Here are two descriptions of the same sofa (an illustration, not a real listing).
Version A
- 3-seater
- Durable fabric
- Modern design
- Easy to clean
Version B
A 2.1 meter, 3-seater sofa that fits a 4 by 4 meter living room. The fabric wipes clean with a damp cloth, so a coffee ring or a cat’s paw prints don’t stay. The cushions are firm: easy to stand up from, and you don’t sink into them. The oak legs are 15 cm high, enough room underneath for a robot vacuum.
Version A is what most stores publish. Version B takes about three extra minutes. And two separate studies, on two different questions, point toward B.
Two studies, one rule
The first is a study titled “Enhancing Personalized E-Commerce Recommendations Under the User-Agent-Platform Paradigm: An LLM-Driven Method”. It looked at what happens when a language model reads product descriptions to recommend products. The finding: descriptions written as a descriptive paragraph worked better for the AI than the same information broken into bullet points.
The second is titled “This tree in the forest is mine: The effect of concreteness on psychological ownership”. It asks a different question, about people: what does concrete wording do to a shopper? The finding: concrete descriptions raise psychological ownership, the feeling that something is already yours before you have bought it. And that feeling raises both the chance of buying and the price people will pay.
A machine and a person, and both reward text written as if the writer has the product in hand.
A note on scope, since I would rather say it than have you find it. The first study tests a recommendation method. It does not tell you how ChatGPT or any specific assistant ranks your store, and I am not claiming it does. It is a direction, not a guarantee.
Why a paragraph helps
Bullets strip out the connections. “Durable fabric” and “easy to clean” sit side by side as two claims. In a paragraph they become “the fabric wipes clean, so a coffee ring doesn’t stay”: a reason, a condition and a result in one line. That is the part a model can lift into an answer when someone asks for “a sofa that survives a cat”. It is also the part a person remembers.
This is not an argument against bullets for people. They scan faster, which is why the next section keeps them.
Why concrete wins
Compare two bikes. “A bike with a star sticker on the right side” against “an eco-friendly, fast bike”. The first one you can see. The second is a list of adjectives every bike has.
When you can see it, a small thing happens: it starts to feel like yours. Once it feels like yours, not buying it feels like giving something up. People dislike losing more than they like gaining the same thing (loss aversion, from Kahneman and Tversky). So the same description that makes the product feel owned also makes walking away cost something. That is the chain from concrete wording to a purchase.
The test for any line in your description: can the reader picture it? “Quality materials” fails. “Oak legs, 15 cm high” passes.
The free raw material: your reviews
The first study added one more thing that helped: information taken from customer reviews. Specifically three things: how the product really is, what is good about it, and who it suits.
You already have this. Open the reviews, the WhatsApp messages and the customer service replies for one product, and in 15 minutes pull out:
- How it really is. The thing customers mention that no spec sheet does. “Smaller than it looks in the photo.”
- What is good about it. The sentence they repeat. “Easy to clean.”
- Who it suits. “Perfect for a small apartment.” “Good for a family with a dog.”
Write those into the paragraph in the customer’s own words. They are concrete already, because a customer wrote them after using the product.
So where do bullets still belong?
Keep a short spec list for facts a person checks: dimensions, material, weight, care. Put the paragraph above it. One tells the story, the other settles the questions. Do not replace the paragraph with the list.
For Israeli stores this matters twice. A lot of Hebrew catalogs are a supplier’s English bullets run through a translator, so they are neither a paragraph nor concrete. “מעוצב, איכותי, עמיד” is true of every product and visible in none.
How to know it worked
Pick your 10 best-selling products and rewrite those first, not the whole catalog. Change one thing at a time, and watch two numbers over the same number of weeks before and after: add-to-cart rate on those product pages, and the number of “does it fit / is it big enough” messages you get. Concrete descriptions answer those questions before they are asked.
And a quick sanity check on the AI side: ask ChatGPT in Hebrew who your product would suit. If the answer is generic, the page is generic. This is a check, not a result from the study.
In short
- Write the description as a paragraph, and keep a short spec list beneath it.
- Make every line something the reader can picture. Replace adjectives with things.
- Mine your reviews for how it really is, what is good and who it suits.
- Rewrite your 10 best sellers first and measure before you touch the rest.
Open one product page right now and read the description. Can you picture the product from it?
That’s it for today. See you in the next tip from Shopping Dev Solutions.
