Write the product description an assistant can quote

Published 27 October 2026 · 5 min read

An AI assistant can only repeat what your product page says. Adjectives give it nothing. Here is what to write instead, with a before and after.

A buyer asks ChatGPT for waterproof hiking boots for wide feet, under €150. Before it answers, ChatGPT runs its own web searches and reads the pages that come back. Then it recommends two or three products and says why.

The “why” comes from those pages. If a product page says the boot is a wide fit, waterproof and made up to size 47, the assistant can say so. If the page says “comfortable, high quality, perfect for any trip”, there is nothing in it to repeat, and nothing that answers what the buyer asked.

Most shop descriptions read like that. They were written for someone already on the page, looking at the photos. An assistant works from the words.

Adjectives give an assistant nothing to say

OpenAI says ChatGPT's shopping results draw on product description metadata, alongside price. Google asks the same of every page it shows: content that answers the reader's question.

“Comfortable” and “high quality” answer nothing. Every shop says them about every product, so they tell nobody anything. Facts do: who the product is for, how it fits, what it's made of, what's in the box, how it differs from the model next to it, how to look after it, and when it arrives.

One boot, two descriptions

Take Granite Trail, an invented hiking-boot shop, and its wide-fit boot. This is the description most shops would have:

Before: “Comfortable, high-quality hiking boot. Perfect for any trip. Order today!”

And this is one that answers questions:

After: “The Ridge Wide is a hiking boot for broad feet. It is cut to an EE width, one step wider across the front of the foot than our standard boots, and leaves room for a thick sock. The upper is waterproof leather with a membrane lining, so wet grass and shallow streams stay outside. It suits day hikes and multi-day trips on marked paths. For scrambling or crampons, the stiffer Ridge Alpine is the better choice.”

• Sizes 39–47, with half sizes from 40.5 to 45.5
• Weight: 1.2 kg a pair in size 42
• In the box: the boots and a spare pair of laces
• Care: brush off mud, dry away from radiators, wax twice a season
• Ships the next working day; free returns within 30 days if unworn outdoors

Now a buyer asks for wide, waterproof boots in size 46. The second page answers every part of that question in its own words. The first answers none of it. An assistant recommending the boot can now say “wide EE fit, waterproof leather, sizes up to 47, free 30-day returns”, and every one of those facts came from the page.

When ten shops use the same words

Many shops paste the manufacturer's description as it arrives. That text is often accurate, and it makes a good start for the spec list. But when a dozen shops carry the same paragraph, nothing in it says why a buyer should get the product from you.

Think of what an assistant has in front of it: several pages with identical text, differing in price and little else. The difference has to go in your own sentences. If they're true, say that you stock the half sizes others don't, that you ship the next day, that you help with sizing by phone.

What every description should answer

Take your five best sellers and hold each description against this list. Every point it misses is a sentence to write.

1. 1 What it is, in one plain sentence, in the words buyers use.
2. 2 Who it's for, and who should pick something else.
3. 3 Size and fit: measurements, weight, how it compares with a standard size.
4. 4 What it's made of, and what that means in use.
5. 5 What's in the box.
6. 6 How it differs from the cheaper and dearer models in your shop.
7. 7 How to look after it, and the warranty.
8. 8 Delivery time and returns for this product.

How to write it so it gets used

• Use your buyers' words. Before it answers, ChatGPT types short, plain searches of its own. If buyers say “wide feet” and your page says “volume-optimised last”, the two never meet. The questions customers email you, the messages you get on Allegro and your reviews all show the words they use.
• A few short paragraphs, then a spec list. The paragraphs say who it's for and why. The list carries the numbers. A single line can't do either.
• Put it on the page as text. Words drawn into a photo or a banner aren't text a robot can quote. Nor is a description that only appears after a script runs: a study by Vercel found that none of the AI crawlers it measured run scripts.
• Prefer numbers to praise. “1.2 kg a pair” tells a buyer more than “lightweight”, and nobody can argue with it. Write only what is true: an assistant repeats it as a fact about your shop.
• One product a day. Start with the products buyers ask about most. Five descriptions rewritten well are worth more than fifty rewritten in a hurry.