AI listing generation means that an AI model writes a product listing from a photo of the item: its title, its description and a suggested price. The seller only takes the photos, checks the draft and posts it. It is used most on resale and classifieds apps, where the sellers are ordinary people, not businesses.
For someone selling a used item, the listing was always the hardest part. You had to name the item correctly, describe its condition, measure it and guess a fair price. Many people gave up before they started. A model now does most of this work.
Here is how it works. First, a vision model looks at the photos. It uses computer vision, software that recognises objects in photos, to find out what the item is, its likely brand, its material and its colour. It also notices defects, such as a scratch or a stain. Next, a large language model (LLM) turns those facts into a title and a short description, in plain sentences. Finally, the app looks up recent sales of similar items in the same city and suggests a price.
Say a student in Mysuru is moving out of a hostel and wants to sell a study table. He takes three photos. The vision model recognises a wooden study table with one drawer, and a scratch on the top. The language model writes the title "Wooden study table with drawer, good condition" and a 60-word description that mentions the scratch. The app finds 40 similar tables sold nearby in the last three months and suggests ₹2,200. The student changes the price to ₹2,500 and posts the listing. The whole job takes about two minutes.
The seller must still check the draft. A model can name the wrong brand, miss a crack or describe a feature the item does not have. If the listing is wrong, the buyer will complain or return the item. So the app shows the draft, and the seller approves or edits each line. This way of working, in which a model drafts and a person approves, is called human in the loop.
AI listing generation is part of recommerce, and it is also used on online classifieds sites. Large sellers of new goods use a similar idea at a much bigger scale. There, models write listings from a manufacturer's data sheets, and product information management (PIM) software keeps those listings in order.