ChatGPT prompt for product descriptions
This prompt writes a structured product description from your own product facts: a benefit-led opening, short paragraphs, a spec list and a closing line. Fill in the variables, paste it into ChatGPT or any other AI model, and compare the result with the example below.
The prompt
You are an ecommerce copywriter. Write a product description for {product_name} by {brand}, sold in the {category} category.
Audience: {target_audience}
Tone of voice: {tone}
Length: about {word_count} words
Product facts (use only these, never invent specifications, certifications or claims):
{specifications}
Structure:
1. Opening sentence: the main benefit for the audience, with the product name.
2. Two or three short paragraphs. Each one turns a fact from the list into what it means for the buyer.
3. A bulleted list of the 4 to 6 most important specifications, written as "Label: value".
4. One closing line that tells the buyer who the product is right for.
Rules:
- Use the primary keyword "{keyword}" once in the first paragraph and at most once more.
- Write in the second person ("you").
- No exclamation marks, no superlatives such as "best" or "ultimate", no prices or shipping promises.
- If a fact is missing, leave it out instead of guessing.Variables to fill in
{product_name}- The product name as it appears on your product page.
{brand}- Brand or manufacturer.
{category}- Your store category, for example Kitchen > Coffee makers.
{target_audience}- Who buys it and why, in one line.
{tone}- For example: clear and practical, warm and personal, premium and calm.
{word_count}- Target length. 120 to 200 words works for most product pages.
{specifications}- The facts the AI may use: dimensions, materials, features, what is in the box.
{keyword}- The search term the page should rank for.
Example output
Product: Ardenne 1.2 L glass pour-over kettle
Input: Brand Ardenne; borosilicate glass; 1.2 litre; gooseneck spout; temperature dial 40 to 100 °C; keep-warm 30 minutes; 1500 W; audience: home coffee brewers; keyword: pour-over kettle
The Ardenne pour-over kettle gives you the slow, precise pour that filter coffee needs, at the exact temperature you choose. The gooseneck spout lets you control the flow drop by drop, so the coffee bed stays even and you get a balanced cup. Set the dial anywhere from 40 to 100 °C: lower for green tea, just off the boil for a light roast. The borosilicate glass body shows you the water level at a glance, and the keep-warm function holds your temperature for 30 minutes while you grind and weigh. - Capacity: 1.2 litres - Material: borosilicate glass - Spout: gooseneck - Temperature: 40 to 100 °C, adjustable - Keep-warm: 30 minutes - Power: 1500 W A good fit if you brew pour-over at home and want the same cup every morning.
Example only: written for an invented product to show the format and length. Check every real output against your own product data.
Tips for better results
- Paste facts from your product data, not from memory. The rule "use only these" is what stops the model from inventing specifications.
- Keep one tone per store. Write it down once and reuse the same {tone} value for every product.
- Run the prompt on 5 products from different categories before you use it on a whole catalog.
- For marketplaces, shorten {word_count} and drop the closing line; for your own store, keep the full structure.
Frequently asked questions
Does this prompt work in ChatGPT, Claude and Gemini?
Yes. It is plain text with placeholders, so it works in any chat model. Replace every placeholder in curly braces with your own product data before you send it.
How do I stop the AI from inventing product details?
Give it only the facts it may use, tell it to leave out anything missing, and check the first outputs against your spec sheet. The prompt already contains both rules.
Can I run this prompt on hundreds of products at once?
In a chat window you paste one product at a time. In WISEPIM you save the prompt in your prompt library and run it on a selection of products, with the variables filled from each product's own data.