How to write a category description: start with what is in the category

What a category description is for

A category page is where a shopper lands when they know roughly what they want: from a search, from the menu, from an ad. The few lines at the top have one job. They help the shopper decide, quickly, whether this page is worth their time. Nielsen Norman Group calls that judgement information scent: a shopper’s “imperfect estimate of the value that the source will deliver”.¹ Good copy makes the estimate easy.

In practice, a category description does three things:

• It confirms. This is the range you were looking for, named the way you would name it.
• It orients. These are the kinds on offer, these are the choices that matter, here are shortcuts to them.
• It helps choose. The one or two questions people ask before buying in this category, answered.

For search engines it does something similar. It tells Google what the page is about, and its links lead to subcategories and products. Google’s own ecommerce guidance asks for links from category pages to subcategory pages and to the products, and warns that without them Googlebot might not find all of your products.²

What it is not

It is not a block of keyword text under the products. Asked about exactly that, Google’s John Mueller told shop owners to think about the purpose of the page and the intent behind it: “Make up your mind, focus, don’t blindly stuff text.”³ Google’s spam policies name the extreme version, keyword stuffing, which includes repeating the same words or phrases so often that it sounds unnatural.⁴ And its guide to helpful content asks a question worth putting on the wall: is the content primarily made to attract visits from search engines?⁵

Length follows from that. A 2025 study by the agency Digitaloft looked at the category pages ranking first on Google UK for 300 commercial searches. The average page had 310 words of its own text, 44% had fewer than 200, and 10% had nothing beyond the heading.⁶ The study does not show that word count causes rankings, and says so. Read it as: the pages that win are mostly short and to the point, not long.

Shopper and search intent

Shoppers rarely search for a category name alone. They add what matters to them. Baymard Institute calls these feature searches, “queries that include one or more product attributes within the search phrase”, and in its 2026 benchmark 39% of sites have issues with them.⁷ Across many rounds of testing, Baymard also finds that roughly half of participants prefer search to the menu for finding products.⁷

Intent comes in layers. The category is the first. Attributes are the second: material, size, type. The third is the use: who it is for, where it goes, what problem it solves. Each layer asks more of your product data.

A good description answers the layers your range can actually answer, and points to the filter or subcategory that answers each one. More and more, the reader is not only a person. In Salsify’s 2026 survey of 2,712 shoppers in the US, Canada and the UK, 22% said they use AI tools to research products.⁸ An assistant summarising your category can only repeat facts it can find, which brings us to the real problem.

The root cause: you cannot describe what you have not structured

Sit down to write a good description for a category and you hit the same wall every time. You do not know what is in it. Not the category name: the actual products. How many are dimmable? Which materials? What heights? Is that reading lamp filed under floor lamps or table lamps? If the answers live in supplier titles and PDF spec sheets, the writer guesses. Guesses come out generic: “a wide range”, “something for every style”, “high quality”.

We learned this building category texts with retailers. Writing was never the hard part. The hard part was knowing what was in the category, and a category with no products attached could not be written for at all, because nothing said what the page was for. The same three data problems came up again and again:

1. Products in the wrong place. The category holds things that belong elsewhere, or misses things that belong in it. The description then describes a different set than the one on the page.
2. Missing attributes. The value exists somewhere, in a title or a PDF, but not as a field. Baymard puts it plainly: “Sometimes all product attributes simply aren’t available for all products in the list.” In its research, 64% of sites fail to show the same attributes consistently across similar products.⁹
3. Inconsistent values. One material spelled three ways cannot be counted. So no one can write “mostly brushed steel”, and no filter can offer it.

There is a second symptom we see often. When attributes are missing, the category tree grows to make up for it. A subcategory appears for every value a filter should have handled, each one needs its own description, and each description is thinner than the last.

This is why the description, the filters and on-site search tend to fail together: they run on the same data. The stakes are not small. In Baymard’s testing, sites with mediocre product lists saw abandonment rates of 67 to 90% when people tried to find a product, against 17 to 33% on sites with a slightly better toolset.¹⁰ Fix the data under the category and the description, the filters and the search results all get better at once.

The anatomy of a great category description

A category description is not one paragraph. It is a few small parts spread across the page, each with its own job and each built from a different piece of product data. Pick a part to see what it does.

Two rules hold the parts together. Keep the top short, so the products stay in view, and put the detail below them for the shoppers who want it. And make sure every part agrees with the grid: Baymard’s product list research recommends category-specific filters for the key attributes of each category, and the description should use the same words as those filters.¹¹

A step-by-step method

Five of the eight steps come before the first sentence. That is the point: by the time you write, the facts are already there and the text mostly writes itself.

1. 01

   Check what is on the shelf

   Open the category and look at every product in it. Pull out what does not belong, find what is filed elsewhere, and only then decide what the page is about. If the category is a mixed bag, no text will fix it.

2. 02

   Decide what matters in this category

   List the five to eight attributes a shopper chooses on here. For floor lamps that is lamp type, height, reach, material, bulb fitting and dimming. For running shoes it is cushioning, drop and surface. These are specific to the category, which is exactly why generic copy misses them.

3. 03

   Fill them on every product, from one list of values

   Each attribute gets an allowed set of values, and every product gets one. Brushed steel is spelled one way. Then measure it: for each attribute, how many products carry a value?

4. 04

   Roll the data up

   Count the values. What dominates, what is rare, what is the span of heights, sizes or prices? This shape of the range is the raw material of the description, and it is the step that is usually skipped.

5. 05

   Match it to demand

   Look at what shoppers search for in this category and market, and which of those words your range can honestly answer. Write about the overlap. Do not promise what the products cannot deliver.

6. 06

   Write a brief, then the text

   The brief is short: the search term for the heading, three to five facts from the roll-up, the links to add, and the one question that decides a purchase here. Then write two sentences for above the products and a short guide for below.

7. 07

   Link it up

   Turn the values with demand into shortcuts: subcategories and filtered views, named in the shopper’s words. The description points the way, the links take people there.

8. 08

   Keep it true

   Ranges change. Re-run the roll-up when products come and go, and check the text still holds. “Every lamp is dimmable” is wrong the day a lamp that is not arrives.

Step four is where the data turns into copy. Here it is for the example range:

Before and after: floor lamps

The same category, written twice. The first version is what you get from a category name and a keyword. The second is written from the structured range, and every sentence shows the attributes it came from. Switch between them.

Notice what changed and what did not. The second version is not longer at the top, it is denser. It does not repeat the keyword, it uses the words shoppers add to it. And none of it could have been written without the data: the types, the heights, the share of dimmable lamps and the bulb fittings all had to exist as fields on every product first.

The checklist

Run through it before you publish a category description, or before you brief someone else to write one.

How Cernel helps

Cernel does not start with the text. It fixes what the text is written from: the first five lines of the checklist, for every category at once.

• Every product sorted into a category

  Imported products are classified automatically into a category tree, Cernel’s own or a custom taxonomy of yours.¹²

• The attributes each category needs

  Attributes are set per category. One set on a parent applies to every subcategory below it, and narrower ones can be added to a single branch, so floor lamps get height and reach while the whole lighting tree shares bulb fitting.¹³

• Filled on every product, from your values

  Each attribute has an AI agent that fills it for every product in the category. Attributes with a value list only take values from that list, so brushed steel stays brushed steel. You choose which values go live automatically and which get a review step, and review steps only ask a person when something fails or needs a decision.¹⁴

• Completeness per category

  Cernel answers the question a writer needs answered first: of the attribute values this category’s products should carry, how many are filled in?¹⁵

• What shoppers search for, per category

  Search demand compares each category with what shoppers search for in a market, using monthly Google Ads figures. A search counts as answered when your product data has the attribute and value it asks for, and Cernel recommends the values and attributes to add next.¹⁶

With that in place, the description is the easy part. Whoever writes it, your team, your agency or an AI writer, writes from facts about the range instead of from a category name. And when the range changes, the facts change with it.

How categories and attributes work in Cernel →