Using ChatGPT to Write Product Descriptions That Don't All Sound the Same
Bulk-generating descriptions turns your catalog into beige. The fix is a per-product angle, not a better prompt.
The trap with ChatGPT and product copy is scale. Give it a spreadsheet of 200 SKUs and a single prompt, and you get 200 variations of 'crafted with care for the modern consumer.' I did this once and had to rewrite a third of them by hand.
What works: write one description yourself for a hero product, then feed that as the voice example and ask ChatGPT to match it for the rest, but only after you paste three specific facts about each item. The model can't invent a real differentiator, so you have to supply 'made from recycled sailcloth' or 'fits a 38mm strap' yourself.
The angle step is the one people skip. Before generating, jot one reason each product exists: who it's for, what annoyance it removes. Drop that line into the prompt. Descriptions with a point read like a person wrote them; descriptions without one read like a template, because they are.
Last tip: cap the batch at twenty and scan for the word 'elevate.' It sneaks in, and when it does, the whole block sounds like the same robot. Swap it for the actual verb and move on.