You can get usable Shopify product descriptions from ChatGPT in a few minutes, but only if you hand it the facts and tell it what it is not allowed to invent. Ask for "a great description of my leather wallet" and you get a paragraph about "premium craftsmanship" and a stitching claim you never made. This post gives you prompts for product descriptions, meta titles and descriptions, collection pages and variant names, plus a brand-voice setup and the failure checks I'd run before anything goes live.
Status of the prompts below: they are untested templates. I did not run them on a live model for this post, so the example outputs are illustrative, written by hand to show the format. Run each on two or three of your own products before you trust it.
Set up your brand voice once, not per prompt
Most weak descriptions come from re-explaining your brand every time. Write one voice block and reuse it. If you use ChatGPT Projects or custom instructions, put it there (see the ChatGPT Projects and custom instructions setup). Otherwise paste it at the top of each chat.
BRAND VOICE (reuse for every product)
Store: [STORE NAME], sells [WHAT YOU SELL] to [WHO, e.g. first-time buyers in India aged 25-40].
Tone: [3 adjectives, e.g. warm, plain-spoken, a little dry].
Always: short sentences, say what the product does before how it looks, use rupees as "Rs." or the symbol, metric units.
Never: "premium", "high quality", "perfect for everyone", "elevate", exclamation marks, claims about health, safety or results.
Reading level: a 14 year old could follow it.
Example of our voice (3 sentences from a real product page you wrote yourself):
[PASTE]
The pasted example does more than the adjectives. Models imitate samples better than they follow tone labels. If you don't have a sample, write three sentences yourself and use those.
The fact sheet: the one input that prevents invented specs
Build a fact sheet per product in a spreadsheet or plain text. Only put in what you can prove from your supplier sheet, the product in your hand or a lab report.
PRODUCT FACT SHEET
Name: [NAME]
Category: [CATEGORY]
Materials: [EXACT MATERIALS or NOT PROVIDED]
Dimensions: [L x W x H, unit]
Weight: [VALUE or NOT PROVIDED]
Colours/variants: [LIST]
Included in box: [LIST]
Care: [INSTRUCTIONS]
Made in: [PLACE]
Certifications: [ONLY if you hold the certificate, else NONE]
Warranty/returns: [YOUR ACTUAL POLICY]
Three real customer questions we get: [LIST]
Who it is for / not for: [TEXT]
The "not for" line is the cheapest honesty upgrade on the list. A description that says who should skip the product converts the right people and cuts returns.
Prompt: product description from the fact sheet
You write Shopify product descriptions.
Use ONLY the facts in the FACT SHEET. If a detail you would normally mention is missing, write [NOT PROVIDED: detail] instead of guessing. Do not invent numbers, materials, certifications, awards or testimonials.
[PASTE BRAND VOICE]
[PASTE FACT SHEET]
Write:
1. A 1-sentence opening that says what it is and who it's for (max 25 words).
2. 3 to 5 bullet points, each starting with the benefit and ending with the fact that supports it.
3. A 60 to 90 word paragraph on how it's used day to day.
4. A "Good to know" line: size/fit/care limits and who should skip it.
Primary keyword to use naturally once in the first 100 words: [KEYWORD]. Do not repeat it more than 3 times in total.
Return Markdown that I can paste into Shopify's description editor.
Illustrative output shape (hand-written, not captured from a model):
A 750 ml insulated steel bottle for people who carry water to work.
- Keeps drinks cold through a commute: double-wall steel, [NOT PROVIDED: tested hours]
- Fits most car cup holders: 7 cm base
...
Good to know: Not dishwasher safe. Skip it if you want a bottle under 400 g.
Notice that the model flagged the missing "tested hours" instead of making up "24 hours cold". That flag is your to-do list. Fill the gap with a real number or delete the line.
Prompt: meta title and meta description
Shopify's help centre says a page title can be up to 70 characters but suggests keeping it to 60 so it isn't cut short, and suggests 160 characters for the description (page checked 2026-10-09). You edit both under "Search engine listing preview" on a product, collection or page. Ask for several options and count them, because models are poor at counting characters.
Write 5 options each for the SEO page title and meta description of this product.
Rules: titles at most 60 characters, descriptions at most 155 characters, include the keyword "[KEYWORD]" in every title, put the brand name last in the title, no all caps, no claims not in the fact sheet.
Output a table: option, text, character count.
[PASTE FACT SHEET]
Then verify the counts yourself. Paste the picks into a spreadsheet with =LEN(A2) and fix anything over. Google may rewrite titles and snippets anyway, so don't agonise over exact phrasing.
Prompt: collection page copy
Collection pages often rank better than individual products because they match broader searches like "cotton kurtas for men". Most stores leave them empty.
Write copy for the Shopify collection "[COLLECTION NAME]".
Inputs: the 8 to 12 product names and one-line facts below, our price range [Rs. X to Rs. Y], and shipping/returns policy [POLICY].
Produce:
- A 40 to 60 word intro that sits above the product grid (what is in this collection and how to choose).
- A 3-question buying guide (size, material, use) answered from the product list only.
- 5 internal link suggestions to other collections, as anchor text only (I will add the URLs).
Do not mention a product that is not in the list. Do not claim "best" or "cheapest".
PRODUCTS:
[PASTE LIST]
Keep the intro short. Shoppers scroll to products. Put a longer buying guide below the grid.
Prompt: variant names and option values
Variants are where catalogues quietly rot: "Blk", "Black", "black - XL", "XL/Black". Clean them in bulk.
Normalise these Shopify variant option values.
Rules: Option 1 = Colour (Title Case, full word), Option 2 = Size (S, M, L, XL, XXL or numeric as given), no slashes, no abbreviations, keep original SKU in a separate column.
Return a table: original title, Colour, Size, SKU. If a value is ambiguous (e.g. "Nvy" could be Navy), put it in a column called CHECK instead of choosing.
DATA:
[PASTE CSV ROWS]
For a store with a few hundred products, export a CSV, run batches of 20, spot-check each batch and re-import. Back up the original export first, always.
What about Shopify's own AI tools?
Shopify's help centre describes Sidekick as an AI assistant in the admin that can generate content including product descriptions, and says it presents changes for your review before applying them (pages checked 2026-10-09). The help pages I could read did not cover plan requirements, supported languages or limits, so confirm those in your own admin before planning around it. The benefit is that it sits next to your product data. The downside is that you still need the fact-sheet discipline, because the same invention risk applies. If you're still deciding on a platform, you can look at Shopify here. Some links on this page are affiliate links; they cost you nothing and I only link tools I'd mention anyway. The prompts in this post work in any chat model, so you don't need Shopify's tools to use them.
Failure cases to check before you publish
Run this checklist on every batch. Each item is something a model does often enough to plan for.
| Failure | What it looks like | Fix |
|---|---|---|
| Invented specs | "Weighs only 250 g", "100% organic cotton" with no source | Compare every number and material to the fact sheet |
| Fake certifications | "ISO certified", "dermatologist tested" | Add "Certifications: NONE" and forbid certification claims |
| Medical or result claims | "Reduces hair fall", "cures acne" | Ban in the voice block; Indian ad rules can apply, so check with a professional if you sell health or cosmetic items |
| Same phrasing across products | Ten pages starting "Elevate your..." | Ask for a different opening structure per product, or vary the fact sheet's "who it's for" |
| Keyword stuffing | The keyword four times in two lines | Cap uses in the prompt and read it aloud |
| Wrong units or currency | Inches for an Indian audience, dollars | State units and currency in the voice block |
For a deeper look at why models fill gaps, read the hallucinations lesson. If your wider content plan needs SEO structure, see AI SEO content strategy prompts, and for what not to paste into a chatbot (supplier pricing, customer data) see the ChatGPT privacy guide.
A workflow for 50 products in an afternoon
- Export your products as a CSV from Shopify admin and keep an untouched copy.
- Add the fact-sheet columns. Fill what you can prove. Leave the rest as NOT PROVIDED.
- Run the description prompt on 5 products. Read them out loud. Fix the voice block, not the output.
- Run batches of 10 to 20. Search each batch for
NOT PROVIDEDand fill the gaps. - Run the meta prompt and check lengths with
LEN. - Import to a few products first, preview them on mobile, then do the rest.
I haven't measured any ranking or conversion change from this process, and I won't quote one. Judge it on whether your return reasons mention "not as described" less often over the next month. That number lives in your own store data. For more like this, browse the prompt library.
