GEO stack for Shopify, which details tools for measuring your presence in LLMs.
Prompt N°10, SEO brief for a new article
Objective: Frame an article before writing to avoid rewriting. Recommended model: Claude or ChatGPT.
You are an SEO project manager at a Shopify agency. Build a complete brief for a blog article on the topic: {{topic}}. The brief contains: 1) main KW and 3 secondary KWs with estimated volume, 2) dominant search intent (info, transact, navig), 3) unique editorial angle vs. the top 3 SERP competitors, 4) detailed plan with H1, 100-word intro, 6 to 8 H2s with IDs, 7-question FAQ, 5) internal linking suggestions (5 precise anchors), 6) external sources to cite (3 minimum, industry authorities), 7) JSON-LD Article + FAQPage + BreadcrumbList instructions.
Expert tip: this prompt replaces 90 minutes of brief work. Have it validated by a human before writing, especially on the editorial angle.
5 prompts for email and Klaviyo loyalty
Prompt N°11, 5-email Klaviyo welcome sequence
Objective: Build a welcome sequence that converts new subscribers. Recommended model: Claude or Klaviyo AI.
You are a Klaviyo email marketer. Build a 5-email welcome sequence for {{brand}}, for new newsletter subscribers who have not yet purchased. For each email: subject (max 50 characters), preview text (max 90 characters), body (150 to 250 words), 1 main CTA, Day +X since subscription. Imposed narrative progression: E1 welcome + brand promise, E2 manifesto and story, E3 flagship product and social proof, E4 expert advice without selling, E5 limited-time first purchase offer. Brand tone: {{3 adjectives}}.
Expert tip: Day +0, Day +2, Day +4, Day +7, Day +12 is the cadence that performs best on average. Adjust according to your purchase cycle.
Prompt N°12, post-purchase email with intelligent cross-sell
Objective: Naturally re-engage with a complementary product. Recommended model: Claude or Shopify Magic.
You are an email marketer. Write a post-purchase email sent 7 days after package receipt for {{marque}}. The customer just bought {{produit principal}}. The email should: 1) thank without flattery, 2) give 1 concrete tip for using the purchased product, 3) suggest 2 specific complementary products: {{cross-sell 1}} and {{cross-sell 2}} with a reason for each, 4) invite them to leave a review. Conversational tone, 200 to 280 words, 1 main CTA towards the cross-sell, discreet mention of the review.
Expert tip: The cross-sell that converts is not the most expensive product; it's the product that logically complements. Prepare your pairs in advance using prompt #23.
Prompt #13, 3-email abandoned cart sequence
Objective: Recover the order without being aggressive. Recommended model: Claude.
You are a Klaviyo email marketer. Build a 3-email abandoned cart sequence for {{marque}}. E1 (1h after abandonment): light tone, simple product reminder, no offer. E2 (24h): more urgent tone, reassurance (free returns, warranty, support), added micro-benefit. E3 (72h): last chance with a 10% off code valid for 24h, clear urgency without aggression. For each email: subject (50 chars), preview (90 chars), body 150 to 200 words, 1 CTA, mention D+X. Include cart contents as a dynamic Klaviyo variable.
Expert tip: Beyond 3 abandoned cart emails, the unsubscribe rate explodes. Stop at 72h, then move the customer into a nurturing sequence.
Prompt #14, win-back inactive customers 90 days
Objective: Re-engage the dormant base. Recommended model: Claude.
You are an email marketer for {{marque}}. Write a win-back email for customers who haven't purchased in 90 days. The email should: 1) open with a compelling question or observation (not a sales pitch), 2) remind them what the brand has recently changed or launched: {{nouveautés}}, 3) offer a coherent offer based on their purchase history: {{indication catégorie passée}}, 4) include 1 simple question they can answer by email reply. Subject 50 chars, preview 90 chars, body 250 to 320 words.
Expert tip: The email reply question is the underestimated lever. Any customer who replies, even negatively, is revived in your CRM.
Prompt #15, monthly newsletter in 5 blocks
Objective: Deliver a newsletter that doesn't feel like a constant promotion. Recommended model: Claude.
You are an email marketer at {{marque}}. Build a monthly newsletter with 5 blocks: 1) 80-word editorial signed by the founder on a current brand topic, 2) "new products" block with 2 to 3 products launched this month, 3) "best seller of the month" block with quantified social proof, 4) "expert tip" block useful without direct selling (150 words), 5) "event" block with an event or main CTA. Subject 50 chars, preview 90 chars. Tone: {{3 adjectifs}}. Current month: {{mois}}.
Expert tip: The expert tip block without selling is the only one that creates long-term retention. Never sacrifice it for an extra commercial insert.
5 prompts for marketing and acquisition
Prompt #16, product landing page brief
Objective: Frame a landing page before writing it. Recommended model: Claude.
You are a UX copywriter. Build a complete landing page brief for {{produit ou service Shopify}}. The brief contains: 1) short H1 max 7 words + subtitle 15 to 25 words, 2) hero block with differentiating promise and 2 CTAs, 3) "the problem" section 100 words, 4) "the solution" section with 3 benefit columns, 5) social proof section with format for 3 testimonials, 6) FAQ section with 6 questions, 7) final CTA section with legitimate urgency, 8) recommendations on background visual and tone of voice.
Expert tip: A successful landing page has 7 sections. Beyond that, you dilute. If you have more to say, write an article and link it to the landing page.
Prompt #17, 5 Instagram posts from a product sheet
Objective: Adapt a product into social content over a week. Recommended model: ChatGPT.
You are a social media manager for {{marque}}. Based on the product sheet {{nom}} below, design 5 complementary Instagram publications: 1) 5-slide carousel with progressive storytelling, 2) story with interactive quiz and poll sticker, 3) 30-second reel script scene by scene with hook, 4) editorial post with client quote as image, 5) UGC-style post with daily staging. For each publication: precise format, caption copy, 3 relevant non-saturated hashtags, 1-sentence visual suggestion. Product sheet: {{coller}}.
Expert tip: Alternate formats throughout the week. Five carousels in a row kill organic reach.
Prompt #18, Meta Ads copy with 3 variations
Objective: Produce testable ads on Meta. Recommended model: ChatGPT.
You are a Meta Ads media buyer for {{marque}}. Write 3 ad variations for the product {{nom}} targeting {{persona}}. Variation 1 short (title 30 chars, body 60 chars, CTA), to test on cold audiences. Variation 2 medium (title 40 chars, body 125 chars, CTA), for retargeting. Variation 3 long (title 40 chars, body 250 chars, CTA), for warm audiences. Each variation must: 1) start with an emotional hook or a strong observation, 2) include 1 concrete benefit with proof, 3) end with an action verb CTA. Prohibit the use of hollow superlatives.
Expert tip: Launch all 3 variations simultaneously in a split test for 7 days. The winner in CTR is not always the winner in ROAS.
Prompt #19, influencer collaboration brief
Objective: Frame a collaboration to avoid disappointments. Recommended model: Claude.
You are the partnerships manager for {{marque}}. Write a complete brief for a collaboration with influencer {{nom et profil}} on product {{nom}}. The brief contains: 1) brand context in 80 words, 2) clear objective and measurable KPIs, 3) precise deliverables (number of posts, format, deadline), 4) tone and key message to respect, 5) 3 mandatory points to mention, 6) 3 points absolutely not to say, 7) mandatory legal mentions (#ad, #partenariat), 8) financial terms if relevant, 9) content validation workflow before publication.
Expert tip: The 3 points not to say are the most important. Without them, you discover errors after publication.
Prompt #20, affiliate program, page and FAQ
Objective: Structure an affiliate program on the store. Recommended model: Claude.
You are the acquisition manager at {{marque}}. Write the affiliate program registration page and its FAQ. Page: H1, 80-word intro setting the promise, "who can join" section 100 words, "how it works" section in 4 clear steps, "commissions" section with transparent scale, registration CTA, legal mentions. FAQ: 8 questions on payment, tracking, cookies, duration, returns, taxation, exclusivity, support. Answers 50 to 100 words each.
Expert tip: Transparency on commissions is the number one enrollment lever. Hide the scale, and you lose good affiliates.
To leverage the new Shopify AI capabilities in your marketing campaigns, also see Shopify Magic and Sidekick new features unveiled at Winter '26, which open up new uses for Meta Ads and Agentic Storefronts.
5 prompts for customer service and operations
Prompt #21, 15 template answers to frequently asked questions
Objective: Populate Shopify Inbox with ready-made answers. Recommended model: Shopify Magic or Claude.
You are the customer service manager of {{marque}}. From the list of the 15 most asked questions below: {{coller liste}}, write a standard answer for each. Each answer must: 1) respect the brand tone {{3 adjectifs}}, 2) start with a short thank you or acknowledgment, 3) answer the question directly with concrete information, 4) suggest an action or useful link (return policy, product sheet, contact), 5) sign "The {{marque}} Team". Format ready to paste into Shopify Inbox.
Expert tip: Read each template aloud. A written answer can sound cold when spoken, and the best customers feel it.
Prompt #22, create a customer segment via Sidekick
Objective: Write a natural language prompt for Sidekick. Recommended model: Shopify Sidekick directly.
You are a Shopify e-commerce analyst. Build a prompt to give to Shopify Sidekick to create the following customer segment: {{description du segment, par exemple acheteuses femmes 30-45 ans, panier moyen supérieur à 100 euros, dernière commande il y a moins de 60 jours, abonnées à la newsletter}}. The prompt should be in natural language, precise enough for Sidekick to create the segment without back-and-forth, include a verification (how many customers in this segment?), and propose a marketing action to take on this segment immediately.
Expert tip: Sidekick loves numerical criteria. Avoid vague terms ("recently," "often"), give precise thresholds.
Prompt #23, analyze 100 reviews and extract signals
Objective: Transform reviews into a product action plan. Recommended model: Claude (large context management).
You are an e-commerce analyst. Here are the last 100 reviews for product {{nom}}: {{coller reviews}}. Analyze and structure your output into 6 blocks: 1) average rating and star distribution, 2) 5 most often cited strengths with verbatim, 3) 5 recurring points of friction or criticisms with verbatim, 4) requests for evolution or missing variations, 5) opportunities for adjacent products to launch, 6) 3 concrete actions to take this week to improve the product sheet and conversion.
Expert tip: Run this prompt every month on your top 10 products. This is product R&D at marginal cost.
Prompt #24, Shopify Flow workflow for returns
Objective: Automate return management on Shopify. Recommended model: Claude.
You are a Shopify ops manager. Build a Shopify Flow workflow to automate return management for {{marque}} orders. The workflow should: 1) trigger when a return request is created, 2) classify the return by reason via tag (size, defect, disappointed, other), 3) route to the correct process (auto exchange, refund, human escalation), 4) send the appropriate customer confirmation based on routing, 5) update stock and internal note. List logical conditions in Shopify Flow pseudo-code, ready to transpose into the editor.
Expert tip: Always keep a "human escalation" path for premium returns and VIP customers. Total automation kills relationships.
Prompt #25, customer service journey audit
Objective: Quickly diagnose customer service. Recommended model: Claude.
You are an e-commerce CX consultant. Audit the current customer service journey of {{marque}} on the following 5 points: 1) average response time (target less than 4 working hours), 2) available channels and their relevance (chat, email, WhatsApp, phone, form), 3) quality of responses (tone, exhaustiveness, first-contact resolution), 4) measured NPS and CSAT, 5) ratio of automatable versus human support. For each point: current observed state, gap vs. sectoral benchmark, priority action with estimated effort and impact (low, medium, high). Provided data: {{coller métriques disponibles}}.
Expert tip: A CSAT score alone means nothing without response cadence. Always cross-reference the two.
5 prompts for strategy and analytics
Prompt #26, complete Shopify store audit in 25 points
Objective: Provide a global diagnosis before an optimization sprint. Recommended model: Claude.
You are a senior consultant at a Shopify Plus agency. Conduct a 25-point audit of the store {{URL}} covering 5 areas (5 points per area): 1) Tech and performance (speed, Core Web Vitals, JS, mobile), 2) UX and conversion (purchase journey, checkout, mobile, reassurance), 3) Content and SEO (meta, structure, internal linking, schema), 4) Acquisition and marketing (channels, mix, CAC, creative), 5) Retention and operations (CRM, returns, NPS, post-purchase). For each point: score from 0 to 10, factual observation, priority action and impact estimate (low, medium, high). Output in a table ready for export.
Expert tip: Use this prompt in the presence of a human who knows the store. AI alone misses the strategic nuances specific to the brand.
Prompt #27, 3 customer personas from Shopify data
Objective: Build data-driven personas, not guessed ones. Recommended model: Claude.
You are a CRM manager. From the following exported Shopify data: {{coller exports avec tranches d'âge, panier moyen, fréquence d'achat, catégories préférées, géographie, canaux d'acquisition}}, build 3 distinct customer personas that cover 80% of the base. For each persona: 1) evocative name and typical photo, 2) precise demographics, 3) purchasing behavior (frequency, cart, preferred channel), 4) main motivations (3 max), 5) brakes or objections (3 max), 6) key message to address them, 7) priority recommended acquisition channels.
Expert tip: 3 personas is the right number. Beyond that, you dilute your campaigns. Below that, you miss profitable segments.
Prompt #28, pricing audit by category
Objective: Test the relevance of current pricing. Recommended model: Claude.
You are an e-commerce pricing analyst. Audit the pricing strategy of {{marque}} for the category {{catégorie}}. Provided data: {{coller liste produits, prix, marges, ventes 90 jours, prix concurrents}}. Expected output: 1) positioning vs. competitors with 3 key benchmarks, 2) observed elasticity by price range, 3) upsell and bundle opportunities (3 concrete ideas), 4) recommendations for price adjustments per product with quantified justification, 5) risk to brand image if adjustment made.
Expert tip: Price adjustments are made in increments of a maximum of 5%, never in one block. Measure the impact over 30 days before proceeding.
Prompt #29, 10 GEO and LLM opportunities for the brand
Objective: Identify ChatGPT and Perplexity queries to occupy. Recommended model: Perplexity (for live search) or Claude.
You are a GEO strategist. Identify 10 specific ChatGPT and Perplexity queries that target buyers of {{marque}} are likely to ask, and on which the brand should be visible. For each query: 1) exact phrasing of the question, 2) search intent, 3) probability that the brand is currently cited (low, medium, high), 4) concrete action to become a citable source (content to publish, authority signal to build), 5) priority from 1 to 5. Brand positioned on: {{positionnement en 2 phrases}}.
Expert tip: Manually test each query in ChatGPT and Perplexity before and after publication. This is the only true indicator of GEO progress. See also the 9 GEO tools to measure ChatGPT and Perplexity visibility.
Prompt #30, 90-day action plan for the brand
Objective: Structure a quarter's work. Recommended model: Claude.
You are a consulting director at a Shopify agency. Build a 90-day action plan for {{marque}}, knowing: positioning {{}}, priority KPIs {{}}, available internal resources {{}}, monthly budget {{}}. The plan must: 1) prioritize a maximum of 5 projects, 2) break down into monthly milestones D30, D60, D90, 3) for each project indicate owner, deliverable, target KPI, necessary resources, dependencies, 4) include 2 AI projects and 3 classic business projects, 5) end with a weekly review framework (who, when, format). Table output format.
Expert tip: A 90-day plan that changes every 2 weeks is not a plan, it's a shopping list. Lock down D30 and D60 milestones, keep flexibility only on D90.
Connecting these prompts to Shopify Magic and Sidekick
The 30 prompts above work on any LLM. But on Shopify, two native tools deserve dedicated use, as they are contextualized to your store without you having to manually re-enter data.
Shopify Magic is used in specific admin fields: product description, meta description, blog, integrated Klaviyo email, FAQ Inbox. Prompts #1, 5, 6, 7, 11, 12, 13, and 21 from the catalog are particularly suited for Magic use. You paste the prompt structure (without variables, which are filled in by the product context) into the Magic input area and you get an output consistent with your catalog. The Brand Voice Cloning feature added in 2026 learns your tone from your existing catalog, eliminating the need to repeat tone adjectives with each prompt.
Sidekick is used conversationally for cross-cutting tasks: creating customer segments, theme adjustments, KPI analysis, admin app generation. Prompts #22, 25, 26, and 30 are calibrated for Sidekick use. With Sidekick Pulse (Winter '26), the assistant automatically brings up weak signals on your dashboard: conversion drop, recurring carts, stock anomalies. You don't need to prompt; Pulse addresses the questions you need to deal with.
For the rest of the catalog (long-form writing, brainstorming, analysis), external models remain more effective. Claude excels at long-form writing and structured data analysis. ChatGPT is versatile and faster for short iterations. Perplexity is unmatched for competitive intelligence and sourced research. The right 2026 stack combines all three tools in addition to Magic and Sidekick. For details on the complete ecosystem, see the complete guide to Shopify AI in 2026.
Limitations and Best Practices
No prompt, however excellent, eliminates the need for human review. Three concrete limitations to integrate into your workflow.
Firstly, AI invents. For factual data (product composition, certifications, dimensions, legal mentions), systematic verification remains mandatory. A well-constructed prompt reduces hallucinations but never eliminates them. If you publish a product page that invents a certification or standard, you commit the brand. The Stellar rule: every number or technical characteristic goes through a second pair of eyes.
Secondly, AI standardizes. Without a strong brand framework, the outputs of the 30 prompts above will resemble those of your competitors using the same tools. Differentiation lies in the context you provide, not in the structural prompt. Invest 1 to 2 days in drafting a brand framework document (positioning, voice, forbidden words, references), then paste it at the beginning of each AI session. This is the most profitable lever in the catalog.
Thirdly, AI tires the reader. An unhumanized AI-generated product page makes you want to close the tab. Conventional phrasing, cliché transitions, rhythm-less lists: the markers are recognizable and harm conversion. Before publishing, run every generated content through the humanization grid described in the 7 levers for humanizing an AI-generated text. Passing through a detector is not the goal; the goal is for the reader to recognize your voice in the text.
On the operational safeguards side, four rules apply daily. One, create an archive file of your successful prompts, by use case. Two, date each version to be able to revert if the output degrades after a change. Three, measure the impact of prompts in production (conversion, CTR, NPS) and abandon those that don't perform. Four, refuse the race for models: changing LLMs every month is useless if your prompts are not stable.
Frequently Asked Questions about Shopify AI Prompts
What is an AI prompt for Shopify?
An AI prompt for Shopify is a structured instruction given to a generative AI model (Shopify Magic, Sidekick, ChatGPT, Claude, etc.) to produce specific e-commerce content or execute a task: product sheet, Klaviyo email, customer segment, FAQ Schema, store audit. A good prompt follows the RCORCF method (Role, Context, Objective, Expected Result, Constraints, Format) and uses customizable variables to be reusable from one product or project to another.
What is the difference between Shopify Magic and a ChatGPT prompt?
Shopify Magic is integrated into the Shopify admin and operates in specific fields (product description, Inbox email, meta). The prompt is partially pre-structured by your store's context. ChatGPT is external, versatile, but ignores your store's context unless you provide it. Magic saves time on repetitive integrated tasks, ChatGPT allows for broader uses (strategy, creative, long-form writing). The two complement each other.
Do I have to pay to use these prompts?
The 30 prompts in the catalog are free and publicly available. You can copy them directly from this article. As for tools, Shopify Magic and Sidekick are included in all Shopify plans. ChatGPT, Claude, and Perplexity have free plans sufficient for getting started, and paid plans (€20 to €30 per month) for intensive use. No additional specific budget for prompts.
How do I adapt these prompts to my brand voice?
Three steps. One, write a brand framework document (positioning, 3 tone adjectives, 5 forbidden words, 3 inspiration references). Two, paste it at the beginning of each AI session as permanent context. Three, adjust the variables {{ton de marque}} and {{persona}} in each prompt. On Shopify, Magic's Brand Voice Cloning feature learns your tone from your existing content and applies it automatically.
Can these prompts be used directly in Sidekick?
Yes, for conversational prompts (especially #22, 25, 26, 30). Sidekick understands natural language and executes the task directly within your Shopify admin. For long-form writing prompts (descriptions, emails, FAQs), Magic is more relevant in dedicated fields, or an external LLM (Claude, ChatGPT) if you want more control over the result.
Should content generated be humanized before publication?
Yes, systematically, even with an excellent prompt. Humanization removes AI markers (conventional phrasing, cliché transitions, rhythm-less lists) that harm conversion and brand voice. See the complete method in the Stellar guide on levers for humanizing an AI-generated text. Passing through an AI detector is not the goal; the goal is for the reader to recognize your brand in the text.
Should product facts be verified before publication?
Absolutely. AI can invent a certification, standard, dimension, or technical detail. All factual data (composition, size, warranty, legal compliance) must be verified by a human before publication. A well-constructed prompt reduces hallucinations but never eliminates them. The Stellar rule: every number goes through a second pair of eyes.
The STELLAR Method book to go further
Prompts are levers; they do not replace a brand strategy. Our book Creating Your Brand in the Age of AI (Florian Pohl and Volkier Bentinck, 2026) outlines the STELLAR Method over 280 pages, including a chapter dedicated to prompts as a lasting editorial asset. A must-read if you want to move beyond the tool race and build a brand voice that endures over time.
To turn these prompts into an internal tool reusable by your entire team, our Shopify AI Studio develops assistants and automations tailored to your processes.
Discover the STELLAR Method book
