This is not a cosmetic edit. It is a change of format.
Queries became conversational — the copy did not keep up
In 2018 a shopper typed “women’s winter jacket” into Google. Today the same shopper types into ChatGPT: “which jacket should I get for a mild Ukrainian winter that does not puff up like a duvet and does not look office-y”. The query grew tenfold and became substantive.
Then query fan-out kicks in — the model breaks that query into 5–15 sub-queries and assembles an answer from dozens of sources. A description that only answers “buy this jacket” matches none of those sub-queries. A description that covers “when it works”, “when it does not”, “how to wash it” and “what it goes with” matches every other one.
AI quotes what sounds like a spoken line
Open a product description and read it aloud. If a sentence sounds like a fragment of a brochure, the AI will skip it. If it sounds like a salesperson answering a question, it will be quoted. AI is trained to recognise a “useful answer” by its patterns: verbs instead of adjectives, specifics instead of superlatives, acknowledged limitations instead of praise.
“Premium material of the highest quality” is almost never quoted. “320 g/m² polyester, feels like dense cotton, does not build static at −10°C, but lets moisture through at −15°C” is quoted almost every time.
The structure that works
Replace “features → benefits → buy” with “problem → solution → proof → limitation”.
Problem. The specific situation the product solves. Not “comfort”, but “the jacket does not rustle when you walk and stays quiet in an office”.
Solution. What exactly about this product closes the problem. Not “innovative material”, but “200 g synthetic down, holds warmth to −15°C in winds up to 30 km/h”.
Proof. A number, a test, a specification. Not “many people love it”, but “survived 200 wash cycles without losing shape”.
Limitation. What the product does NOT do. Not “suits everyone”, but “not for active sport below −20°C — you need a membrane for that”.
Acknowledging a limitation is the single most quoted element. Its presence is what separates “text written by a seller” from “text written by an expert”.
What to do today
Take the description of your best-selling product. Read it aloud. If even three sentences sound like advertising, rewrite it to “problem → solution → proof → limitation”. A week later, check whether ChatGPT and Perplexity are quoting you for queries in your niche.