Category page SEO copy: what really works, and what is useless

The block of copy at the bottom of category pages is one of the most debated topics in e-commerce SEO. Some consider it essential, others see it as a relic of the 2010s. Both camps are partly right: the problem is not the text, it is text written for no one.

This guide reviews what Google says, what the tests show, and how to write category copy that serves both SEO and sales.

What Google says

Google has spoken about the subject several times, through John Mueller.

In 2019, he warned against adding text with no purpose: random informational content on a category raises the question of whether the page still addresses users who want to buy. His advice, in substance: decide which intent the page answers, focus on it, and do not stack text blindly.

In 2021, he clarified the other side. Google needs a minimum of information on a page to understand its topic, but that minimum is often low. If product names are explicit, the page almost explains itself. If they are not, for example a pair of jeans named with just a first name, two or three sentences of context are enough to remove the ambiguity.

The message is consistent: a little useful text, yes; a lot of useless text, no.

What the tests show

Public tests and studies paint a nuanced but consistent picture:

• In 2020, SearchPilot removed the SEO copy placed below the fold on the transactional pages of a travel site: -3.8% organic sessions.
• In 2024, SearchPilot removed the SEO text at the bottom of a retailer's categories: mobile organic traffic increased significantly, a result confirmed on another part of the site. The hypothesis put forward: this copy lengthened the page without adding anything.
• In 2026, SearchPilot added an FAQ below the existing text on another retailer's categories, with questions aligned with user searches: +9.7% organic clicks.
• Glenn Gabe documented the case of a site that replaced long category descriptions, partly hidden, with a few useful sentences: the categories then improved markedly.
• On length studies, results diverge: 310 words on average on first-ranking categories in the United Kingdom according to Digitaloft (2025), a median of 580 words on top 5 categories according to 1Digital. Neither establishes a causal link between length and position.

The practical conclusion: length is not what makes the difference, it is what the text brings. Additional content that answers real buyer questions can improve a category; generic filler can hold it back, especially on mobile.

The four functions of good category copy

Useful category copy fulfills one or more of these functions. If it fulfills none, it is useless.

1. Define the scope

Say in one or two sentences what the page contains and who it is for. "160x200 fitted sheets for mattresses from 15 to 35 cm thick, in cotton, percale, sateen and linen." It is useful to the visitor, who knows they are in the right place, and to Google, which understands the topic precisely.

2. Help users choose

This is the most underused function. A user landing on a category often hesitates between several criteria: which material, which pocket depth, what is the difference between percale and sateen? Answering briefly on the category page, instead of sending them to a blog post, helps them buy without leaving the page.

3. Point to the right subcategories

The copy is an excellent vehicle for internal linking: it allows you to link, with descriptive anchors, to the most in-demand facets and subcategories ("extra deep fitted sheets", "linen fitted sheets"). Visitors find things faster, and those pages receive contextual links.

4. Use the vocabulary of buyers

Users do not all search with the same words: "fitted sheet" or "fitted sheets", "160x200" or "double", "deep pocket" or "extra deep". Copy written naturally covers these variants without forcing them. If you have to line them up in a list to fit them in, they have no place there.

Where to place the copy

A category page exists to show products. The text must never push them off the screen.

• At the top of the page: a short introduction, two or three sentences, defining the scope. On mobile, it can be truncated with a "Read more" button, provided the full text is present in the page HTML.
• Below the product list: the supporting block, for choice support and links to subcategories. It speaks to the visitor who has browsed the products without finding what they need, or who needs reassurance before buying.

A short FAQ at the bottom of the category, with two or three real buyer questions, is often the most natural form of this block.

How long should the copy be?

Google sets no length, and public data does not point to an ideal length. It does show what matters: every sentence must answer a real buyer question.

The in-depth format is not better on principle: it is better when the added content matches real searches, as in the 2026 SearchPilot test, and it becomes counterproductive when it is filler, as in the 2024 one. The only way to decide for your categories is to test both formats on comparable batches, with a control batch.

In unblind, Standard mode (250 to 400 words for a category) is the starting point. MAX mode (400 to 900 words) adds research and decision-support sections when the page's queries justify it; it is the format to test on the categories where buyers ask the most questions.

Example: the "160x200 fitted sheet" category

The first text repeats the query five times and teaches nothing. The second contains fewer repetitions, but far more relevant vocabulary (pocket, thickness, percale, sateen, linen) and answers the two questions that make the buyer hesitate.

What about facets?

An indexed facet, such as "160x200 linen fitted sheet", targets its own query and deserves its own copy. Short: two or three sentences saying what distinguishes this selection (the qualities of linen, its care, what it brings in summer and winter).

The classic mistake is to reuse the parent category text with one word changed. The result is hundreds of near-identical pages, exactly what Google seeks to filter out. Simple rule: if a facet does not justify copy of its own, it probably does not justify being indexed.

Producing this copy at scale without falling into spam

A catalogue of 3,000 categories and facets cannot be written by hand. Automated generation, including with AI, is the only realistic option. It is allowed, but it has a precise limit.

Google does not penalize AI as such: its guidance recalls that what counts is the quality of the result, not the way it is produced. However, its policy against scaled content abuse targets the generation of many pages with no added value, notably through synonym rewriting or automatic assembly of existing texts. Generated copy stays on the right side of the line if it relies on the real data of each page, answers its search intent and passes a genuine quality check.

unblind's quality gates

To hold that line, unblind does not settle for "generating text". Every page goes through a chain of checks:

1. Analysis of the real page. Our crawler reads the page like a browser: products displayed, brands, materials, sizes, breadcrumb, subcategories. The page type is identified (category, facet, brand, search, comparison, buying guide, product page) because each has its own rules.
2. Real queries. When Google Search Console is connected, the queries that already generate impressions on the page are taken into account, with brand queries separated. A targeting plan assigns each page its main query, and cannibalization detection flags pages competing for the same queries.
3. A brief per page. Before writing, a brief sets the main query, the query groups to cover and the purchase decisions the copy must answer.
4. An intent-driven opening. The first paragraph must answer the visitor's goal and, on a category, make the offering and the next action visible. Copy that merely describes the inventory is flagged.
5. An independent review. A second pass, separate from the writing, reviews the final text against around twenty defects to detect: redundant sections, sentences that describe the interface ("use the filters"), jargon, overlong sentences, obvious statements, forcibly inserted queries, interchangeable generic formulas, off-topic content, promotions or collections that will be outdated next season.
6. Fact checking. Any claim about the merchant's offering (brands sold, products, services, figures) must be confirmed by the page content. An unsourced claim is flagged and corrected. If more than 20% of sentences are copied from the page, the text is rewritten.
7. Length budgets, not quotas. The text stays within the range of the chosen mode, Standard or MAX, with no padding to reach a word count.
8. Your rules. Banned word lists and lengths per page type: generation follows your settings.

The result: text that differs from one page to another because it starts from different pages, not from a template with variables. It is the same approach applied to the title and the meta description of each page.

And if you are still wondering why to concentrate this effort on categories rather than the blog, our analysis of e-commerce SEO traffic does the math.

Checklist

• Every category has a two or three sentence introduction that defines its scope.
• The bottom block helps users choose and links to the most in-demand subcategories.
• The text does not repeat the main query artificially.
• Every indexed facet has short copy of its own.
• Products remain visible without scrolling, on mobile and desktop.
• Generated texts rely on the real data of each page and are reviewed on a sample basis.