Ecommerce AI Prompt Templates

AI prompt templates for ecommerce teams: write product descriptions, answer order issues, recover carts, build FAQs from reviews, and handle returns. Use the variants as-is, edit the placeholders, or download the editable Word doc.

Format
Editable Word .docx
Length
6 variants · copy-paste
Price
100% free
Setup
Copy or download

Get the editable Word doc in one click.

4.9·Free · No signup · Instant download
Connects to WebsiteWhatsAppTelegramInstagram
Support

Template preview

Write a product description

Prompt

You are a role for brand. Write a product description for product aimed at audience. Lead with the outcome they want, answer objection before they raise it, and end with a single clear reason to buy now. Use this context: context. Keep it under word limit words in a tone tone.

How to adapt it

  • Replace context with the real spec sheet and review quotes -- that is where the specifics come from.
  • Name the objection precisely, such as sizing, shipping cost, or durability, so the copy defuses it.
  • Tighten word limit until every line earns its place.

Why it works

It assigns a role, supplies real product data, names the buyer and their hesitation, and constrains length and tone -- which is what turns a spec dump into copy that converts.

Download free template

6 ready-to-use variants

1

Write a product description

When to use: Use it whenever you add a product and the copy is still placeholder text.

Prompt

You are a role for brand. Write a product description for product aimed at audience. Lead with the outcome they want, answer objection before they raise it, and end with a single clear reason to buy now. Use this context: context. Keep it under word limit words in a tone tone.

How to adapt it

  • Replace context with the real spec sheet and review quotes -- that is where the specifics come from.
  • Name the objection precisely, such as sizing, shipping cost, or durability, so the copy defuses it.
  • Tighten word limit until every line earns its place.

Why it works

It assigns a role, supplies real product data, names the buyer and their hesitation, and constrains length and tone -- which is what turns a spec dump into copy that converts.

2

Draft an order-issue reply

When to use: Use it when an order complaint lands and you want a first draft you can edit in under a minute.

Prompt

You are a role at brand. Draft a reply to a customer whose order had this problem: issue. Open by naming what went wrong and apologising for that specific thing, with no excuses and no explanation of our internal process. Then state remedy and confirm timeline. Close by telling them what they need to do next, or that they need to do nothing. Use these facts and do not invent any others: order facts. Keep it under word limit words in a tone tone.

How to adapt it

  • Put the real carrier status and dates in order facts so the AI cannot guess at a delivery date.
  • State remedy exactly as your policy allows, including any conditions, so the draft never overpromises.
  • Adjust tone up in warmth for a first-time buyer and down to brisk efficiency for a repeat one.

Why it works

It forces the apology to be specific before any remedy appears, which is what actually lowers a customer's temperature. Pinning the draft to supplied order facts stops the AI fabricating dates, and naming remedy and timeline means the reply commits to something rather than merely sounding sorry.

3

Write a cart-recovery email

When to use: Use it when building or refreshing your cart-recovery sequence.

Prompt

You are a role for brand. Write a cart-recovery email for someone who left cart contents behind. This is sequence stage. The likeliest reason they stopped is abandon reason -- address that reason directly rather than simply reminding them the cart exists. Include incentive and nothing more generous. Write one subject line under nine words and a body under word limit words in a tone tone. End with a single call to action.

How to adapt it

  • Set abandon reason from your own checkout analytics rather than guessing -- shipping cost and fit doubt need completely different copy.
  • Use sequence stage to control pressure: the first nudge should be helpful, the last one honest about the cart expiring.
  • Set incentive to none on early stages so you are not discounting people who would have bought anyway.

Why it works

Cart emails fail when they assume forgetfulness. Naming abandon reason makes the email answer an actual objection, and constraining incentive by sequence stage protects your margin from customers who learn that abandoning earns a code.

4

Turn reviews into an FAQ

When to use: Use it when a product has enough reviews and support tickets to show a pattern.

Prompt

You are a role for brand. Read this customer text about product: review data. Extract the faq count questions that come up most often, ranked by frequency, using the customers' own wording rather than marketing phrasing. Answer each one using only these verified facts: known answers. If a question cannot be answered from those facts, list it separately as needing an answer rather than guessing. Keep each answer under word limit words in a tone tone.

How to adapt it

  • Include one and two star reviews in review data -- the sharpest recurring questions live there.
  • Keep known answers strictly to things you can verify, so the FAQ never publishes a guess.
  • Raise faq count for complex products and drop it to three or four for simple ones.

Why it works

It extracts real questions rather than imagining plausible ones, ranks them by how often they actually appear, and refuses to answer beyond your verified facts. The separate list of unanswerable questions is often the most valuable output, because it shows exactly where your product page is silent.

5

Draft a return or refund reply

When to use: Use it when a return request needs a reply that follows policy without reading like policy.

Prompt

You are a role at brand. A customer has asked for this: request. Our policy is: policy text. The decision is decision. Draft a reply that gives the decision in the first two lines, explains the one policy reason behind it in plain language rather than quoting clauses, and then sets out next steps. Do not apologise for the policy and do not hedge the decision. Keep it under word limit words in a tone tone.

How to adapt it

  • Paste policy text in full every time, since a summarised policy is where invented refund promises come from.
  • Set decision yourself -- let the AI write the reply, not make the call.
  • Make next steps concrete, naming the label, the photo, or the window, so the customer is never left guessing.

Why it works

Leading with the decision respects the customer's time and stops the reply reading as a build-up to bad news. Supplying the full policy text keeps the explanation accurate, and separating decision from the drafting keeps a human accountable for the outcome while the AI only handles the wording.

6

Write a restock or launch announcement

When to use: Use it when stock lands or a product goes live and the waitlist is waiting.

Prompt

You are a role for brand. Write an announcement that product is available, sent to audience. Open with the news in the first line, since that is the only reason they opened it. Note whats changed if it matters to them. State the true stock position exactly as given here and invent no urgency beyond it: stock reality. Write one subject line under nine words and a body under word limit words in a tone tone, ending with one call to action.

How to adapt it

  • Put the real numbers in stock reality and instruct the AI to state them plainly -- honest limits outperform manufactured countdowns.
  • Use whats changed to give returning buyers a reason to look again rather than assume it is the same email as last time.
  • Keep word limit tight, because a waitlist audience needs news and a link, not persuasion.

Why it works

It leads with the news for an audience that already wants the product, so no persuasion is needed. Binding urgency to stock reality keeps the scarcity real, which matters because the people on a waitlist are your most loyal buyers and the quickest to notice when a countdown is theatre.

Do it in sem.chat

Let your AI agent put this to work

Load this template into sem.chat and your agent uses it automatically, in your brand voice, around the clock.

  • Save as reusable replies, scripts, or rules
  • Keeps every message on-brand and consistent
  • Hands the hard cases to a human

How to use this template

  1. 1

    Pick the closest variant. Choose based on the situation, not only the channel.

  2. 2

    Replace every placeholder. If you cannot fill a field, ask one clarifying question first.

  3. 3

    Save the final version into sem.chat, your CRM, or your help desk so the team stays consistent.

  4. 4

    Review results weekly. Drop variants that create confusion and improve the ones that work.

Frequently asked questions

Can I use these templates commercially?
Yes. Copy, edit, and use them in your business, client work, CRM, help desk, or sem.chat workspace.
Why are there six variants?
One generic template rarely fits every situation. Six variants give your team practical choices without a messy library.
Should I paste these into sem.chat?
Yes. Save the best variants as canned replies, knowledge base entries, routing rules, or CRM notes so your AI agent and team stay consistent.
Should AI write my product descriptions unedited?
No. Use it for the first draft and edit for accuracy. The AI can only work from the specs and reviews you give it, so anything it says beyond that context is invention. A human should verify every claim about materials, sizing, compatibility, and shipping before it goes live, since those are the claims that generate returns and complaints.
How do I stop AI drafts from inventing policy or delivery dates?
Paste the real source text into the prompt and instruct the AI to use only those facts. Full policy text, actual carrier status, and real order dates leave nothing for it to fill in. Prompts that ask it to summarise a policy from memory are where invented refund promises and imaginary delivery windows come from.
Is it safe to let AI answer customer emails directly?
Draft yes, send no. The pattern that works is AI produces the draft, a human makes the decision and approves the wording. Keeping the decision with a person and the drafting with the AI gives you the speed benefit without handing refund authority or policy interpretation to a tool that cannot be held accountable for it.
Why should I feed negative reviews into these prompts?
Because that is where the real questions are. One and two star reviews name the sizing confusion, the missing spec, and the shipping surprise that your product page never addressed. An FAQ built only from positive reviews answers questions nobody was worried about, while the objections that actually cost you sales stay invisible.

Put this template to work in sem.chat

Use this in sem.chat and let your agent handle it, in your voice, around the clock.