Which tool to choose
For a handful of products, any free AI assistant writes a good description: the latest available model of ChatGPT, Claude, or Gemini is fine.
When the products are many, the choice changes. You need an assistant that can digest a long list pasted all at once and return the descriptions in a table, ready to re-import into the site: Claude holds the thread on long texts better than the others and is the handiest for bulk runs. Alternatively there are tools made specifically for catalogs (bulk generators built into Shopify or your management system), which import the product sheet and rewrite everything: handy if you manage thousands of items and want to publish with one click.
How to do it
From computer or phone the reasoning is the same, but for bulk runs the computer is almost mandatory: pasting and re-copying a table on a phone is awkward.
- Gather the product's real data: name, what it's for, material or ingredients, dimensions or sizes, what sets it apart from the cheaper similar one.
- Decide the tone once and keep it for the whole catalog: sober, warm, technical. Changing it from description to description confuses anyone browsing.
- For a single product use the first request below. For many, prepare a table (even in a spreadsheet) with one row per product and use the second.
- Always reread the dimensions and materials in the generated descriptions: the AI sometimes rounds off or invents a plausible detail. You double-check the numbers.
The operational syntax for a single product:
Write me the description of this product for my online shop, in English. Open with the concrete benefit for the buyer, then the practical details, then the technical features. Warm but not over-the-top tone. No empty superlatives like "the best ever." Length about 90 words. Product data:
[name, materials, dimensions, who it's for, what sets it apart]
The operational syntax for bulk runs:
Below is a table of products, one row each. For each row write me a description in English of about 70 words. Each description has to open with a specific, different detail of that single product, never with the same phrase. Open with the benefit, then the practical details. Return the result in a table: product-name column, description column. Here's the data:
[paste the table]
A real example
You sell insulated water bottles and you have twenty models that differ only by color and capacity. If you ask for twenty descriptions with no constraints, the AI gives you twenty times "Keep your drinks at the ideal temperature." Google reads them as duplicate content and hides most of them.
So you prepare a table: model, capacity, color, unique detail (the green one has a child-proof cap, the black one is in matte scratch-resistant steel). You run the second request. Now the green one's description opens with "The child-proof cap makes it the no-surprises backpack bottle," the black one with "The matte scratch-resistant finish stands up to a backpack full of keys." Same product, twenty different openings, no photocopies.
When it does NOT work (and how to fix it)
If the descriptions all seem the same
It's the number-one problem with bulk runs. The AI recognizes that the products resemble each other and recycles the opening phrase. Remedy: in the table add a "unique detail" column and order the AI to start from that. The different detail forces the different opening.
If it invents a feature that isn't there
The AI fills the gaps with plausible details: if you don't tell it the material, it might write "stainless steel" because it's common in that category. A description that promises something false exposes you to complaints and returns. Remedy: give it only the data you have and add "don't add features that aren't in the provided data; if a piece of data is missing, omit the sentence."
If the tone isn't yours
The descriptions come out correct but cold, or too jargon-heavy for your audience. Remedy: paste a description you wrote yourself and like, and ask "imitate this tone for all the others." A real example guides the AI better than ten adjectives.
A tip from someone who actually uses it
The description shouldn't describe the product, it should answer the silent question of someone about to buy: "does this solve my problem?" Before you launch the AI, write next to each product the sentence the customer says to themselves ("I need a bottle that doesn't leak in my backpack") and pass it as context. The descriptions stop being catalogs and start selling. And always publish in small batches: ten descriptions, check, publish. You catch a tone mistake before it multiplies across the whole catalog.
Frequently asked questions
How long should a product description be?
It depends on the product, but for most items 60-100 words are enough: the customer wants to understand fast, not read an essay. Complex or expensive products (an appliance, a course) can carry more, with subheadings that separate features and content.
Can I use the same description on Amazon or the marketplaces too?
Better not in the identical version. The marketplaces and your site are different pages, and identical descriptions compete with each other. Ask the AI for a variant rewritten for each channel starting from the same data: change the order and a few sentences, not the substance.
Do AI-written descriptions penalize the site on search engines?
No, if they're unique and say true things. Google doesn't penalize text because it's written by an AI, it penalizes duplicate text and empty text. A generated description, but with real data and a specific opening, is worth as much as a hand-written one; a photocopy description written by a human gets hidden just the same.