Writing thousands of product descriptions is one of the most time-intensive tasks for webshops. AI offers a solution, but it requires a solid approach. Here is how a mid-sized e-commerce company tackled this.
For a webshop with more than 5,000 products, keeping product descriptions up to date was a permanent bottleneck. New products arrived weekly, existing descriptions were inconsistent in quality and tone, and hiring a copywriter for everything was not financially viable. The solution lay in a structured AI approach.
The webshop sold consumer electronics in the mid-market segment. The range grew continuously: new models, seasonal products, bundles. Existing product descriptions varied widely in quality: some were copied from manufacturers (technically correct but not SEO-optimised), others were manually written but inconsistent in style.
The result: low conversion on products with poor descriptions, mediocre search visibility and a content backlog that kept growing.
The solution consisted of three components:
1. Structuring product data: all product attributes were standardised in a database: brand, model, technical specifications, target audience, use cases and distinguishing features. Without structured input, AI output is also unstructured.
2. Developing description templates: a template was developed for each product category with the desired structure: opening sentence, core benefits, technical specifications in readable language, use case, CTA. The templates were written out by a copywriter.
3. AI fills in based on data and template: the model used the product attributes as input and the template as structure to generate a unique description per product. Multiple variations per product enabled A/B testing.
The initial batch of 2,000 descriptions was generated in two weeks. An editor reviewed a 10 percent sample and adjusted where necessary. The findings:
Time savings compared to fully manual writing: approximately 90 percent. Quality improvement compared to manufacturer texts: considerable.
The approach failed for products with little structured data. If a product had only a name and an EAN code, without specifications or use context, AI produced text that said nothing. The quality of the input determined the quality of the output.
Also for premium products where the description needed to convey a luxury experience, AI fell short. Those descriptions required a writer with a feel for tone and aspiration, not a template.
After the initial production, maintenance was needed. New products were added weekly via a semi-automated process: enter product data, AI generates description, editor approves. Average processing time per product: 20 minutes instead of 2 hours.
The templates were reviewed twice a year based on conversion data: which description structures led to higher add-to-cart rates?
Scalable AI content production requires three things: structured data, good templates and human oversight. Those who use AI without that foundation quickly get a lot of text of varying quality. Those who lay the foundation properly gain a structural advantage.
A webshop that uses AI well for product descriptions can save considerably on writing costs without quality loss, provided the approach is properly set up. The 90 percent saving in this case was not accidental, but the result of good preparation.
Mach8 helps e-commerce companies set up AI-assisted content production for large catalogues. View our content production service or get in touch.
We help you go from strategy to implementation. Schedule a no-obligation call.
Schedule a call