Writing Product Descriptions: Which Method Fits Your Catalog?
A few weeks ago, a shop owner reached out to me. He sold industrial supplies – 800 products, most of them still running on manufacturer copy. He’d spent weeks researching methods: content agencies, freelancer platforms, AI tools, automation solutions.
“I’ve spent more time researching than actually using any solution. And I haven’t touched a single product.”
I’ve seen this before. The question “How do I write good product descriptions?” is actually three questions at once: Who should write them? What can I spend? And most importantly: Does it scale?
In my article on AI product descriptions, I described what I learned from 60,000+ generated texts. But before you decide how to use AI, there’s a more fundamental question: Is AI even the right approach for you?
This article compares all common methods – honestly, with real costs and timelines. No sales brochure, no “Top 10 Tools” listicle.
The Scale Problem: Why the Method Depends on Catalog Size
Most comparisons of product description methods ignore the decisive factor: the number of products.
Quick math: if you need 10 minutes per description – and that’s optimistic – 500 products is roughly 83 working days. Full time. No breaks. Just copy.
That’s why I split this into three tiers:
- Under 50 products: Almost anything works. You can write each text individually and have something good in a few days.
- 50 to 200 products: The gray zone. Manual feels tedious, but automation feels like overkill.
- 500+ products: This is where it gets real. Methods that work brilliantly for 20 products fall apart at 500.
Most shop owners who come to me are in the third tier – and they’ve been using methods from the first. That’s why they’re stuck.
Method 1: Manual Writing and Freelancers
Writing it yourself
The most obvious path: you know your products best, so you write the copy yourself.
For small catalogs with premium products, this genuinely is the best method. If you sell 30 handcrafted pieces of furniture, you want to write each description yourself – there’s passion in it, and the customer can tell.
With 500+ products? Forget it. Not because you can’t, but because you’re running a shop, not a content factory.
Freelancers
Freelancers cost between €5 and €20 per description depending on quality. Good writers charge €15–20, content-mill level is more like €5–8.
The issue isn’t the per-unit price – it’s the management overhead. Writing briefs, revision rounds, quality control, ensuring consistent tone. With 500 products at €10 per description, that’s €5,000 – plus weeks for briefing, revision, and quality control. It quickly becomes a part-time job. And you still end up with different freelancers writing in different voices.
Agencies
Content agencies deliver professional quality with consistent tone. You pay €20–50 per description for that. With 500 products, we’re talking €10,000–25,000.
The honest verdict: Manual, freelancers, and agencies deliver excellent results for small catalogs. Beyond 200+ products, it becomes a cost-time problem that’s increasingly hard to solve.
Method 2: Templates and Generic AI
Template-based systems
Some shop owners work with Excel templates: “[Product name] is a [category] made of [material]. Ideal for [target audience].” CSV import pushes it straight into the shop.
The result: cookie-cutter text that Google recognizes for what it is – mechanically assembled. Duplicate content in a new disguise. As a stopgap it might do, but long-term it helps neither with rankings nor with sales.
Using ChatGPT and Claude directly
This is where it gets interesting – and I want to be fair, because many shop owners use exactly this approach, and it’s not bad.
What works well: The cost is minimal (€20/month for a subscription), you’re flexible with phrasing, and for the first 10–20 products you learn an incredible amount about what makes a good prompt.
Where it breaks down: No Shopware integration. No batch processing. Every product one at a time via copy-paste. At 2–3 minutes per product (craft the prompt, check the result, copy it into the backend), 500 products is 16–25 hours of work. That’s half a work week – just for copy that still needs SEO optimization afterward.
The biggest drawback: the more products you create this way, the more inconsistent the tone becomes. You tweak your prompts, you’re tired some days, motivated on others – and after 200 products you notice that text 1 and text 200 sound like they’re from different shops.
My advice: Start with ChatGPT or Claude. Seriously. Learn what makes a good prompt, what your products need. But plan from the start that beyond a certain volume, you’ll need a different solution.
Method 3: Specialized AI Platforms
This is about tools built specifically for e-commerce product descriptions – not general-purpose AI assistants.
E-commerce AI platforms (as a category, there are various providers) offer batch processing, pre-built templates for product copy, and CSV export. You export your product data, run it through the platform, and import the results back.
Shopware-native solutions go a step further by integrating directly into the shop – no export/import cycle. I need to be transparent here: mitKai is one such solution, and I built it myself. So take the following table knowing I’m not entirely neutral – but the numbers are realistic.
| Generic AI (ChatGPT) | E-Commerce Platforms | Shopware-Native | |
|---|---|---|---|
| Shopware Integration | Manual copy-paste | CSV export/import | Direct |
| Batch Processing | No | Yes | Yes |
| Time for 500 products | 16–25 hours | 2–3 hours | Under 1 hour |
| Cost for 500 products | Time = €500–1,000 | €200–400/month | Credit-based |
| German language optimization | Manual per prompt | Limited | Native |
The difference shows up most in day-to-day work: if you regularly receive new products (and most shops do), the time savings compound massively.
The Hybrid Approach: The 80/15/5 Strategy
Most successful shops I know don’t use one method for everything. They combine:
- 80% of products: Fully automated via AI. These are standard products where a solid, SEO-optimized description is enough.
- 15% of products: AI-generated with light manual editing. Top sellers, seasonal highlights, products with high return rates (where better descriptions directly save money).
- 5% of products: Fully manual or from a professional copywriter. Hero products, own-brand items, products with an emotional purchase decision.
In practice: an outdoor shop with 800 products has 640 standard items described by AI, lightly edits the 120 best-selling products, and invests in professional copy for its 40 flagship items.
This sounds like common sense – and it is. But most shop owners try to treat all 500 products the same. Either all manual (impossible) or all AI (wastes potential on the most important products).
The Decision: Three Questions That Clarify Everything
Instead of a complicated matrix – answer these three questions:
Question 1: How many products do you need?
- Under 50 → A good freelancer or ChatGPT is perfectly sufficient.
- 50 to 200 → An e-commerce AI platform will save you weeks. Or a motivated freelancer with a solid brief.
- 500+ → You need automation. Everything else consumes your budget or your time.
Question 2: What’s your budget?
- Small (under €1,000) → ChatGPT/Claude with good prompts. Invest the time in learning.
- Medium (€1,000–5,000) → E-commerce AI platform or a combination of AI and freelancer for top products.
- Large (€5,000+) → You have options. Agency for premium products, automation for the rest.
Question 3: What platform are you on?
- Shopify/WooCommerce → Most e-commerce AI platforms support these shops well.
- Shopware → A Shopware-native solution like mitKai is particularly worthwhile here, because the export/import cycle goes away.
- Custom system → You need an API-based solution or a custom approach.
The Most Common Mistakes
Regardless of which method you choose – these traps are everywhere:
Copying manufacturer content. Pasting manufacturer copy into your shop as-is is the most common mistake. Google recognizes identical text and rewards nobody for it. Every description needs a unique angle – benefits over features, your target audience rather than everyone.
Keyword stuffing. “Buy sneakers sneakers cheap sneakers online shop sneakers” – that didn’t work in 2015 and it definitely doesn’t work today. Write for humans, not crawlers. If the text sounds natural, the SEO is usually fine on its own.
Ignoring product specifics. A template that sounds the same for every product is barely better than no text at all. A hiking boot description should sound different from a tent description – even if both are in the same shop. Good AI copy needs product data as input, not just the product name.
No conversion optimization. Product descriptions should sell, not just inform. Listing features isn’t enough – you need to explain why this feature matters to them. “Waterproof up to 5,000mm water column” means nothing to most people. “Keeps you dry even when it rains all day” – that sells.
Conclusion: Scale Changes Everything
The right method depends on your situation. With 30 products, a good copywriter is unbeatable. With 800 products, they’re unaffordable – literally.
Remember the shop owner from the beginning? The one with 800 products and weeks of research? He went with the hybrid approach: AI automation for the bulk, manual editing for his 40 top sellers. Within two weeks he had unique descriptions for all 800 products – instead of continuing to research, he started doing.
There’s no single best method. There’s only the method that fits your catalog, your budget, and your platform.
If you want to go deeper on the AI approach, read my article on AI product descriptions – it covers in detail what separates good AI copy from bad.
If you’re running a Shopware shop and want to test the automated route: you can try mitKai on a few products for free and see for yourself whether the results work for you. But regardless of which tool you pick – starting matters more than choosing perfectly.
And if you’re unsure which path is right for your shop – reach out. I’m happy to help, even if the answer isn’t “use my tool.”