eCommerce GEO: A Guide to AI Search Optimisation for Online Stores

Product discovery has moved. AI search for eCommerce is now a significant and fast-growing traffic source, and that traffic converts at a materially higher rate than organic search. Generative engine optimisation (GEO) for online stores is no longer optional: it is the discipline that determines whether products show up where buyers are looking.

• +693% YoY growth in retail traffic from generative AI platforms. Source: Digital Commerce 360 / Adobe, Jan 2026
• +254% YoY increase in revenue per visit from AI-referred traffic. Source: Digital Commerce 360 / Adobe, Jan 2026
• 50% Higher conversion from AI-referred shoppers vs organic search (product pages). Source: Shopify AI Search Insights, Q1 2026

These numbers reflect the 2025/2026 holiday season, but the trend is structural. On Google’s AI Mode, product searches now return AI-generated summaries with embedded comparisons and ranked lists; and in regular search on Google, organic blue links are pushed well down the page, below AI Overviews. 

On ChatGPT’s Shopping Assistant, users run full conversational research sessions without leaving the chat. 

On Perplexity Shopping, product carousels come with clearly sourced results.

This is the new shelf. Products that do not appear here when buyers are researching are simply absent from the consideration set.

eCommerce Product Pages and Commercial Intent

That same shift plays out differently lower in the funnel, where buyers already know roughly what they want to buy. When a user types a high-intent query into Google (for example, one containing buy), what they see is product grids with advanced filtering, not organic links, and often not AI Overviews either. The Google Shopping layer dominates the commercial SERP.

What this means is that a perfectly optimised product page can still lose out if the structured product data feeding into Google Shopping surfaces is incomplete. Page optimisation and data optimisation are separate problems, and both need attention.

eCommerce GEO: The Technical Blueprint

Closing that data gap starts with three technical fundamentals. Everything else in eCommerce GEO builds on these:

1. AI bot access: can AI crawlers actually reach the store?
2. Structured data: is the product data giving AI the signals it needs?
3. Product feed optimisation: is the product data accurate, complete, and machine-readable?

Pillar 1: AI Crawlability and Technical eCommerce SEO

Crawlability is the foundation. If AI bots cannot reach a product page, they cannot index, understand, or surface it. That sounds straightforward (and it is), but the most common crawl blockers are not where people expect to find them.

robots.txt is the right starting point but it is not the full picture. Blocks regularly appear at deeper layers of the crawl stack, and several of them are silent: the bot appears to be allowed in robots.txt but the request never reaches the page. 

So for an eCommerce website that strives to appear in AI results, the technical SEO audit needs to cover the full infrastructure, not just the surface layer.

Layer	Common Issue
Hosting provider	Infrastructure-level restrictions that bypass robots.txt entirely
CDN / Cloudflare WAF	AI bots silently blocked despite being allowed in robots.txt
robots.txt	Key bots blocked, or allowances missing
Product page	Staging environment access rules carried into production

Pillar 2: Structured Data for eCommerce Product Pages

Once a store is reachable, the next question is what AI finds when it lands there: structured data. The relationship between schema markup and AI visibility in eCommerce needs careful framing. Two pieces of current evidence:

• Google’s AI Optimisation Guide (May 2026) states that schema is not required to appear in AI results. AI can extract from well-structured content without it.
• Ahrefs research found no significant correlation between schema markup and increased AI citations.

So the case for product schema in eCommerce is not about citations. For SEO on eCommerce product pages, the goal is product visibility in Shopping surfaces. Product schema powers Rich Results and unlocks full-featured merchant listings. Those surfaces are increasingly connected to AI-powered discovery, particularly in Google’s AI Mode. That is the chain that matters.

Don’t implement product schema expecting more AI citations. Implement it because AI shopping experiences rely on those product signals.

Filip Ruprich, Chief Performance Officer, SUSO

Product schema properties for eCommerce SEO

In practice, that means prioritising a handful of properties over a blanket implementation:

Property	Why It Matters	Example
name, image, sku	Core identification	“Nike Air Zoom Pegasus 41”
offers	Pricing and availability	£129, In Stock, 2-day delivery
aggregateRating, review	Trust and CTR signals	4.8/5 from 1,200 reviews
hasVariant, ProductGroup	Consolidates product variants	Same t-shirt in S/M/L, 6 colours
isSimilarTo	Surfaces alternatives in AI results	Stanley Quencher ↔ Yeti Rambler
isAccessoryOrSparePartFor	Product relationships	Laptop charger → MacBook Pro
additionalProperty	Decision-making attributes	Waterproof, BPA-free, Material: Steel
FAQ / inline Q&A	Conversational search queries	“Will this fit a 15-inch laptop?”

The most commonly overlooked properties are isSimilarTo, hasVariant, and isAccessoryOrSparePartFor. These help AI understand what is related, what works together, and what serves as an alternative: context that is increasingly valuable in conversational shopping scenarios.

Product image optimisation for AI search

Schema covers structured product data. The other half of that picture is visual, and it matters just as much for discoverability as for conversion. A generic stock photo with a machine-generated filename and no alt text is basically invisible to visual search. A well-optimised product image should have:

• A descriptive filename (not IMG_4832.jpg)
• Meaningful alt text
• High resolution with multiple angles where relevant
• The image property linked in Product schema

Pillar 3: Google Merchant Center and Product Feed Optimisation

Schema and imagery shape what AI understands on the page itself. The data feeding shopping surfaces sits one layer further back. Google Merchant Center (GMC) is no longer just the infrastructure behind shopping ads. AI systems now verify stock levels, pricing, and shipping speed directly from product feeds, without the user visiting the store. A product with stale pricing, inaccurate inventory, or incomplete attributes may be deprioritised or dropped from AI-generated results entirely.

GMC optimisation checklist:

• Real-time stock accuracy (AI deprioritises out-of-stock products)
• Competitive pricing (AI compares across merchants)
• Shipping speed (faster fulfilment may rank more prominently)
• Accurate product attributes: title, GTIN, category

New: Google Merchant Center conversational feed attributes

In May 2026, Google introduced a set of supplemental feed attributes that most stores have not yet implemented. These sit on top of a standard Merchant Center feed and give AI systems richer product context. They can be submitted via a supplemental feed (TSV/XML/Sheet) or the Merchant API.

Attribute	What It Captures	Example
question_and_answer	Q&A pairs for long-form queries (highest leverage)	“Is it dishwasher safe?” → “Yes, top rack only.”
document_link	Links to supporting documentation	Size guide, care instructions, spec sheet
related_product	Complementary items	Coffee machine → matching descaler
item_group_title	Human-readable variant group name	“Organic Cotton Men’s T-Shirt”
variant_option	How a variant differs	Colour: Forest Green · Size: M
popularity_rank	Relative bestseller signal	95.5 (top seller) vs 34.1

The question_and_answer attribute is the highest-leverage entry point. It lets you pre-feed AI the exact questions shoppers ask and the answers you want it to provide (“Is this waterproof?” or “How long does the battery last?”), rather than leaving that to inference. Most stores have not implemented these yet, making it a clear early-mover opportunity.

You’re effectively feeding AI the exact questions your customers may ask and the answers you want it to give. That’s incredibly useful for conversational shopping.

Filip Ruprich, Chief Performance Officer, SUSO

Agentic Commerce and the Universal Commerce Protocol

Bot access, schema, and a clean feed are the groundwork. What they increasingly feed into is full transactional AI, known as agentic commerce. The Universal Commerce Protocol (UCP) is Google’s open eCommerce standard that enables AI agents to find products, apply discounts, and complete purchases, moving beyond reading product data into actively supporting transactions.

What AI agents can do inside UCP today:

• Purchase directly inside Google AI Mode and Gemini (no click to your site required)
• Pay via tokenised checkout through your existing payment stack
• The merchant stays as the Merchant of Record, retaining the customer relationship and data

On the roadmap:

• Build and edit multi-item carts in conversation
• Link loyalty and account programmes
• Handle post-purchase flows: tracking and returns

How to get started with UCP and agentic commerce

Google provides documentation for how to implement the Universal Commerce Protocol. Here are the essential steps to get started:

Step	Action
1	Prepare Merchant Center: feed, shipping, and returns in good standing
2	Join the UCP waitlist and await Google approval
3	Set up Google Pay for checkout via Google Wallet
4	Publish your UCP profile so Google can locate your endpoints
5	Build native checkout: create, update, and complete sessions
6	Choose identity method: guest checkout or account-linked sign-in
7	Sync orders: push status updates back to Google

UCP access is not automatic. It requires waitlist approval, and a self-serve integration tab in Merchant Center is currently US-only in a limited pilot. If this is on the roadmap, the time to apply is now.

What Merchant Center needs for UCP

Step one above, preparing Merchant Center, breaks down into five concrete requirements:

Requirement	Detail
Baseline	Account in good standing; products approved for free listings
Return policy and support info	Required at account level
native_commerce attribute	Opts each product into agentic checkout. Missing or false = excluded.
consumer_notice	Required for products with regulatory warnings
merchant_item_id	Maps feed ID to Checkout API ID where they differ

Product types not eligible for UCP checkout

Not every product can use this flow yet. The following categories are excluded for now:

• Subscriptions and instalment plans
• Personalised or made-to-order goods
• Used, refurbished, and no-return items
• Pre-orders and bundled services
• Age-restricted and prohibited categories
• Services, rentals, digital goods, and virtual goods

Shopify & AI Search: The Agentic Sales Channel

Everything in the blueprint above applies across platforms. Shopify gets its own chapter because it has built agentic discovery directly into the platform, with its own catalogue, dashboard, and diagnostic tools. This chapter covers how those Shopify-specific tools work, with input from Mack Johnson, Director of Commerce at eHouse Studio, a Shopify Platinum Partner.

How Shopify’s agentic commerce channel works

As of 1 May 2026, Shopify’s agentic sales channel is automatically enabled on all eligible live stores. No setup required. Shopify treats this as a core channel alongside the traditional online store, POS, and B2B, connecting products to AI-powered shopping surfaces such as ChatGPT, Microsoft Copilot, Google’s AI Mode and Gemini.

To access it, go to Settings → Sales Channels → Agentic, or append /mappings/shopify-catalog-mapping to the admin URL.

The agentic dashboard shows:

• Sessions and revenue attributed to AI channels
• Test query results: how products actually appear in AI chat search
• Completeness scores with deficiencies flagged against Shopify’s global catalogue top ten

Shopify also offers an Agentic plan for brands not running a full Shopify storefront. As long as products are in the catalogue and meet eligibility requirements, brands can participate in AI-driven product discovery without a full platform migration.

The Shopify catalog and AI search optimisation

That agentic dashboard runs on a single underlying system, so it is worth understanding what is actually in it. The Shopify catalog is a standardised, machine-readable index of every eligible product across Shopify’s merchant base. When AI surfaces product recommendations, it queries this catalog in real time: pricing, availability, variants, product relationships. If the data is accurate and complete, AI platforms can find and recommend the product. If it is messy, incomplete, or absent, the product is either invisible or misrepresented.

Every Shopify store is automatically included. Whether what is in there actually performs well is a separate question entirely.

The gap between ‘yeah, we’ve got product data’ and ‘our product data is actually working for us in AI, and we know it is’ is where most brands are sitting right now, and they don’t know it.

Mack Johnson, Director of Commerce, eHouse Studio

By default, Shopify maps product title, description, and category into the catalogue representation. That is fine as a baseline. The problem is that most product descriptions were written for a human browsing a traditional product page, not for the conversational queries AI is now fielding. Queries like “blue jeans under £80 with free shipping arriving in two days” or “a cosmetic bag as a gift for a work colleague” were never the target when that copy was written. If those intent signals are absent from the data, the product will not show up for those queries.

When data is incomplete, Shopify attempts to fill gaps using AI inference. If the underlying inputs are vague or off-purpose, the inference will be equally vague, and the brand loses control of how its products are represented.

Thoughtfully mapped data puts you in control, rather than leaving it to chance.

Mack Johnson, Director of Commerce, eHouse Studio

AI store optimisation: Shopify completeness checks

Shopify scores all of this automatically. The agentic dashboard runs completeness checks against four baseline factors, and flags exactly where a store falls short:

Factor	What Shopify Checks
Description length	Sufficient detail to match conversational queries
Image count	Multiple images covering angles and variants
Review volume	Quantity of customer reviews
Average rating	Aggregate trust signal

Store policies (returns, shipping) are also checked.

In practice, most stores pass on descriptions and policies. Images and reviews are where the gaps consistently appear. A single product image and a handful of reviews (or none) will knock products out of the global top ten results, even when everything else is solid. Both are fixable, and the dashboard surfaces exactly which products need attention.

Using the Shopify catalog query tool

For a more hands-on view of the same gap, Shopify also provides a catalog query developer tool. It lets you run real queries against the full Shopify catalogue and see the exact output an AI agent would receive. Type in a shopper’s query, and Shopify returns a JSON output showing whether the store’s products rank in the global top ten, why or why not, and what the competing products that do appear look like.

This is the most direct diagnostic available. It shows the gap between what a merchant believes their catalog says and what an AI agent actually sees, including which fields Shopify is inferring rather than reading from structured data.

Common eCommerce GEO pitfalls on Shopify

Run that query tool against enough stores and the same five issues turn up again and again.

1. Empty category metafields

When a product is assigned to a category taxonomy, Shopify automatically generates suggested metafield values: colour, materials, style, and other product-specific attributes. Those suggestions sit unactioned until someone accepts them. The taxonomy gets assigned; the metafields stay empty. These fields are exactly what help AI understand the specific context of a product, not just its name.

2. Weak or mismapped product descriptions

Descriptions written for traditional product pages tend to lead with specifications (“100% cotton, machine washable”). AI search queries are intent-driven and occasion-based (“perfect for gifting”, “suitable for sensitive skin”). If that language is absent from the product data, the product is invisible to those queries.

3. Non-standard product titles

Titles that work well in brand-led editorial (creative, evocative, non-taxonomic) can undermine AI’s ability to categorise and match products to structured queries. Titles need to follow a predictable format aligned to how shoppers search.

4. Fragmented variants

Variants set up as separate products rather than grouped correctly make it harder for AI to understand the full product offering and present it accurately. A shopper asking for “the grey version” of a product should not be sent to an unrelated listing.

5. Missing intent keywords

Product data needs to contain the language of intent, not just specification. If the data does not include signals like “gift”, “travel-size”, “quick-dry”, or “eco-friendly”, those queries will not return the product, even if it is objectively a good match.

Measuring eCommerce AI Search Performance

Fixing the pitfalls above is half the job. The other half is knowing whether the fixes are actually working, which comes down to three places to check.

Step 1: Technical eCommerce SEO audit for AI bot access

Auditing a store’s AI search optimisation starts with the full crawl stack. Confirm access for:

• GPTBot
• GPT-User
• OAI-SearchBot
• Googlebot
• Gemini-Deep-Research
• PerplexityBot
• Perplexity-User
• ClaudeBot
• Claude-User
• Claude-SearchBot

Then check for silent blocks in CDN/WAF rules (Cloudflare in particular), confirm hosting provider defaults are not restricting AI crawlers, make sure that robots.txt allows AI bots, and verify that staging access rules have not carried over to production.

Step 2: Search Console reports for eCommerce product pages

With bot access confirmed, the next check is whether the structured data behind those pages is actually valid. Search Console offers the most direct health check for structured product data, even though it is not built for AI-specific reporting. Three reports are relevant:

Report	What It Shows
Product Snippets	Confirms Product structured data is valid
Merchant Listings	Confirms full-featured merchant listing data is valid
Merchant Opportunities	Flags missing merchant data (delivery policy, returns, payment methods)

Filtering the Performance report by Search Appearance gives click and impression data broken down by rich result type. Use this to track whether structured data improvements are moving CTR, and to identify which pages and queries are generating rich results.

Step 3: Product analytics in Merchant Center

Search Console confirms the data is structured correctly. Merchant Center goes a step further and shows whether it is actually performing. The product analytics report surfaces signals that directly inform both data quality and wider catalogue strategy.

Signal	Action
Popular now and trending up	Keep in stock. Fix any “needs attention” flags immediately.
Trending down	Investigate early before the drop compounds.
High impressions, low CTR	Review the image, title, and price against competing listings.
Organic vs ads, online vs local	Segment to isolate what is actually driving demand.

Upcoming AI search measurement tools for eCommerce

Everything above is built for shopping ads and structured data rather than AI specifically, which makes it necessarily indirect. That is starting to change. Two measurement tools currently in rollout will make AI-specific performance data more visible:

• Google AI Performance Insights (Merchant Center): direct visibility into product performance across AI-generated results. 
• Shopify Agentic Readiness: Shopify’s own assessment of how prepared a store is for agentic commerce. Available at shopify.com/agentic-readiness.

FAQs

• What is eCommerce GEO?

  Ecommerce GEO (Generative Engine Optimisation) is the practice of making online stores and product data visible and understandable to AI-powered search platforms: Google’s AI Mode, Gemini, ChatGPT, and Perplexity, among others. Where traditional SEO is concerned with ranking in organic blue-link results, eCommerce GEO is about appearing in AI-generated product summaries, recommendations, and agentic shopping experiences.

• How is eCommerce GEO different from traditional eCommerce SEO?

  Traditional eCommerce SEO focuses on ranking for keywords in Google’s organic results. eCommerce GEO focuses on making product data accessible and understandable to AI systems that generate answers and recommendations directly. The biggest practical difference is that AI does not just crawl pages: it queries product feeds, reads structured data, and verifies information without the customer visiting the site at all. Page-level optimisation still matters, but it needs to be paired with clean product data and feed accuracy.

• Does schema directly improve AI citations for eCommerce stores?

  No. Google’s AI Optimisation Guide confirms that schema is not required to appear in AI results, and Ahrefs research found no significant correlation between schema and AI citation rates. For eCommerce, product schema matters because it powers Rich Results and Shopping surfaces that feed into AI discovery, not because it generates citations directly.

• What is agentic commerce? Do I need to prepare my store for it now?

  Agentic commerce refers to AI agents that can complete product purchases autonomously on a user’s behalf: finding products, comparing prices, and checking out without the user visiting a retailer’s site directly. Google’s Universal Commerce Protocol and Shopify’s agentic channel are the two primary live implementations. Consumer adoption is still early, but the data foundations required for agentic commerce (clean Merchant Center feeds, accurate product data) are also the foundations for strong AI search performance today. Preparing now serves both goals.

• If a store is on Shopify, is it already set up for AI search?

  Shopify stores are in the agentic catalog by default, which is a head start. Being in the catalog and performing well in it are different things. Most stores have incomplete category metafields, product descriptions that do not map to conversational queries, and limited image and review coverage. The agentic dashboard and catalog query tool show exactly where the gaps are.

• Can brands be part of Shopify’s catalog without a full Shopify storefront?

  Yes. Shopify’s agentic plan allows brands that do not run a full Shopify storefront to participate in AI-driven product discovery, provided products meet eligibility requirements. It is a lower-barrier entry point than full platform migration.

• How do we check which AI bots can crawl a store?

  Make sure that AI bots are allows in robots.txt, check WAF rules at the CDN level (Cloudflare is the most common source of silent blocks), and confirm hosting provider defaults. SUSO’s free AI Search Visibility Checker is a practical first step before running a full technical audit.

• What are conversational feed attributes in Google Merchant Center?

  They are a set of supplemental attributes submitted alongside a standard Merchant Center feed to give AI additional product context. The most valuable is question_and_answer, which lets merchants pre-answer the questions shoppers are likely to ask (“Is this dishwasher safe?”, “Does this come in wide fit?”), rather than leaving AI to infer. Other attributes cover related products, variant descriptions, documentation links, and popularity signals. Most stores have not implemented these yet.

• How is eCommerce AI search performance measured right now?

  Current tools: SUSO’s AI Search Visibility Checker for an initial visibility read, Search Console shopping reports for structured data health, and Merchant Center product analytics for feed performance and demand signals.

  Google Analytics also shows traffic, revenue and conversions directly attributable to LLMs in the Traffic Acquisition report filtered down to the AI Assistant Channel. You can also use the Attribution Paths report to see if AI traffic participated in a conversion path when a conversion got attributed to another source. This is provided you have conversions tracking set up properly.

  Google’s AI Performance Insights (rolling out in Merchant Center) will add more direct AI-specific measurement as they become widely available.

  If your online store is a part of the Shopify catalog and , you should also be able to how many sessions and what revenue is attributed to AI channels. Additionally, Shopify offers a catalog query developer tool that allows testing whether your products appear for your target conversational queries when checked against the entire Shopify catalog. And finally, Shopify provides an Agentic Readiness assessment tool for testing individual product pages.

• Should eCommerce stores invest in blog and editorial content for AI search?

  Yes. Semantic and conversational search means product discovery is frequently also brand discovery. A shopper asking questions about products in your niche is often also a potential buyer, and the store with the most detailed, contextually rich answer to that question builds the trust signal AI needs to recommend it with confidence.

  The highest-value content for eCommerce is experience-led: product comparisons, expert reviews, use cases, original research. The closer the content is tied to specific products and buying decisions, the more useful it is for AI referral. Basic informational topics are lower priority.

• If a customer completes a purchase through AI Mode without visiting the site, what data does the merchant receive?

  The order and shipping data needed to fulfil the purchase comes through. What is lost is the analytics layer: pixel data, session duration, source attribution, and the broader behavioural data typically captured through a site visit.