How to Optimize for AI Overviews on Shopify: Two Tracks, Not One

Someone has probably told you AI Overviews are eating your traffic. But the first useful question isn’t what to fix. It’s which of your queries even have an AI Overview on them. We have four other posts in this neighbourhood, and none of them answers that. AI SEO for Shopify is about using AI to do SEO work across a catalogue. Shopify agentic commerce is about being transacted with by an AI agent. The best Shopify AI SEO apps and the best apps for AI search visibility are app-buying guides. This post is narrower than all four. It covers getting your store surfaced and cited inside a Google AI Overview, and which of two fix lists applies to which kind of query.

Because the answer really does split in two. And the data behind that split is not subtle.

Step one: find out which of your queries have an AI Overview

The bottom line: prevalence swings by roughly a factor of seven depending on search intent. So measure your own split first.

Seer Interactive tracked 53 brands across 5.47 million queries. The study ran from January 2025 to February 2026. The gap by intent was wide, and it held all year.

Shopping queries have moved fast, though. A March 2026 study pulled 20.9 million shopping keywords from Ahrefs’ database. Every one had a shopping box on the results page. Then it checked how many also carried an AI Overview.

Direction of travel differs by intent too, not only the level. Semrush watched 600,000 US desktop keywords across ten industries. Over six months, the two intents moved opposite ways.

Read those three together and the picture is clear enough. Research-phase queries are where AI Overviews are arriving. Ready-to-buy queries are, for now, mostly holding.

The two tracks, and how to tell which one a query is on

The bottom line: sort every keyword you care about into ready-to-buy or still-deciding. The fix list is genuinely different for each.

Track 1: the shopper has already decided

These queries name the thing. A product name plus a model number. A brand plus a size. “Price”, “buy”, “free shipping”, or a bare SKU. An AI Overview is usually absent here. A shopping module often sits in its place instead, and that module is fed by your product data rather than your page copy.

Track 2: the shopper is still choosing

These queries describe a job to be done. “Best X”, “X vs Y”, “X for Y”, “how to choose X”, “is X worth it”. Anything phrased as a question belongs here. An AI Overview is usually present. Semrush’s numbers say it is becoming more present, not less.

A twenty-minute diagnostic you can run today

Do this by hand once. You learn more from watching ten results pages than from any dashboard.

1. Export your top 100 queries. In Search Console, open Performance, switch to the Queries tab and export. Our Search Console setup guide covers the connection if you haven’t done it yet.
2. Tag each query Track 1 or Track 2. Anything with “best”, “vs”, “for”, “how”, “why” or a question mark goes in Track 2. Product names, model numbers and price words go in Track 1.
3. Spot-check ten from each list. Search them in a private window. Note whether an AI Overview appears. If it does, write down which domains it cites.
4. Tally the result. Now you know your real split rather than an industry average. You also know who is getting quoted instead of you.
5. Work the matching fix list for whichever track holds most of the traffic you care about.

One caveat before you start: prevalence moves month to month. BrightEdge measured AI Overviews on roughly 48% of all tracked queries in February 2026. A year earlier it was about 30%. So treat your tally as a snapshot, and re-run it a couple of times a year.

Track 1: optimize for AI Overviews on transactional queries

The bottom line: on ready-to-buy queries your product data feed does more of the work than your page copy. Google’s own guidance points at Merchant Center rather than at new markup.

Google is unusually direct about this. Its documentation on AI features says there are “no additional requirements to appear in AI Overviews or AI Mode, nor other special optimizations necessary”. The same page names keeping Merchant Center information up to date as ordinary best practice. So the work here is accuracy, not novelty.

Four things to check, in rough order of how often they’re wrong.

1. Availability that says the same thing in three places

Google’s product data specification asks you to “accurately submit the product’s availability and match the availability from your landing page, checkout pages, and structured data”. Three surfaces, one answer. Drift usually creeps in when stock runs through a separate app or a manual CSV feed. So start there.

2. Identifiers you haven’t invented

Per the same specification, “providing the correct GTIN for a product will ensure the best user experience and result in the best performance”. It also warns that “products with a GTIN but submitted without one may have limited visibility”. Some products genuinely have no GTIN, like a store brand you’re the only seller of. For those, Google’s guidance is to submit brand and MPN rather than guess at a number.

3. Attribute completeness against what shoppers type

Merchant Center’s AI performance insights let you “identify popular product specifications searched by users (for example, color, style, material)”. An attribute completeness score then flags products missing those attributes. Colour and material feel like Shopping housekeeping. On conversational surfaces they are how a product gets matched to a described need. Our Merchant Center guide has the connection steps and the field-by-field detail.

4. On-page structured data that agrees with the feed

Product schema on the page should report the same availability and price as the feed. Plug In SEO adds Product, Organization, WebSite and BreadcrumbList JSON-LD automatically, which keeps that half consistent as the catalogue changes. The feed half stays your job in Merchant Center. Our structured data guide has the full schema set if you’d rather do it manually.

Track 2: optimize for AI Overviews on research queries

The bottom line: here the AI Overview is usually already there. So the job is being quotable: indexed, snippet-eligible, and answering the question in a form that can be lifted whole.

Google sets one hard floor. To be shown as a supporting link in AI Overviews or AI Mode, “a page must be indexed and eligible to be shown in Google Search with a snippet”. Check that literally. A noindex tag, a nosnippet directive or a tight max-snippet limit will each keep a page out. Those are easy to inherit from a theme or an old app without noticing.

Clearing that floor pays measurably. Being cited and merely ranking are not the same outcome.

Then there’s the number in the hook, which is the most useful fact in this post. BrightEdge found that few cited sources are also the ones ranking. Ecommerce came in lowest of the nine verticals it breaks out.

Read that both ways, because both directions matter. You don’t have to outrank everyone to get quoted. But outranking everyone won’t get you quoted either. So the fix list looks less like classic ranking work and more like editing.

What to change on the page

• Have a page that answers the question. “Best gifts for someone obsessed with coffee” needs a page that answers it in sentences. A collection page with forty products and thirty words of copy answers nothing.
• Put the answer in the first sixty words. No wind-up, no scene-setting. Lead with the conclusion, then justify it underneath.
• Make claims specific and sourced. A number with a link behind it is quotable. An adjective isn’t.
• Write headings that read like the query. One question per heading gives a retrieval system an obvious anchor.
• Add a real FAQ block. Phrase it the way shoppers phrase things, with FAQPage structured data behind it.
• Show a last-updated date, and actually update the page when the answer changes.

Three things Google says you don’t need

The bottom line: three widely-recommended tactics are not requirements, on Google’s own published account.

1. New files or special markup. Google’s documentation says plainly that “you don’t need to create new machine readable files, AI text files, or markup to appear in these features”. It adds that there is “no special schema.org structured data that you need to add”. Product schema still earns its keep for rich results and feed consistency, as above. It just isn’t a citation switch.
2. A separate set of AI-only pages. The same guidance holds that “all existing SEO fundamentals continue to be worthwhile”, and points at “creating helpful, reliable, people-first content”. One good page beats a parallel version written for machines.
3. A whole new measurement stack. Your AI Overview impressions are already in Search Console, which is the next section.

How to check whether your AI Overview work paid off

The bottom line: impressions from AI features already land in your existing Search Console Performance report, under the Web search type.

Google’s wording is specific. Sites appearing in AI features “are included in the overall search traffic in Search Console. In particular, they’re reported on in the Performance report, within the ‘Web’ search type”. In June 2026 Google also announced dedicated generative AI performance reports. Those are rolling out to a subset of sites first, so a separate view is coming even if you can’t see it yet.

Until then, write down three things before you change anything.

• Impressions, clicks and average position for the Track 2 pages you’re about to edit.
• Which domains the AI Overview currently cites for your ten sample queries.
• The date, so your before-and-after has an actual before.

Re-check after four to six weeks. Re-run the ten-query spot check at the same time. Prevalence shifts underneath you, so a flat result sometimes means the AI Overview simply stopped appearing.