If you run an online store in Lagos, Accra or Nairobi, you have probably generated product descriptions with a chatbot. So has every competitor, and the results are indistinguishable: "Elevate your everyday with this premium, versatile piece designed for the modern lifestyle." Nobody buys because of that sentence. This post is a process to humanize marketing copy so it sounds like your shop, with examples, and a note on doing it for five hundred products at once.
Why generated copy does not convert
Shoppers read product pages for two things: is this the right item, and can I trust these people? Generated copy fails both. It is vague ("premium materials", "versatile design") so it does not answer the first question, and it sounds like every other page so it does not build the trust that answers the second. Worse, it often makes claims you did not approve. "Crafted by artisans" is a lovely phrase until a customer asks which artisans.
The five-step process
1. Strip the adjectives, keep the nouns
Delete "premium", "elevate", "versatile", "stunning", "perfect for", "designed to" and "whether you're". What is left is usually the actual product: "cotton shirt, two pockets, available in four sizes." Start from there.
2. Answer the three questions a buyer has
- What exactly is it (material, size, weight, what is in the box)?
- What is it for, in one concrete scene?
- What is the catch (care instructions, delivery time, what it is not)?
Answering the third question is what builds trust. "Hand wash only" written plainly beats "premium delicate fabric" every time.
3. Write one sentence only you could write
Where it was made, who chose it, why you stock it, what a real customer said. "We tried six suppliers before this one held its colour after ten washes" is worth a paragraph of adjectives.
4. Vary the rhythm
Product copy generated by a model comes out as three even sentences. Break it. One short line for the headline fact. One longer sentence for the scene. A fragment for the catch, if that suits your brand.
5. Protect the facts, then humanize
Add your product names, prices, dimensions and brand terms to protected words so they are never rewritten, then run the description through the humanizer in Standard mode. The engine removes the remaining tell-words, varies sentence openers and swaps generic vocabulary while leaving every figure untouched.
Before and after
Generated: "Elevate your wardrobe with our premium Ankara midi dress. Crafted from high-quality fabric, this versatile piece is perfect for any occasion, whether you're heading to the office or a weekend brunch. Experience unparalleled comfort and style."
Humanized and edited: "An Ankara midi dress in a wax print we picked up in Balogun Market. Cotton, fully lined, with pockets (yes, pockets). It falls just below the knee on a five-foot-six frame. Machine wash cold, hang dry, and it will not fade. Ships from Lagos in two working days."
Same product. The second version answers what it is, where it comes from, how it fits and what the catch is. It also reads as if a person wrote it, because the rhythm is uneven and every sentence carries a fact.
Doing it at scale
Nobody hand-edits five hundred descriptions. This is where the process becomes a system.
- Build a protected-words library once. Every product name, material, size code and brand phrase. In Humanize 360 this is saved to your account and applied to every run.
- Create a style profile. A short sample of copy in your voice. The engine uses it to decide about contractions, sentence length and formality so the whole catalogue sounds like one shop.
- Batch upload. Put the descriptions in a DOCX or TXT file with one product per paragraph and upload it. Standard plans and above process several files at once; the output comes back in the same order.
- Spot-check ten percent. Read every tenth description aloud. Fix any that lost a fact (rare, because the facts are protected) or that still sound generic (common, because the engine cannot invent your supplier story).
- Add the human sentence over time. You do not need step three of the process for all five hundred on day one. Do it for the twenty best sellers, then the next twenty.
Costs for a catalogue
A typical description is 80 to 120 words. Five hundred products is roughly 50,000 words. At launch pricing that is one Project pack (₦7,500 for 40,000 words, valid for twelve months) plus a Starter pack, or one month of the Standard plan (₦4,900, 50,000 words) with batch upload included. Agencies managing several stores should look at the enterprise page for wholesale rates and team seats. Everything is on the pricing page.
What not to do
Do not use a humanizer to launder claims. If the generated draft says "dermatologist tested" and it was not, no amount of rewriting fixes that; delete it. Do not humanize reviews; write none you did not receive. And do not aim at a detector score. Aim at a customer with a question, and answer it in your own voice.
Our business page covers Formal mode for proposals, the repurposer for turning one description into a social post and an email, and invoicing in local currency for teams.
Common questions
Will the humanizer change my prices or product names? Not if they are in your protected words, which is the first step in the process above. Even without that, numbers and capitalized product names are skipped by default.
Can I humanize copy in Pidgin or a local language? The engine is built for English, including Nigerian, Ghanaian, Kenyan and South African English. Mixed-language copy is returned with the English sentences edited and the rest untouched.
What about meta descriptions and ad copy? The same process applies, but use the repurposer: paste one good description and get a short social post, a meta description and an email line from it, each within the character limits.
Does this help with search rankings? Indirectly. Search engines reward pages people stay on and buy from, and specific, trustworthy copy does both. Duplicate generated descriptions across hundreds of stores do the opposite.