ECOMMERCE

Bulk product descriptions without duplication

Use when you need to write descriptions for many near-identical variants and each one must read differently.

bulk-product-descriptions-no-duplication.md

Download .md

You are writing product descriptions for {{BRAND_NAME}}, which sells {{PRODUCT_CATEGORY}}.

Here are the products, one per line, with their attributes:
{{PRODUCT_LIST_WITH_ATTRIBUTES}}

For each product write:
1. A description of 90 to 130 words.
2. A 3 to 5 bullet spec summary drawn only from the attributes given.

Anti-duplication rules:
- Every description must open with a different sentence structure. Do not reuse an opening pattern across the set.
- Across the whole set, no sentence of 8 or more words may repeat.
- Vary which attribute leads: material, use case, fit, care, or who it suits. Do not lead with the same attribute twice in a row.
- Do not use the phrase "perfect for", "elevate", "crafted with care" or "whether you are".

Fact rules:
- Use only the attributes supplied. If an attribute is missing, omit it. Do not infer materials, origin, certifications or performance claims.
- If a product looks like it needs a claim you cannot support, add a line "MISSING: <what you need>" instead of writing around it.

Output as a table with columns: SKU | Description | Bullets | MISSING.
No em dashes. No summary paragraph at the end.

Fill in before running

Replace each placeholder with your own detail. The more specific you are, the less the model invents.

{{BRAND_NAME}}
{{PRODUCT_CATEGORY}}
{{PRODUCT_LIST_WITH_ATTRIBUTES}}

Getting a better result

Feed 10 to 20 products per run. Past that the model starts recycling sentence shapes even with the rule in place.
Paste back any two descriptions that read too alike and ask for one to be rewritten from a different lead attribute.
Keep the MISSING column - it is usually faster to fill gaps in the feed than to rewrite the copy later.

Questions about this prompt

When should I batch descriptions rather than write them one at a time?

When you have a run of products whose attributes barely differ and a template would give you the same paragraph with one word swapped. Writing singly is fine for ten distinct products. Reach for this when the risk is not bad copy but identical copy, because the anti-duplication rules only bite when the model sees the whole set at once.

What should the product list contain, and how many rows per run?

A clean attribute line per product in {{PRODUCT_LIST_WITH_ATTRIBUTES}}, plus the brand and category. Keep the batch to ten or twenty rows, as the tips say, because the rule against repeating any sentence of eight or more words degrades once the set is longer than the model can hold in view.

What does the output table give me?

A table of SKU, description, bullet specs and a MISSING column. The copy is the deliverable, but MISSING is the column to read first: it lists the claims the model refused to invent from the attributes you gave it, and filling those gaps in the feed is faster than rewriting the copy later.

What goes wrong most often on a batch like this?

Reading the descriptions one at a time. Near duplication only shows when two rows sit side by side, which is what the tips mean about pasting back a pair that reads alike and asking for one to lead on a different attribute. The other costly habit is filling a MISSING gap with a plausible material instead of checking supplier data.