The collection copy playbook

A collection page with no words is a warehouse aisle. The copy’s job is the shop assistant’s: orient the buyer, explain what matters in this category, and point them to the right shelf.

Collection pages carry the highest-intent category searches and usually offer nothing but a grid — because nobody knows what the copy should say. The answer is buying guidance: what someone needs to know to choose well within this category. This playbook covers researching that guidance from real buyer behaviour, structuring it on the page so it serves shoppers and search simultaneously, and producing it across dozens of collections without lapsing into templated filler.

Stage 1: Research what choosing requires

Good collection copy answers the questions buyers weigh in this category. Those questions are discoverable: on-site searches within the category, the filters people actually use, review language comparing options, and support questions asking which one to buy. The output is a short list of the deciding factors — material, fit, use case, price logic — that the copy must address.

  • On-site search terms and filter usage within the category

  • Review language where buyers compare or justify choices

  • "Which one should I buy" support and chat questions, harvested

  • The category’s genuine deciding factors, ranked

Stage 2: Structure above and below the grid

Above the grid: two or three sentences that orient — what this category covers, who it serves, the one distinction that matters most. Enough to confirm arrival, not enough to delay browsing. Below the grid: the fuller guidance — how to choose, what the differences mean, honest trade-offs between sub-types, links to the sibling collections that might fit better. This placement serves shoppers who need it without taxing those who do not, and gives the page genuine substance to rank on.

Stage 3: Scale without templating

The structure repeats across collections; the content must not. "Our {category} collection offers something for everyone" is filler that shoppers skip and search engines discount. Each collection’s copy comes from its own deciding factors and its own review language. Where a collection is too thin to say anything true — near-duplicate filtered views, tiny seasonal sets — the honest answer is less copy or a noindex, not invented prose.

How it works

01

Extract deciding factors

Search terms, filters, reviews, and support questions per category, distilled into what choosing requires.

02

Write to the two-position structure

Brief orientation above the grid, real buying guidance below, links to adjacent collections.

03

Roll out with a quality gate

Per-collection content from per-collection evidence; thin collections get restraint, not filler.

What you get

  • Collection pages with a reason to rank beyond their product count

  • Buying guidance drawn from how buyers actually choose

  • A repeatable structure that never repeats its sentences

  • Filler prose eliminated instead of multiplied

Frequently asked questions

How long should collection copy be?

Above the grid: two or three sentences. Below: as long as genuine guidance requires — often 200–500 words for a considered category, less for a simple one. The moment it stops helping someone choose, it is too long.

Does text below the grid actually get read?

By the shoppers who are stuck choosing — the exact ones about to leave without it. And it is read by crawlers regardless, which is why it beats an empty grid for rankings. Both audiences, one placement.

Should every filtered view get copy?

No — only the filtered views with real search demand that you deliberately index. The rest should canonicalise to the parent, copy-free. Writing prose for facet noise is effort spent making duplicate content worse.

Which agents run this?

Visibility runs the research and owns which collections deserve investment; Builder ships the copy and structure. Review language and search data flow from the shared memory both already use.

Put the team to work on your store

Connect your storefront, analytics, and email stack, set a goal, and let the agents run the work end to end. Start on the free plan with your own model key.