Mastering e-commerce FAQ generation with Claude is one of the most effective ways to reduce repetitive support tickets for DTC brands. If you run an online store, you know the pattern: an order ships, and within 48 hours your inbox fills with the same three questions: “Where is my order?”, “How do I return this?”, and “Will this fit me?”
This guide gives you a reusable Claude workflow for e-commerce customer service FAQ generation. You feed Claude 3.5 your real policies, and it returns a structured FAQ library with customer-facing answers, agent macros, escalation rules, and proactive delay messaging. It doesn’t invent policies or write vague, generic filler.
Table of Contents
Why This Prompt Works for e-commerce FAQ generation with Claude
Most “write me an FAQ” prompts fail because they give the AI too little ground truth. Claude fills the gaps with plausible but wrong details, like a 60-day return window you don’t offer. This prompt is built to prevent that:
- Grounded inputs first. You supply your actual shipping, returns, and sizing data, and Claude is told to use only that.
- Explicit “no guessing” rule. Missing information gets flagged as [NEEDS INPUT] instead of being fabricated. This matters most for refund timelines and carrier promises.
- Multiple customer phrasings. Customers don’t write questions the way brands do. Each FAQ includes real-world variants, which helps with help-center search, chatbot training, and macro matching.
- Three output layers per FAQ. You get a short answer for your site, a fuller version for email or chat, and an internal agent note with escalation triggers. One generation covers deflection and handling.
- Built-in self-check. Claude audits its own output for policy conflicts, tone drift, and unsupported claims before finishing.
- Role and tone control. Your brand voice is defined up front, so answers sound like you rather than a generic corporate bot.
The Master Prompt
Copy the prompt below, replace every {{VARIABLE}} with your own information, and paste it into Claude 3.5 Sonnet.
You are a senior e-commerce customer experience strategist and support copywriter. You specialize in reducing repetitive post-purchase tickets for direct-to-consumer brands while protecting customer trust and brand voice.
YOUR GOAL
Create a complete, ready-to-publish FAQ library and agent macro set for {{BRAND_NAME}} that deflects repetitive post-purchase tickets about shipping delays, returns/exchanges, and sizing/fit.
[BRAND CONTEXT]
- Brand name: {{BRAND_NAME}}
- Products sold: {{PRODUCT_TYPES}}
- Customer profile: {{CUSTOMER_PROFILE}}
- Brand voice: {{BRAND_VOICE}} (example: warm, direct, lightly playful, never corporate)
- Support channels: {{SUPPORT_CHANNELS}} (example: email, live chat, Instagram DMs)
- Support hours and average response time: {{SUPPORT_HOURS_AND_SLA}}
[SOURCE OF TRUTH: POLICIES AND DATA]
Shipping:
{{SHIPPING_POLICY}}
(Include: processing time, carriers, delivery estimates by region, tracking details, international rules, what happens with lost or delayed packages, current known delays.)
Returns and exchanges:
{{RETURNS_POLICY}}
(Include: return window, condition requirements, who pays return shipping, refund method and timeline, exchange process, final-sale items, return portal link.)
Sizing and fit:
{{SIZING_INFO}}
(Include: size chart, fit notes per product, model measurements, common fit complaints, fabric stretch or shrinkage details.)
Top customer complaints or ticket themes from the last 90 days:
{{TOP_TICKET_THEMES}}
[STRICT RULES]
1. Use ONLY the information provided in the SOURCE OF TRUTH section. Never invent policies, timeframes, fees, or guarantees.
2. If information needed for a good answer is missing, write [NEEDS INPUT: describe what is missing] in that spot instead of guessing.
3. Never blame the customer or the carrier. Be accountable, specific, and helpful.
4. Write at an 8th-grade reading level. Use short sentences and plain language.
5. Where order-specific details are needed, use placeholders such as [ORDER_NUMBER], [CUSTOMER_FIRST_NAME], [TRACKING_LINK], [RETURN_PORTAL_LINK].
6. Match this brand voice exactly: {{BRAND_VOICE}}.
[TASK: BUILD THE FAQ LIBRARY]
Organize the output into three categories: (A) Shipping Delays and Tracking, (B) Returns, Exchanges and Refunds, (C) Sizing and Fit.
For each category, create {{FAQS_PER_CATEGORY}} FAQs. Prioritize questions that match the ticket themes provided. For EACH FAQ, use this exact format:
FAQ TITLE: (the question, written the way a customer would ask it)
CUSTOMER PHRASING VARIANTS: (3 alternate ways customers actually word this, including casual or frustrated versions)
SHORT ANSWER (for website FAQ page): (40 words or fewer)
FULL ANSWER (for email or chat): (80 to 150 words, with clear next steps)
AGENT MACRO: (a ready-to-send reply with placeholders and a friendly sign-off)
ESCALATION TRIGGER: (specific conditions under which an agent should stop using the macro and escalate, for example a package stuck over 10 days, a damaged item, or a repeat contact)
POLICY REFERENCE: (quote or paraphrase the exact line from the source material that supports the answer)
[TASK: PROACTIVE MESSAGING]
After the FAQ library, write:
1. A shipping delay notification email (subject line, preview text, body) that tells the customer about a delay before they have to ask. Include a clear new timeline or [NEEDS INPUT] if unknown.
2. A post-delivery sizing check-in message that invites customers to exchange before the return window closes.
3. A returns confirmation message that sets expectations for refund timing.
[TASK: GAP ANALYSIS]
End with a section called "Gaps and Risks" containing:
- Every [NEEDS INPUT] item in one checklist
- Any contradictions or ambiguities you found in my policies
- 3 suggested policy or product-page changes that would likely reduce ticket volume further
[SELF-CHECK BEFORE FINALIZING]
Review your own output and confirm: (a) no claim appears that is not in the source material, (b) tone is consistent, (c) every FAQ has all required fields, (d) escalation triggers are specific and measurable. Fix any problems silently, then output the final version.
[OUTPUT FORMAT]
Use clear headings for each category. Do not add introductions or commentary before or after the deliverable.How to Customize & Use It
Fill in the variables carefully
- {{BRAND_NAME}}, {{PRODUCT_TYPES}}, {{CUSTOMER_PROFILE}}: Keep these short and concrete. “Sustainable activewear for women, ages 25 to 40” works better than “clothing for everyone.”
- {{BRAND_VOICE}}: Describe it in three or four adjectives, and add a sample sentence from your site or a past email if you can. Claude 3.5 mimics voice well when it has an example.
- {{SHIPPING_POLICY}}, {{RETURNS_POLICY}}, {{SIZING_INFO}}: Paste your actual policy text, size charts, and fit notes. This is the most important step. The more precise your source material, the fewer [NEEDS INPUT] flags you’ll get.
- {{TOP_TICKET_THEMES}}: Export 50 to 100 recent tickets from Gorgias, Zendesk, or Shopify Inbox, ask Claude to summarize the top 10 themes, and paste the result here. This keeps the FAQs aimed at real volume.
- {{FAQS_PER_CATEGORY}}: Start with 6 to 8. That’s enough to cover most repeat questions without making review a chore. If you want longer, richer answers, run one category at a time (see Step 3 below).
Step-by-Step: How to Run This Workflow
Gather your source material. Collect your shipping policy, returns policy, size chart, and fit notes. Paste them as plain text. Remove anything outdated before you start.
Export your ticket themes. Pull 50 to 100 recent tickets, ask Claude to list the top 10 recurring themes, and keep that list for the {{TOP_TICKET_THEMES}} field.
Run one category at a time. Paste the master prompt and add this line at the end: “Generate category (A) only. Wait for me to type ‘next’ before continuing.” Shorter runs produce sharper answers and make review easier.
Resolve every [NEEDS INPUT] flag. Fill the gaps with real information, then ask Claude to regenerate only the affected FAQs.
Review, then publish. Read every answer against your actual policy. Publish the short answers on your FAQ page and load the macros into your help desk.
Before and After: What the Output Looks Like
Here is a simplified, illustrative example for a fictional activewear brand. The numbers below are placeholders, not real policy.
Before (generic prompt: “Write an FAQ about shipping delays”):
Q: What if my order is late?
A: We’re sorry for the delay! Orders usually arrive within 5 to 10 business days. If your order is late, please contact our support team and we’ll be happy to help.
The delivery window here was invented, the answer is vague, and there’s no next step.
After (this workflow):
FAQ TITLE: Where is my order?
CUSTOMER PHRASING VARIANTS: “Tracking hasn’t moved in days” / “Order still not here??” / “Can you check my shipment?”
SHORT ANSWER: Orders ship within [PROCESSING_TIME]. Use the tracking link in your confirmation email to check progress. If tracking hasn’t updated in [NEEDS INPUT: days before escalation], contact us and we’ll investigate.
ESCALATION TRIGGER: No tracking movement for the number of days defined in policy, a “delivered” scan with no package received, or a second contact about the same order.
POLICY REFERENCE: [exact line from your shipping policy]
The “after” version uses only your data, flags what’s missing, and tells the agent exactly when to stop using the macro.
Common Mistakes to Avoid
Pasting vague policies. “Returns accepted within a reasonable time” will produce a vague FAQ. Use exact numbers and conditions.
Skipping the review. Claude follows your rules well, but a human must confirm refund timelines and fees before publishing.
Generating everything in one run. Very long outputs lose quality toward the end. Split by category.
Forgetting to update. When your policy changes, rerun only the affected FAQs and update your macros the same day.
Ignoring escalation triggers. These protect you from sending a canned reply to an angry customer with a damaged package.
Pro Tips
Add one real email from your brand to the voice field as a sample. Claude mimics it far better than adjectives alone.
Ask for a version in a second language if you ship internationally, then have a native speaker check it.
Turn the “Customer Phrasing Variants” into training data for your chatbot or help-center search.
Re-run the workflow every quarter with fresh ticket themes.
Frequently Asked Questions
Can I use this workflow with a free Claude account?
Yes. The prompt works in any Claude chat. Free plans have message limits, so run one category per session.
Will Claude make up policies if my information is incomplete?
The prompt tells Claude to flag missing details as [NEEDS INPUT] instead of guessing. You should still check every answer against your real policy.
How many FAQs should I generate?
Start with 6 to 8 per category, then add more based on ticket volume.
Can I connect the output to Gorgias, Zendesk, or Shopify Inbox?
Yes. Copy the agent macros into your help desk’s saved replies and publish the short answers on your FAQ page.
Does this work for stores that sell products without sizes?
Yes. Replace the sizing category with another high-volume topic, such as product care, compatibility, or warranty.
Conclusion
A good FAQ library saves hours every week and keeps your answers consistent. This workflow works because you give Claude your real policies and clear rules. Copy the master prompt, fill in your variables, and run it one category at a time. Then check the result against your policy before you publish.
Want more prompts like this? Browse our Workflows category, or read how we test every prompt on our Methodology page.