First, get the exposure map right, because it changes where the work goes.

AI Overviews concentrate on research queries, not checkout queries. Seer Interactive's analysis of 49,353 queries found informational searches trigger an AI Overview 36 percent of the time, commercial searches 8 percent, and transactional searches 5 percent. BrightEdge's tracking even shows AI Overview prevalence on ecommerce queries falling to around 4 percent, as Google protects Shopping units and ads on money terms.

So your collection page is rarely fighting the AI Overview head-on for "buy linen sheets". It is losing or winning one step earlier, on "best sheets for hot sleepers" and "linen vs percale", where the Overview now sits and organic CTR runs meaningfully below non-Overview queries.

That means collection page optimization is now a system, not a page. The system has three parts.

Part one: make the collection page itself extractable and specific. Give it a descriptive H1 that names the category the way people search it. Write a unique intro of 100 to 200 words that answers the category's core question directly, in the first sentence, because engines lift direct answers and skip padded openers.

Below the product grid, add a short buying-criteria section and an FAQ built from the questions your customers actually ask. This is the same direct-answer discipline that Shopify's own AI optimization guidance applies to product pages.

Add the structured data layer: CollectionPage and ItemList schema, breadcrumbs, and complete Product markup on the items inside. Remember that AI surfaces increasingly read your catalog, not your HTML, so attribute-complete product data determines whether the products on that collection appear in AI shopping answers at all. Keep filtered and parameter URLs canonicalized to the main collection so you do not split the page's authority across duplicates.

Part two: build the citation layer around the collection, because that is what AI Overviews quote. Publish the comparison, "best for", and how-to-choose posts that target the research queries upstream, and link each one to the collection with descriptive anchors.

The citation door is genuinely open here: only 17 percent of AI Overview citations come from top-10 organic pages, and getting cited earns 120 percent more clicks per impression on that query.

Organize this as a cluster so coverage compounds; that is what a topical map is for, and the posts should carry product and collection embeds so the click has somewhere to convert.

Part three: treat the collection as the landing page for a warmer visitor. People who click through an AI answer arrive pre-filtered and convert well above organic norms. Reward that intent: visible availability and shipping, review counts on cards since reviews are a signal agents and shoppers both weigh, and a fast path from grid to checkout.

Measure it by intent, not in aggregate. In Search Console, split the collection's queries into research terms and buying terms, and watch impressions against CTR on each. Rising impressions with falling CTR on research terms means the Overview absorbed the click, and the fix is winning the citation, not rewriting the meta description.

The straight version: stop optimizing the collection page as if it were the whole strategy. Make the page answer-ready and schema-complete, surround it with citable content that funnels down, and let the page do the one job AI cannot take from it, which is converting the buyer.