How to appear in Google AI Overviews as an online store

On a growing number of shopping searches, Google shows a boxed answer before the classic list of ten links. It usually has a short written summary, sometimes a row of product cards with photos and prices, and it sits directly above the results a shopper used to scroll through first. That's an AI Overview, and for many "best X for Y" and "where to buy Z" searches, it's the only thing a shopper reads before clicking away, checking a specific store, or refining the question.

For a store owner, the tempting question is "how do we rank in AI Overviews." The more useful question is "which part of the box are we even talking about," because an AI Overview is not one thing built from one signal. It stacks two different mechanisms in the same visual box, and each one rewards a different kind of work.

What a shopper sees in an AI Overview

Open an AI Overview on a shopping query and there are usually two distinct pieces, even though they read as one panel.

The first is a short written answer: two to five sentences or a few bullet points that summarize the topic, often with small linked citations pointing to a handful of pages it drew from. This is the part that shows up on comparison-style questions - "best budget espresso machines," "is [product category] worth it for a small kitchen."

The second is a horizontal row of product cards: a photo, a price, a store or brand name, sometimes a rating, each one a separate tile a shopper can tap through to a listing. This shows up more on transactional queries - "buy espresso machine under 300," a specific model name plus "price."

Not every query triggers both pieces, and not every store that appears in one appears in the other. A store can be cited in the written answer and completely absent from the product row on the same page, or the reverse. Treating the whole box as a single ranking to chase is the first mistake worth avoiding.

Two tracks: cited pages vs product cards

The written summary and the product cards are built from different pipelines, and knowing which one a store is trying to influence changes what work actually matters.

The written overview: crawled and cited pages

The text portion works the way most AI-generated answers work: a system reads pages across the web that look relevant to the query, synthesizes a short answer from what several of them say, and links to a handful of the pages it drew from. Being cited here depends on the same things that decide whether a page gets pulled into a normal featured snippet: is the page indexed at all, is it reachable by a crawler, and does it state a plain answer to the actual question a shopper is asking, in text a system can extract with confidence.

Product cards: the Shopping Graph

The product row draws from Google's own shopping data - what's usually called the Shopping Graph. That graph is built primarily from product feeds retailers submit through Google Merchant Center, combined with structured product data Google's crawlers find directly on a page. Getting a specific product into that carousel is a Merchant Center and feed-management question, handled through Google's own tools and account settings. No third-party visibility or GEO tool submits a feed on a store's behalf or manages a Merchant Center account - that setup and its data quality stay squarely with the store and whoever runs its feed.

The two tracks share one thing: both need Google to have a working, crawlable, accurately-described version of the page in the first place. Past that shared foundation, they diverge.

What still looks like normal SEO

It's tempting to look for a special tag, a hidden schema type, or an "AI Overview meta field" that unlocks the written summary. There isn't one. Google has been consistent that AI Overviews are generated from the same crawled and indexed web content that feeds its regular search results, and that the way to be eligible is the same as the way to rank well normally: a page needs to be crawlable, indexed, and worth citing.

In practice that means the boring list still applies. A page needs to load without errors for a crawler, avoid blocking the bots that gather this content, state its actual content in HTML rather than only in an image or a script that never renders for a bot, and directly answer the kind of question a shopper is likely asking rather than only carrying a marketing tagline. A comparison page that never states which product is better for which use case, or a category page built entirely from hero images with no real text, gives a summarizing system nothing to extract and cite - the same problem that keeps a page out of a classic featured snippet keeps it out of an AI Overview citation.

None of this is unique to AI Overviews. It's the same groundwork covered in a GEO checklist for online stores: crawlable pages, real answers in text, and content that states facts plainly instead of hinting at them.

Product data that shopping surfaces actually need

The direct route into Google's shopping carousel runs through Merchant Center, and that stays true regardless of what's on the page. But on-page Product and Offer schema, matching what a shopper actually sees, is still worth getting right for two separate reasons that have nothing to do with a feed.

First, it's one of the inputs the Shopping Graph itself can pull directly from a crawled page, on top of whatever a feed supplies. Second, and often more immediately useful, it's exactly what other AI shopping assistants read when deciding whether to name a product with confidence. A model reading a product page to decide whether to recommend it in an answer leans on the same structured facts: current price, currency, availability, and a link back to the exact page those facts came from. That mechanic is covered in more detail in does ChatGPT recommend your store, and it applies whether the assistant asking is ChatGPT, Gemini, or the written layer of an AI Overview.

A few things are worth checking on the pages that actually sell:

• Price and currency in the schema match what's rendered on screen. A price that's only correct in an image or only correct in the schema, but not both, reads as unreliable to anything trying to cite it.
• Availability reflects real stock. A product marked in stock in schema while the page shows "sold out" is the kind of mismatch that erodes trust in the whole page, not just that one field.
• A stable identifier is present - SKU or GTIN where one exists - so the same product can be matched consistently across a feed, a page, and anything reading the page directly.
• The schema is generated from the product data, not copy-pasted once and left static across every product on the catalog.

None of this replaces a Merchant Center feed for the shopping carousel specifically, and no visibility tool fixes a feed on a store's behalf. It does raise the odds a page is read and cited confidently everywhere else a shopper might ask, including inside the written half of an AI Overview and inside other AI assistants.

How to check if you show up

Checking the two tracks means checking two different things, and neither one is a single yes or no from a single search.

For the written overview, the direct method is manual: search a handful of the actual questions shoppers ask before buying in a given category, read whatever overview appears, and note whether a domain is one of the linked citations. Run the same query again a few days later - overviews change as pages get recrawled and as the underlying answer shifts, so one look is a snapshot, not a verdict.

For the product carousel, that's a Merchant Center question first: whether products are actually approved and eligible in the account, which Google's own diagnostics inside Merchant Center report on directly. That side sits outside what a GEO or AI visibility tool tracks.

Where a tool like Aigely helps is the written, cited layer: tracking, on a repeating schedule, whether a store gets named in Google AI Overviews answers to real shopping prompts, alongside how it does in ChatGPT and other assistants on the same prompts. That's engine coverage, not a stand-in for Merchant Center diagnostics.

Worth being precise about scope here: the free Aigely Scan runs ten shopping prompts against ChatGPT only, for one domain, once a week, with no card required - it does not check Google AI Overviews. Tracking whether a store is named in AI Overviews answers is part of the paid engine set: Starter tracks ChatGPT and Google AI Overviews together on a weekly schedule, Pro adds Gemini, and Business adds Claude on top of the rest. A 7-day trial includes AI Overviews and Gemini alongside ChatGPT, though not Claude, which is enough to see the written-citation side across the assistants that matter most before choosing a plan.

A practical order of work

Given two tracks that respond to different work, a sensible order looks roughly like this:

1. Get the crawlability basics right first. Both tracks need Google to be able to reach, render, and index a page before anything else matters. A page that errors out, blocks crawlers, or hides its content behind client-side JavaScript that never renders for a bot loses on both tracks before content quality is even in play.
2. Write pages that plainly answer the question a shopper is actually asking. This is what feeds the written overview: a real answer in text, not just a headline and a hero image, on the category and comparison pages a shopper is likely to land on before buying.
3. Add or fix Product and Offer schema so it matches the visible page exactly. Price, currency, availability, and a stable identifier, generated from real product data rather than copy-pasted once. This helps the Shopping Graph read a page directly and it's exactly what other AI assistants check before naming a product.
4. Keep the Merchant Center feed itself in sync separately. If a store runs Shopping ads or free listings, the feed's own accuracy and approval status are a Merchant Center task, checked through Google's own account diagnostics rather than anything a GEO audit measures.
5. Track it on a schedule instead of checking once. Run a baseline with a free ChatGPT scan, then move to a plan that also tracks Google AI Overviews once weekly tracking matters. See how tracking works before deciding how much coverage a store needs.

Start with the free Aigely Scan to see where a store stands on ChatGPT today, and compare plans to add Google AI Overviews, Gemini, and Claude tracking once weekly visibility across engines is worth paying for.