TL;DR
A good collection description does two jobs: rank for one long-tail keyword and route the shopper to the right product fast. Feed the AI your real top sellers, your top analytics filters, and the cliché you want to avoid — then ask for a 60-word intro above the grid plus a 200-300-word “how to choose” and a short FAQ below it. Shopify Magic (free, built in) is fine for a first draft; ChatGPT, Claude, or Gemini give you more control over voice and structure. One change to note for June 2026: Google retired FAQ rich results in May, so the FAQ now earns clicks through AI Overviews and on-page conversion, not a rich snippet.
The task
Your Shopify store has a collection page for “Women’s Sneakers” with 47 products. Search Console says the page gets 12 organic clicks a month, mostly from your own brand name. The description above the grid is one sentence: “Our women’s sneakers collection.” Conversion on the page sits below the site average — shoppers land, scroll, get overwhelmed, and bounce. You want a collection description that does two real jobs: rank for “best women’s sneakers for standing all day” (the long-tail term that has real search volume and buying intent), and route the shopper to the right sub-collection or product within 10 seconds so they actually buy.
Which AI tool to use
You do not need a fancy setup. Three realistic options, as of June 2026:
| Tool | Cost | Best for | Watch out |
|---|---|---|---|
| Shopify Magic | Free on every Shopify plan | A fast first draft right inside the collection editor; it can read the products already in the collection | Generic voice, tends toward filler; weak at the “how to choose” routing logic |
| Sidekick (Shopify) | Free, conversational agent in admin | ”Suggest which products belong in this collection” and bulk edits | Still light on long-form copy; better as an organizer than a writer |
| ChatGPT / Claude / Gemini | Free tiers work; paid is $20/mo (ChatGPT Plus, Claude Pro, Google AI Pro $19.99) | The full structured prompt below — you control voice, length, and routing | You must paste in your real top sellers and filters; it does not see your store |
The honest recommendation: draft in Shopify Magic if you want speed, but run the structured prompt in ChatGPT, Claude, or Gemini when the page actually matters for revenue. The paid models (GPT-5.5, Claude Sonnet 4.6, Gemini 3.1 Pro) hold a longer brief and follow the “route every line” rule more reliably than the free editor widget.
Where AI helps — and where it does not
AI is genuinely good at writing an SEO-friendly intro that threads one long-tail keyword naturally, building a “how to choose” framework that routes buyers by use case, and producing tight Q&A pairs. It holds the format constraints (60-word intro, 200-300-word body, 3-question FAQ) without you having to count words.
What AI cannot do: know your top sellers, know what shoppers actually filter by, or know your specific category clichés. Every athleisure brand writes “for the woman who does it all,” and Google’s index has seen that sentence a million times. So feed the model your top three sellers, your top three filter clicks from analytics, and the line you have noticed every competitor using so it can avoid it.
A specific failure to head off: AI tends to repeat the collection name three times in the first 60 words. That reads as keyword stuffing to both buyers and Google. Tell it plainly: “use the keyword once in the first 30 words and at most one more time across the rest.”
What to feed the AI
- Collection name plus 1-2 sentences on what makes this collection distinct (not the slogan; the real reason buyers narrow to this page)
- The one long-tail keyword you want to rank for (one, not five — a page diluted across several keywords ranks for none)
- Your top three sellers in this collection, each with a one-line “who buys this”
- The top three filters shoppers click in your analytics (“by occasion,” “by season,” “by price band,” “by material”)
- 2-3 buyer types to route — each with a use case and the sub-collection or product for them
- The one cliché phrase you have noticed every competitor uses (so the model avoids it)
- Existing brand voice — formal, casual, expert, friendly — with one example sentence
- Internal links to surface (sub-collections, size guide, returns policy) — paste the URLs so the model uses them
Copy-ready prompt
Write a Shopify collection description for [collection name].
Long-tail keyword to rank for: [one keyword]
What makes this collection distinct: [1-2 sentences]
Top 3 sellers + who buys each: [paste]
Top 3 filter clicks from analytics: [paste]
Buyer types to route (use case + sub-collection or product): [paste]
Cliche phrase to avoid: [paste]
Brand voice + one example sentence: [paste]
Internal links to surface (URLs): [paste]
Structure:
1) 60-word intro for above the product grid. Use the keyword once in the
first 30 words. End with the line that tells the shopper this page is
for them.
2) "How to choose" — a 200-300 word section organized by use case, not by
feature list. Map each use case to the specific sub-collection or
product that fits. Use the verb the shopper would use ("if you stand
all day," not "for prolonged-standing wearers").
3) "Best for" — 3 buyer-type rows. Each row: a one-line buyer description
plus the sub-collection or product they should look at first, linked.
4) FAQ — 3 questions a buyer actually asks before checkout (returns,
sizing, materials, durability). Each answer 30-40 words.
Rules:
- Keyword appears once in the first 30 words and at most once more after.
No stuffing.
- Replace any cliche if it slips in. Specifically: do not write
"[cliche phrase]."
- Every sentence in "How to choose" must route the buyer somewhere
specific. Cut generic phrases like "we have something for everyone."
- The intro is 60 words for above the grid; the body sits below the grid.Shorter variant — refresh an existing description
Below is my current collection description. Rewrite it to:
1) Add the long-tail keyword "[keyword]" once in the first 30 words.
2) Replace any line that does not route the buyer to a specific
sub-collection or product.
3) Add a 3-question FAQ at the end.
Keep the voice, and keep length within 10% of the original.
Current: [paste]Sample output
A useful 60-word intro: “Women’s sneakers built for standing all day — without the cushion-collapse most pairs hit by month three. We obsess over three things: a midsole that holds shape, a heel cup that does not slip when you are on your feet for 10 hours, and uppers that breathe. If you commute, work retail, or run a kitchen, this collection is for you.”
A useful “how to choose” passage: “If you stand all day, prioritize midsole density over weight — the Cushion Lace line is our top pick. If you walk a lot but sit some, the Loafer Knit is lighter and packs well. If you commute on transit and want one pair that handles rain, the Slip-On Pro stays dry through 2 inches of standing water. Switch between two pairs and you extend the life of both by about 40%.”
A useful FAQ entry: “How do they fit if I am between sizes? Cushion Lace runs true to size; Loafer Knit runs half a size large. Between sizes, size down on Knit and stay true on Cushion. We accept returns within 60 days, worn or not, including size swaps.”
How to refine the draft
- Cap keyword density: “Use the long-tail keyword once in the first 30 words and at most one more time after that. Remove any instance beyond. Re-read and confirm it does not read as keyword stuffing.”
- Force routing in every ‘how to choose’ line: “Each sentence in the ‘How to choose’ section must end with a specific sub-collection, product line, or filter. Replace ‘we have something for everyone’ with a concrete route.”
- Match the buyer’s own words: “Rewrite each use case in the voice the buyer would say out loud. ‘If you stand all day’ beats ‘for prolonged-standing footwear.’ If a sentence sounds like a brochure, it is wrong.”
- Keep the FAQ tight: “Add 3 Q&A pairs. Each question is one a buyer actually asks before checkout (sizing, returns, materials, durability). Answers 30-40 words, plain English.”
- Differentiate from category clichés: “Read the description and flag any line that could appear unchanged on a competitor’s site. Replace those lines with claims specific to our products and our buyers.”
What changed with FAQ schema in May 2026
If you wrote collection FAQs in 2024 for the star-rated rich snippet, that incentive is gone. Google stopped showing FAQ rich results across all sites in early May 2026, and is removing the search appearance, the rich-result report, and Rich Results Test support through June 2026 (Google’s developer docs still describe FAQPage as a valid Schema.org type). Two practical takeaways:
- The
FAQPagemarkup itself is harmless and still valid; you do not need to rip it out. - The reason to keep an FAQ on a collection page is now AEO and conversion, not a rich snippet. Concise Q&A pairs covering price band, sizing, materials, and durability are exactly what Google’s AI Overviews and shopping answers pull from, and they cut pre-purchase hesitation on the page itself. Write the FAQ for the buyer reading it, not for a snippet that no longer renders.
Common mistakes
- Skipping the “how to choose” section, so the page is an inventory grid instead of a navigation experience; shoppers bounce when they have to route themselves.
- Using the same description style across every collection. Google rewards distinct, useful pages; each collection deserves its own voice and decision framework.
- Keyword-stuffing the intro. Repeating the keyword four or more times in 60 words trips quality signals and reads as spam to a human.
- Writing a thin one-liner. A 200-300 word body above (or just below) the grid is the commonly cited sweet spot; one sentence gives Google and the AI engines nothing to rank or cite.
- Putting the long description above the product grid. Buyers want products first: a 60-word intro above, the long form below. Dawn, Sense, and Refresh all support this split natively.
- Listing features instead of routing. “Made with X material” matters less than “the Cushion Lace is the pair if you stand all day.”
- Writing for the brand instead of the buyer. “For the woman who does it all” routes nobody anywhere; use the buyer’s actual use case.
- Letting AI invent product names or claims. Feed it your real top sellers and specs; fabricated specifics get returned and trust drops.
FAQ
- How long should a collection description be? Aim for a 60-word intro above the grid plus a 200-300-word body below it. Under ~150 words gives the page little to rank or get cited for; past 500, shoppers skim and only the FAQ gets read.
- Above or below the product grid? Short 60-word intro above (with the keyword and the routing line), long form below. Most Shopify themes — Dawn, Sense, Refresh — support this split natively.
- Do I still need FAQ schema after May 2026? It no longer earns a rich snippet, but
FAQPagemarkup stays valid and the FAQ content still feeds AI Overviews and reduces pre-checkout questions. Keep it; write it for the reader. - Should I use Shopify Magic or ChatGPT/Claude/Gemini? Magic is free and fast for a first draft inside the editor. For pages that drive revenue, run the structured prompt in a paid model — you get tighter voice control and more reliable per-line routing.
- What about translated or multilingual collections? Write each language separately, do not machine-translate. The French keyword is rarely the literal translation of the English one; do keyword research per market.
- How often should I refresh collection descriptions? When top sellers shift (often quarterly) or when the page ranks for a different keyword than you targeted. Both signals show up in Search Console.
Related
- AI Etsy listing
- AI marketplace listing
- Write Abandoned Cart Emails With AI
- Product Description Prompts: Bullets That Convert
- AI Product Title
Tags: #AI writing #E-commerce #Workflow #Shopify #Collection