AI Overviews and eCommerce Search: What Store Owners Need to Know

A client opened Search Console with me a few weeks back and pointed at a graph that made no sense: impressions for “best waterproof dog coat” holding steady, clicks quietly sliding month on month. No algorithm update, no ranking drop, the listing was still exactly where it always had been. What had changed was that Google was now answering the question itself, above the results, before anyone reached that listing.

Most advice about AI Overviews treats this as one problem with one fix, usually “add more schema” or “write more helpful content.” It isn’t. For a store, AI Overviews and their cousins in ChatGPT and Perplexity are two separate systems making two separate decisions: whether your brand gets mentioned at all, and whether your product gets sold through. Getting the first without the second is common, and invisible in Search Console until you know what to look for.

This goes deeper than the short section on AI Overviews in our guide to choosing an eCommerce SEO agency, which covers who does this work for you. This one is about what the work actually is: what triggers these features on shopping queries, and what you can check yourself this week.

Two Different AI Systems Are Deciding Whether Your Products Get Seen

I think of this as two separate systems, because that’s genuinely how it behaves, even though Google gives neither an official name. Call the first one the citation: the AI Overview text block with linked sources, drawn from Google’s standard web index, the same one that’s always crawled and ranked your pages. Google’s own documentation is unambiguous about it: “There are no additional requirements to appear in AI Overviews or AI Mode… You don’t need to create new machine readable files, AI text files, or markup to appear in these features. There’s also no special schema.org structured data that you need to add.” Good content and normal SEO fundamentals earn this one.

Call the second the card: the Shopping Graph, the price-rating-image results that show up inside AI Overviews and AI Mode for shopping queries specifically. It draws on your Merchant Center feed, but not only that: Google also crawls structured data in your page’s HTML directly, and runs its own shopping crawler, Storebot, over product descriptions, pricing and review content on the page itself. A feed alone won’t fully populate it, and neither will page content alone. Google’s VP of Ads & Commerce, Vidhya Srinivasan, confirmed at Google I/O 2026 that the Shopping Graph now holds more than 60 billion product listings, a figure that’s already moved up twice in eight months (Google’s own number, an indicator of scale rather than an audited stat). A beautifully written product page with no Merchant Center feed behind it will still struggle here, and so will a pristine feed sitting behind a page Storebot can’t parse.

The practical split still holds: content quality and structured data earn citations in the text; a complete feed plus crawlable structured data earns placement in the cards. A store needs both, won through overlapping but distinct work. This is exactly the groundwork our generative engine optimisation service is built around: getting a site where both systems can find and trust what you sell, not treating “AI visibility” as one checkbox.

Why AI Overviews Show Up on Some Shopping Searches and Not Others

Here’s a pattern worth knowing before you chase AI Overviews on the wrong pages: Google shows them far more readily on research queries than ready-to-buy ones. Industry tracking through 2026 has AI Overviews appearing on the clear majority of evaluation-style queries like “best air fryer,” a sharp rise on a year earlier, while transactional queries such as “buy air fryer” sit meaningfully lower and have held there. Exact numbers vary by tracker, and different studies measure different things here, trigger rate versus composition, so treat any single percentage with caution. Some trackers also reported Google pulling back coverage over the 2025 holiday season on specific product names, prices and “deals” keywords, while holding steady on comparison terms such as “best X” and “X vs Y.” Treat the figures as directional, not confirmed policy, the pattern holds well enough to plan around: don’t expect an AI Overview on the product page itself. The opportunity sits upstream, at the comparison stage, so buying guides, “X vs Y” pages and “best X for [use case]” content matter more here than the transactional pages you’d normally optimise hardest.

There’s a mechanical reason AI handles these queries differently, too. Google’s systems break a query like “backpack for hiking in wet weather” into sub-searches on individual attributes, weather resistance, capacity, material, before matching products against them. Google itself calls this “query fan-out,” a term it introduced when launching AI Mode, a genuine shift from keyword matching. A page that only says “backpack” without addressing weather resistance can lose out to one that does, even at similar rankings.

How Placement in the Cards Actually Gets Decided

Back to the Shopping Graph. It isn’t an auction: there’s no Shopping Ads-style bid behind AI Mode’s product cards. Placement comes down to how completely your product data matches the query: feed completeness, review depth, specificity of attributes.

Two features make the point. Virtual try-on, launched in Search Labs in the US at Google I/O 2025, lets a shopper upload a photo and see how apparel would look on them, fabric fold and fit included. Agentic checkout, sometimes called “buy for me,” lets a shopper set a target price and specs; Google adds the item to cart and checks out via Google Pay once the price drops, no site revisit needed. Both are gated on your Merchant Center feed being complete and current, not on how well-written your product description is.

Baseline eligibility requires a compliant Merchant Center feed (ID, title, link, image, price, description, availability, condition), a visible return policy and shipping settings, required in most major markets including the UK. Storebot also crawls your structured data and page content to cross-check the feed, useful for catching drift, not a substitute for submitting one. Free listings now extend to Gemini too, alongside Google’s other surfaces. Better blog content doesn’t fix a thin feed: no amount of on-page writing buys you eligibility for the cards.

What Makes a Product Page “Citable”: Structured Data That Actually Matters

Schema isn’t required for the AI Overview text citation. But it is required for the Shopping Graph track, and it’s the single most effective thing a store owner directly controls: not a guarantee, but the biggest lever available. For a developer checklist:

• Product and Offer schema: name, image and offers (with price and priceCurrency) are required. Worth adding anyway: aggregateRating, brand, gtin/mpn/sku, review, availability, hasMerchantReturnPolicy and shippingDetails.
• Put it in the initial HTML, not injected by JavaScript. Google is explicit that dynamically-generated markup makes Shopping crawls less reliable, especially for fast-changing data like price and availability. A common trap on Shopify stores running JS-rendered review or price widgets that only render client-side.
• Genuine product reviews are fine to mark up. Google’s 2019 policy banning “self-serving” star ratings applies to LocalBusiness and Organization markup (you can’t mark up “5-star reviews of us” on your homepage), not to Product review markup. Real customer reviews on a product page remain eligible for star snippets.

Two smaller things worth checking too, since both get missed even on otherwise well-built product pages:

• Check identifier_exists in your feed. GTIN, MPN and brand are Google’s three product identifiers. A common, easily fixed error is setting identifier_exists: false when a real GTIN or brand exists.
• FAQ schema is now cosmetic. Google dropped FAQ rich results from the SERP on 7 May 2026 after abuse of the markup for extra real estate. It still validates and won’t hurt rankings, it just doesn’t earn the visual accordion. The markup was never really what earned AI citation, the answer-shaped writing underneath it was, so well-written FAQ content still pays off.

The Comparison and Review Trap: Why AI Mentions Your Brand But Links to Someone Else

Gradial’s 2026 study analysed 28 retail brands across ChatGPT, Claude and Perplexity, over 1,600 data points. Average brand mention rate: 44%. Average URL citation rate, meaning the AI links to the brand’s site: 8%. One beverage brand was mentioned in 92% of relevant responses but linked to in just 3%. It’s vendor research on one sample, but the gap shows up repeatedly across similar studies.

Separate research from Aleyda Solis, an international SEO consultant, adds a retail angle: Shopping and Retail sites earn roughly 2.3 times more AI citations per site than sectors like finance, yet most AI-driven retail traffic (around 57.6%) lands on a brand’s homepage, not a product or category page, and a real share of AI-recommended purchases route through marketplaces instead. Brands get talked about, just not always linked to, and not always sent the shopper who holds the stock.

The honest takeaway isn’t “add schema and you’ll get cited.” The realistic goal splits into two jobs: be known and mentioned at all, which structured data, review volume and Merchant Center completeness genuinely help with, and make sure the third-party sources AI is actually citing, genuine reviews, Reddit threads, comparison sites, have accurate and current information about you. You’re not only competing with rival stores for the citation. You’re competing with Reddit, and increasingly with the marketplaces selling the same product.

Beyond Google: ChatGPT and Perplexity Are Building Their Own Shopping Layers

ChatGPT rolled out Instant Checkout, built on OpenAI’s Agentic Commerce Protocol, to US users including the free tier during 2026, though this space is still moving fast. Merchants submit a compressed product feed and checkout happens inside the chat, no site visit required, and OpenAI takes a percentage fee on completed purchases. More recently, OpenAI has said it’s pivoting away from pure in-chat checkout toward product discovery, admitting the initial version “did not offer the level of flexibility that we aspire to provide.” Not settled infrastructure, in other words.

Perplexity’s Merchant Program, free with no transaction fees, gets a merchant into Perplexity Shopping results plus a trends dashboard, and Buy with Pro offers one-click in-chat checkout via PayPal for US Pro subscribers. Eligibility for both leans on the same fundamentals as Google: clean Product schema, reliable inventory and price data, real review depth.

Your product feed has three audiences now: Google Merchant Center, ChatGPT and Perplexity. Worth watching, not a roadmap yet, given OpenAI’s pivot mid-2026.

How to Actually Check If This Is Happening to Your Store

Until recently, there was no way to isolate AI-driven traffic in Search Console: AI Overview clicks were folded invisibly into the standard “Web” search type.

That changed on 3 June 2026, when Google launched a Generative AI performance report inside Search Console, separating AI Overview and AI Mode impressions by page, country, device and date. No clicks or CTR yet, but it’s the first real visibility store owners have had. It’s rolling out to a subset of UK site owners first, so check whether you have it yet.

Some rigorous US data helps set expectations here. Pew Research Center analysed real browsing behaviour from 900 US adults across nearly 69,000 Google searches. When an AI summary appeared, users clicked a traditional result in 8% of visits, against 15% when no summary appeared. Only 1% of AI-summary visits involved clicking a link inside the summary itself, and over a quarter ended the session entirely: the user got an answer and left. A real, measurable shift, not a catastrophe, US data best treated as directional here.

On crawlers: GPTBot and ClaudeBot train models, distinct from OAI-SearchBot, PerplexityBot and Claude-SearchBot, which power live answers, so blocking one doesn’t affect the other. And a myth that won’t die: llms.txt does nothing for AI visibility, Google has said plainly it isn’t supported and isn’t planned.

What This Means for Your SEO Foundations

None of the above works without ordinary fundamentals underneath it: clean technical SEO, genuinely useful content, real customer reviews, a fast and well-structured site. AI systems grade a store on largely the same signals a human shopper would use, just automatically and at scale.

To be upfront about what our own evidence proves: our platform-SEO case study for Small Smart, a children’s products store, isn’t an AI Overview case study. It shows a 220% increase in ranking keywords and an eightfold increase in Christmas-intent traffic, through technical fixes, content and structure, the same groundwork AI answer engines now depend on to find, trust and cite a store, even though the work predates any of this.

If the foundational side is where you’re weakest, that starts with the same SEO fundamentals that have always mattered. The schema, crawler access and citability groundwork covered above is what our GEO service adds once the foundations are solid.

FAQ

Do I need schema markup to appear in AI Overviews?

Not for the text citation, Google is explicit about that. It’s the main lever for the Shopping Graph and product-card track though, so functionally yes, if you want the fuller picture.

What’s the difference between an AI Overview and Google’s AI Mode shopping features?

An AI Overview is the summary block with text citations, drawn from the normal web index. AI Mode’s shopping features, virtual try-on, agentic checkout, product cards, are a separate layer built primarily on your Merchant Center feed, cross-checked by Google’s Storebot against your page’s structured data.

Will ChatGPT or Perplexity actually send my store traffic?

Some, through Instant Checkout and the Perplexity Merchant Program, but both are new and still changing shape, not something to plan revenue around yet.

Can I see how much AI Overview traffic my store is getting?

Only very recently, and only partially. Search Console’s new Generative AI performance report, rolling out to UK site owners first, shows impressions by page, country and device, but no clicks or CTR yet.

Does llms.txt help my store show up in AI answers?

No. Google has said directly it isn’t supported and isn’t planned, despite how often it’s sold as a fix.

Should I focus on product pages or buying guides for AI visibility?

Buying guides and comparison content first, product pages still matter, but mainly for the Shopping Graph track rather than AI Overview text citations.