AI Prompts for Online Stores: 10 Ready-to-Use Tasks You Can Hand to AI
Prompt Hack

AI Prompts for Online Stores: 10 Ready-to-Use Tasks You Can Hand to AI

Tanyalak Techapeompol
September 9, 2026
6 min read

An online store can hand 10 tasks to AI right away: writing product descriptions, briefing product photos, answering repetitive chat questions, reading reviews to find why customers hesitate, writing ad copy in multiple versions, translating product pages into English and Chinese, summarizing weekly sales, writing review-video scripts, analyzing competitors, and finding keywords on Shopee/Lazada. Every task has a ready-to-use prompt to copy below. Start with the 1–2 tasks that take up the most time in your store right now; you do not need to do all 10 at once.

These 10 tasks have one thing in common: they repeat every week, the volume is high, and the quality depends on how tired the person doing them is. Those are exactly the conditions where AI can take over best. The size of the opportunity is not small: ETDA reported in its 2024 survey that Thai e-commerce was worth 5.96 trillion baht in 2023, with retail and wholesale the largest segment at 2.83 trillion baht, and e-marketplaces the most popular sales channel at 24.58%. The principles behind the prompts below are the same ones Custocare uses in its training for business owners, such as Upskill AI for Real Business Growth (May 2026, with the Chiang Mai Provincial Industry Office × Decentric × NDEA, 200 participants) and an in-house AI Content Creator program for Pisona Group Phuket, 40 people per cohort.

Works with Claude or ChatGPT. Copy the prompt, paste it in, and replace each [ ] with your actual product details.

Which 10 tasks should online stores stop doing by hand?

The ten tasks to hand to AI fall into 3 groups: content production (product descriptions, image briefs, ad copy, video scripts), customer replies (standard answers for chat), and data reading (customer reviews, weekly sales, competitor moves, keywords). Each task below comes with a copy-ready prompt and instructions for what to do next.

1. Write product descriptions for 100 SKUs

Problem: Typing them one by one takes days, and quality is inconsistent.

Prompt:

Help me write a product description for an online store (Shopee/Lazada).

Product information:
- Product name: [product name]
- Features/specs: [enter specs, e.g. material, size, color, weight]
- Key selling points to emphasize: [e.g. durable, easy to use, good value]
- Target customers: [e.g. working adults, homemakers, students]

Please write it in 2 versions:
1) Short version (for the product listing page, no more than 2 lines)
2) Long version (for the product detail page, with features as bullet points)

Tone of voice: [friendly / professional / direct sales]

How to use:

  1. Prepare the product data as an Excel table (name, specs, selling points) covering several SKUs at a time
  2. Have the AI write one at a time, or paste them in batches of 5-10 per round
  3. Check for words that Shopee/Lazada prohibit (such as exaggerated advertising claims) before posting for real

2. Brief white-background product photos and lifestyle photos

Problem: You are not sure how to brief the photographer or design team to get the images you want, and you lose time on rounds of revisions.

Prompt:

Help me write a photo shoot brief for the design team/photographer for this product: [product name]

I need 2 sets:
1) White-background images (Product Shot) — specify the camera angles to shoot, the lighting to use, and what must be clearly visible (logo/key features)
2) Lifestyle images — specify the scene, the mood, and the people in the frame (if there are models) that fit the target customers: [target group]

Desired image style: [minimal / warm / luxurious / fun and bright]

How to use:

  1. Use this prompt to create the brief before booking the shoot, which cuts down revision rounds
  2. If you want images straight from AI, take it further with an image generation tool (such as Midjourney) by turning the brief above into an English prompt

3. Answer repetitive chat questions (how many days for delivery, what colors are available, is it authentic)

Problem: The same questions come in every day, and the admin team wastes time typing the same answers over and over.

Prompt:

Help me create a set of standard answers (FAQ Script) for customer chat. Store: [store name]
Products: [product type]

Information to use in the answers:
- Delivery time: [specify]
- Shipping carriers used: [specify]
- Colors available: [specify]
- Authenticity/warranty policy: [specify]
- Exchange/return policy: [specify]

The answers should be:
- Polite and friendly, not stiff
- No more than 2-3 lines per answer
- With a moderate number of emojis

How to use:

  1. Have the AI create the set of answers and save them as Quick Replies in your chat system (Shopee/LINE OA/Facebook)
  2. Update them whenever a policy changes, then ask the AI to rewrite the messages

4. Read reviews/customer chats to find why customers hesitate

Problem: Customer information is scattered, and no one has time to read through every message.

How to prepare the data:

  1. Collect past chat messages/customer reviews into a single file (Word/Excel/CSV)
  2. Upload the file into the chat by pressing the "+" button to attach it

Prompt (use after attaching the file):

From the attached file of customer chats/reviews, please analyze:

1. Why customers hesitate and do not decide to buy (group them into topics, with the number of times each appears)
2. The 5 most frequently repeated questions
3. Suggestions on where to fix the product page/store policy to reduce these concerns

Summarize in a table, ordered by importance

How to use the results:

  • Use the analysis to adjust the product detail page (task 1) or add FAQs (task 3) that target the right points

5. Write 20 versions of ad copy to test

Problem: You can only come up with a few styles of ad copy on your own, not enough to test for a winning ad.

Prompt:

Help me write 20 versions of ad copy for Facebook/TikTok Ads
Product: [product name]
Main selling point: [specify]
Promotion (if any): [specify]
Target audience: [specify]

Please divide them into a variety of approaches, such as:
- Problem-solution (Pain point)
- Review/Social proof
- Value for money/price
- Urgency (Limited time)
- Storytelling

Each version should be no more than 3-4 lines long, and end with a CTA

How to use:

  1. Group the ad copy by the approaches the AI provided, and set each up as a separate Ad Set in Ads Manager
  2. Test with a small budget first to see which approach gets the best CTR, then increase the budget

6. Translate product pages into English and Chinese to sell across borders

Problem: Hiring translators language by language is expensive and slow, and you miss the launch window for new products.

Prompt:

Help me translate the following product description into English and Chinese (Simplified)
As you do:
- Use natural language suited to selling on a marketplace (not word-for-word translation)
- Keep technical terms/brand names as they are
- Adjust units of measurement (if any) for international readers, if necessary

Original text (Thai):
[paste the product description text]

How to use:

  1. Start with your best-selling products, then expand to other products
  2. Have a native speaker (if you have a team) review it once more before real publication, especially for words with culturally sensitive meanings

7. Summarize the weekly sales report

Problem: You pull data from the Shopee/Lazada back office and then have to summarize it yourself every week.

How to prepare the data:

  1. Export the sales report from the back office as an Excel/CSV file
  2. Attach the file to the chat

Prompt (use after attaching the file):

From the attached sales file, please summarize a weekly report that includes:

1. Total sales compared with the previous week (up/down by how many %)
2. Top 5 best-selling products and Top 5 products with declining sales
3. The best-performing channel (if you sell on several platforms)
4. Observations/what to do next week

Summarize it as a short, easy-to-read report that can be sent straight to the team's LINE group

How to use the results:

  • Make it a routine every Monday before the team meeting starts, so the data is ready to discuss right away

8. Write scripts for product review videos

Problem: You cannot think of a script, the videos are boring, and there is no twist to keep viewers watching to the end.

Prompt:

Help me write a short video script (TikTok/Reels) 30-45 seconds long
for a review of this product: [product name]
Key points I want to mention: [specify]
Problem this product solves for customers: [specify]

Desired structure:
1. Hook (0-3 sec) — grab attention immediately
2. Body — describe the problem, then show how the product solves it (use spoken language, not written language)
3. Closing CTA — tell viewers what to do next (order now/chat to ask/follow)

Also specify how many seconds each segment should take

How to use:

  1. Adjust the script into your own way of speaking, or the presenter's, before the real shoot
  2. Make several versions of the Hook and test which one gets more viewers to watch to the end

9. Analyze what competitors changed in their prices and content

Problem: You cannot keep up with competitors and do not know what strategy changes they have already made.

How to prepare the data:

  1. Take screenshots of competitors' storefronts, or copy the prices/promotions you see, at regular intervals (for example, every week)
  2. Collect them in a single file in chronological order

Prompt:

From the attached data (prices/promotions/content of the competitor store [store name] over time)
please analyze:

1. What price changes were made, when, and what the likely reasons were
2. What pattern or trend the changed content/promotions show
3. Where we can adjust our strategy to compete better

Summarize in short, easy-to-read bullets

How to use the results:

  • Make it a monthly routine and keep it as a database for long-term comparison; competitor patterns will become clearer over time

10. Find the keywords people actually search on Shopee/Lazada

Problem: You name products and set tags by guesswork, without knowing what words people really search with.

Prompt:

Help me brainstorm the search terms (keywords) that customers are likely to use to find this product on Shopee/Lazada
Product: [product name]
Category: [specify]
Target customers: [specify]

Please divide them into:
1. Main keywords (short, direct) — 5 terms
2. Long-tail keywords (e.g. specifying the use or the problem solved) — 10 terms
3. Keywords people might misspell or use under different names (synonyms/slang) — 5 terms

Then recommend which terms to put in the product name, and which to put in the tags/description instead

How to use the results:

  1. Compare the results with the popular search terms in the Shopee/Lazada back office (Seller Center has this data) to confirm they match real behavior
  2. Put the main keyword at the start of the product name, because it has the biggest effect on being found

If you only have one week, which task should you start with?

Start with the 1–2 tasks that take up the most time in your store right now, not the ones that sound most cutting-edge. To choose, time yourself for one week and see what takes the most time. If it is typing repeated chat replies, start with task 3. If it is writing descriptions for new products, start with task 1, then expand to other tasks once you are comfortable.

The order matters because changing many tasks at once leaves no one on the team with time to check quality, and unchecked output costs more than the time it saves. An observation from McKinsey's State of AI, 2026 edition (published Aug 2026) points the same way: 8 in 10 people say AI genuinely makes their own work more efficient, but only 37% say their organization sees results at the profit level. The gap comes from plugging AI into existing processes without redesigning the work. For an online store, redesigning the work means turning prompts into routines with a clear day and a clear owner, such as summarizing sales every Monday morning, or updating the chat answer set every time a policy changes.

Usage summary

| Step | What to do | |---|---| | 1 | Pick the 1-2 tasks that take the most time in your store right now | | 2 | Copy the prompt that matches that task and replace the [ ] with your own product/store information | | 3 | Let the AI do the work, and always review the output before using it for real | | 4 | Repeat it as a routine (weekly/monthly) to keep saving time |

What should you check in AI output before using it?

Check 4 things before every post: exaggerated advertising claims that Shopee/Lazada prohibit, product spec numbers the AI may have added on its own, translations with culturally sensitive meanings, and a brand tone of voice that must be the same across every channel. These four checks take under five minutes per item, but they are the line between stores that use AI well and stores that have their listings banned.

The mistake that happens most often is product specs, because AI writes to read smoothly by adding details you did not give it, such as an approximate material or size. The fix that works is to put the specs into the prompt as complete, itemized points, then add the instruction: "If any information is not in what I provided, do not guess; leave it blank." Another point is translation: have a native speaker check it before publishing, as noted in task 6, especially for beauty products, supplements, and baby products, where each country has different advertising restrictions. Finally, keep the prompts that work in a central file for the store, because the real value is not in a single answer but in getting the same result every week even when the person doing the work changes.

A real example from a Custocare client

[Pending case from: an online store/brand that uses this prompt set — needs the store name or industry, the number of SKUs handled, the time before → after, and the measurement period, along with a link to the full case study page]

Frequently Asked Questions

Which AI should I use, and do I have to pay? All 10 prompts work with the free versions of Claude and ChatGPT, but the tasks that require attaching a file (tasks 4, 7, 9) work better on versions that support file uploads and have longer usage quotas. To start at no cost, begin with tasks 1, 3, 5, and 8, which need no attached files at all.

How long does it take to write descriptions for 100 SKUs? It depends on how complete your starting data is. If you have prepared an Excel table with names, specs, and selling points, sending 5-10 SKUs at a time as suggested in task 1 is faster than sending the whole store at once, because it is easier to review and the quality is more consistent. Most of the time goes into preparing the data, not waiting for the AI.

Will the AI write the same thing as other stores, so it looks copied? You can prevent this by putting information that only your store has into the prompt, such as selling points real customers praise, your store's warranty policy, and your specific customer group. If you provide only a generic product name, the output will be generic, in line with the information given. The difference comes from the input, not from the model.

Can I let the AI answer customer chats automatically? The task 3 prompt is designed to create a "set of standard answers" saved as Quick Replies, not to let the AI answer live, because answers about price, warranty, and returns are binding on customers. Having a person press send from a reviewed set is safer, and it is still fast enough to genuinely reduce admin time.

Is it safe to upload product and sales data? Before attaching sales files or customer chats, remove personally identifiable information such as full names, phone numbers, addresses, and order numbers. The analyses in tasks 4, 7, and 9 need only the text and the numbers, not customer identities, and you should also check the account's setting on whether your data is used for training.

How do I measure whether using AI is worthwhile? Look at the time per piece of work before and after, for example how many SKUs per hour you used to write descriptions for and how many you do now, and add the relevant end-result numbers, such as the CTR of each ad copy approach in task 5 or the video completion rate in task 8. Capture the numbers for one week before you start, and the comparison will be clearer than going by feel.


Author: Tanyalak Techapeompol Managing Director of Custocare, an AI and Digital Transformation consultancy, with 17 years of IT experience Former CIO of Oasis Spa Group, overseeing IT across 14 branches in 4 provinces · Speaker and guest lecturer on AI for Chiang Mai University and private-sector organizations

Profile: tanyalak.net · LinkedIn · 6 Anthropic certifications

Published: 09/09/2026 | Last updated: 09/09/2026

References

  1. ETDA — Thai E-Commerce Value Survey for 2023 (published 2024)
  2. McKinsey — The state of AI in 2026: On the road to ROI (Aug 2026)
  3. Anthropic Certifications — tanyalak.net