500 Product Descriptions in 3 Weeks with AI

The challenge

A fast-growing webshop had 500+ products but fewer than 30% had a complete product description. The marketing team needed two part-time employees just to write copy: not scalable with a planned catalogue expansion.

Our approach

AI pipeline that automatically generates SEO-optimised product descriptions based on product attributes, specifications and brand information. Includes quality control and a human validation flow for edge cases.

1. Phase 1

   Capture the brand voice

   We analysed the 150 existing descriptions the team considered good and distilled tone, sentence length and recurring phrasing. Without that baseline, AI produces correct but characterless copy.

2. Phase 2

   Get product data in order

   Output quality depends entirely on input quality. Missing specifications and inconsistent material names were cleaned up first: this was the most time-consuming part of the project.

3. Phase 3

   Generate with SEO structure

   Every description follows a fixed structure covering features, application and search terms in natural positions. No keyword stuffing: copy that works for the customer also ranks better.

4. Phase 4

   Human final check

   Every text passed an editor before publication, with extra attention for bestsellers. That review took roughly one minute per product versus half an hour of writing.

The results

• 92% time savings on content production
• Full catalogue coverage achieved in 3 weeks
• Organic search traffic +34% in 3 months
• Conversion rate improved by 21%

Frequently asked questions

Does Google penalise AI-generated product copy?

Google judges quality and usefulness, not how a text was produced. Unique, informative descriptions with human review perform well; mass-duplicated copy does not.

Do all descriptions not end up sounding the same?

That risk is real with a naive approach. By varying structure and angle per product category, and through the final review, the catalogue stays varied.

How much human review remains necessary?

About one minute per product. For top products the team deliberately invests more, because that is where the traffic and revenue are.

Can this work in multiple languages?

Yes, and it is one of the strongest applications. Do define the brand voice separately per language rather than translating literally.