Using OpenAI GPT-3 to write E-Commerce Product Descriptions

3AI January 13, 2023

Featured Article:

Author: Raj Bhatt, CEO, Knowledge Foundry

The Problem
Writing e-commerce product descriptions has traditionally been a manual process. This activity necessitates a substantial amount of creative effort given the hundreds of new products produced by each brand/marketplace each year.

While large companies use copywriters to create these product descriptions, many smaller companies opt for simpler solutions:

1. Instead of creating product descriptions, list the features of the product in bullet points.

2. Create a templated description with placeholders for the features of the product. 3. Utilize a brand-level product description without mentioning any characteristics.

Figure 1: Example of product feature list used instead of description

Figure 2: Example of templated product descriptions

Figure 3: Example of brand-level product description

These three methods all result in product descriptions that might not be compelling to consumers. According to our polls, people like to read product descriptions that creatively highlight the advantages that they may obtain from the product by utilising its characteristics.

The Solution

Original, legible, and cost-effective E-Commerce product descriptions can be created with OpenAI GPT-3. Using the GPT-3 DaVinci algorithm, we trained models on hundreds of products across a range of E-Commerce categories. Tonality has been adjusted in the models (professional, witty, etc.,). Additionally, we employ algorithms to evaluate the originality and readability of the created descriptions.

The created product description can be compared to Google APIs to acquire insight into the internet traffic for the keywords mentioned as well as alternative phrases that may generate significant search traffic. This improves SEO for product pages.