The short version
- Supplier copy is duplicate content shared with every retailer carrying the product, so search engines pick one page and suppress the rest — usually not yours.
- Lead with a 40-word direct answer: what it is, who it suits, the deciding feature. That is the span a snippet lifts.
- Specs are the most-searched and thinnest part of most product pages. Complete them, in the units buyers use, with no rounding into vagueness.
- Your support inbox is the keyword list. Every pre-purchase question belongs on the page as a real heading.
- At scale, sort by revenue: full treatment for the products that earn, generated-from-attributes for the tail, never invented specifications.
A product page has to do two jobs that pull in different directions. It has to answer the question a stranger typed into a search engine, and it has to close a sale for someone already on the page. Descriptions that do only the second read well and never get found. Descriptions written only for the first read like keyword soup and sell nothing. The structure below does both, and it is the same structure whether you sell forty products or four thousand.
Why the manufacturer paragraph never ranks
If you pasted the supplier's copy, so did every other retailer carrying that product. Search engines pick one page to represent duplicated text and suppress the rest, and the one they pick is usually the brand's own site or the largest marketplace. Your page is not being ignored on merit — it is being deduplicated.
This is the single highest-leverage fix in ecommerce SEO, and it is unglamorous: original text on the product page, written once, for the products that actually earn money. If you are on Shopify, it sits alongside a short list of platform-specific fixes worth doing in the same week.
What to write, in order
The title: what it is, then what makes it specific
Product titles should read as a person would search: product type first, then the attributes that distinguish this variant — material, size, capacity, compatibility, colour. "Leather laptop sleeve 14-inch, full-grain, dark brown" is searchable. "The Marlowe" is not, however good the branding is. Keep the brand name if it carries weight; put it after the descriptive part, not before.
The first paragraph: the answer, not the atmosphere
The first 40 words decide whether the page is used as an answer by a snippet or an AI overview. Say what the product is, who it is for, and the one thing that makes it the right choice. Save the mood-setting for later — or delete it. A first line that begins "In a world where…" is dead weight in both jobs.
Specifications: complete, structured, and honest about what is missing
Specs are the most-searched part of a product page and the part most often left thin. Every dimension, weight, material, compatibility note and included item belongs in a list or table, in the units your customers use. Two rules: never leave a field blank rather than saying "not specified", and never round a measurement into vagueness — "about 30cm" loses the query "29.5 cm laptop sleeve".
The benefit block: translate each spec into a consequence
A spec is a fact; a benefit is what the fact means for the buyer. "Full-grain leather" is a fact. "Scuffs darken instead of showing white, so it looks better after a year than it did new" is the reason to buy. Do this for the three or four specs that matter and skip it for the rest — a page where every line is a benefit reads like a pitch and stops being believed.
Objections: answer them on the page
Your support inbox is a keyword research tool — and, as keyword research for small sites argues, a better one than most exports. Whatever people ask before buying — will it fit, how long is delivery, can it be returned, is it waterproof, does it work with X — should be answered on the page, as real questions with real headings. This is also where most of your long-tail traffic comes from: nobody searches your product name, but plenty of people search the question.
The specifics that build trust
- Where it is made and by whom, if that is a reason to choose it.
- What is in the box, itemised.
- Care, warranty and returns in plain numbers — "30 days", not "hassle-free returns".
- Who it is not for. The fastest way to be believed about fit is to say where it does not fit.
The technical layer
- One canonical URL per product. Colour and size variants should not each be a separate indexable page unless people search for them separately.
- Product structured data with price, currency, availability and, if you have genuine reviews, aggregate rating. This is what earns the price and stock line in the result.
- Unique meta title and description per product — templated is fine, identical is not.
- Descriptive alt text on every image that says what the image shows, not the product name repeated eight times.
- Images compressed and sized correctly. Product pages are where mobile performance is most often lost, and speed is a ranking factor on a page that already loads twelve photos.
- Internal links up to the category and across to genuine alternatives. A product with no inbound internal links is the last page crawled and the first forgotten.
Doing this at four thousand products
Nobody hand-writes four thousand descriptions, and nobody needs to. Sort the catalogue by revenue and by impressions, and treat the two lists as one queue:
- Products that already earn money: full treatment, written or reviewed by a human. Usually 5–10% of the catalogue and 60–80% of the revenue.
- Products with impressions but no clicks: rewrite the title and first paragraph only. They are being seen and rejected — that is a snippet problem, not a content problem.
- The long tail: generate from your real attribute data, with the specs table populated from the feed, then spot-check a sample. Generated-from-attributes beats duplicated-from-supplier every time.
- Discontinued or out of stock permanently: redirect to the closest alternative rather than leaving a dead page or a soft 404.
The rule that keeps this honest: generation is allowed to describe attributes you have. It is never allowed to invent a specification. A confident wrong measurement is a return, a refund and a bad review, and it costs more than the page ever earned. This is the same boundary that decides what to automate and what to keep manual.
A worked skeleton
- H1: descriptive product title.
- 40-word direct answer: what it is, who it suits, the deciding feature.
- Bulleted key facts: five lines, the ones a buyer scans for.
- Full specification table.
- Two or three benefit paragraphs, each anchored to a spec.
- Questions section: the four things support gets asked, as h3 questions with direct answers.
- Care, warranty, delivery and returns in numbers.
- Links: parent category, two genuine alternatives, one complementary product.
Fifteen minutes per product for the top tier, near-zero for the tail once the attribute data is clean. The reason it works is not clever copywriting — it is that almost nobody else in your category has bothered.
Frequently asked questions
How long should a product description be?
Long enough to cover specs, the main benefits and the common pre-purchase questions — typically 150–400 words for a simple product, more for a technical one. Length is an output of covering the ground, not a target.
Can I use the manufacturer's description?
You can, but it is duplicate content shared with every other retailer, so the page is unlikely to rank. Use the supplier data for the specification table and write the surrounding copy yourself.
Do product descriptions really affect SEO?
Yes, primarily through long-tail queries: attribute and question phrases that only appear on pages with real, specific copy. They also affect conversion, which affects the behavioural signals a search engine can observe.
Should each colour or size variant have its own page?
Only if people search for the variant specifically — "black" or "14-inch" as part of the query. Otherwise keep one canonical product page with variant selectors, so ranking signals are not split across near-identical URLs.
What structured data should a product page have?
Product schema with name, image, description, brand, SKU, and an Offer carrying price, currency and availability. Add aggregateRating only if the reviews are genuine and visible on the page.