Bol.com lists over 34 million products. Even at scale, product description quality directly affects both search ranking and conversion rate. For a mid-size e-commerce business with 2,000–10,000 SKUs, poor product descriptions are a conversion problem that's almost invisible, nobody tracks which descriptions are losing sales, but entirely fixable.
Why most product descriptions are bad
E-commerce product descriptions are typically written in one of three ways: copied from the manufacturer, templated from a spreadsheet field, or written by whoever uploaded the product. None of these produce good descriptions. Manufacturer copy is written for spec sheets, not for buyers. Template copy is mechanically accurate but not persuasive. Upload-time copy is rushed and inconsistent.
The result is a catalogue where product pages feel like they were produced by a different company than the homepage. The homepage has personality, a clear voice, a considered customer experience. The product pages feel like a database printout.
The SEO case for fixing it
Thin product descriptions are an SEO problem. Google doesn't rank pages with minimal unique content well. When your product descriptions are 30–50 words of manufacturer copy that appears on 40 other e-commerce sites, your page is not differentiated. You're competing on price and domain authority alone.
Unique, specific, keyword-informed product descriptions improve ranking for long-tail product searches. These searches, "waterproof hiking boot with wide fit size 46", convert at 3–5× the rate of generic category searches because the buyer knows exactly what they want. Capturing that traffic requires descriptions specific enough to match those queries.
A 500-SKU refresh of product descriptions on one mid-size e-commerce site produced a 23% increase in organic product page sessions within three months. The pages that improved most were in categories where competitors had equally thin descriptions, the improvement came entirely from being less bad.
Why AI changes the economics
Writing 2,000 product descriptions at agency rates (€60–€120 per description) costs €120,000–€240,000. That's a project no mid-size e-commerce team can approve. At freelance rates it's cheaper but slower, a capable freelancer writes 15–20 descriptions a day. 2,000 descriptions takes three months.
AI production changes both numbers. With a properly configured content workflow:
- Descriptions can be generated at 200–500 per day, not 15–20
- Brand voice and tone are consistent across all descriptions, not variable by writer
- SEO requirements, keyword inclusion, structure, length targets, are baked into the workflow
- Variants for different markets or audience segments can be produced simultaneously
- Cost per description drops by 80–90% versus agency production
What the production setup looks like
The workflow starts with structured product data: category, attributes, key features, use cases, target customer. That data feeds the generation workflow, which applies brand voice and SEO requirements to produce a description in the correct format and length for the category.
Different product categories need different description structures. A technical product needs specification-first copy with clear feature explanations. A lifestyle product needs benefit-led copy that creates desire. A commodity product needs differentiating copy that justifies the choice. Those structures are built into the workflow templates, not left to the generator to decide each time.
The review process is where human judgment adds value: spot-checking descriptions for accuracy (AI sometimes confuses specifications), approving category-level tone decisions, and flagging edge cases. A team of two can review 500 AI-generated descriptions in a day when the base quality is high.
Prioritising the catalogue
Not all product descriptions are equally worth improving. Start with high-traffic, high-margin categories where thin descriptions are most costly. Then move to new product launches, where a good description at launch is worth more than a retrofit six months later. Leave the long-tail, low-traffic, low-margin SKUs for last, or automate them fully without manual review.
Pick a category of 100–200 products as your pilot. Run the AI production workflow, review a sample of 20%, publish, and measure the impact on organic sessions and conversion rate over 60 days. That's enough data to justify expanding the production programme across the full catalogue.
Further reading
Want to see what Content-Agent makes for your brand?
We load your brand rules and styling once. After that you make images, video, audio and text in minutes, always on brand.