Best Practices for Ecommerce Sites in Google Search (2026 Complete Guide + AI Citations)

Most ecommerce SEO advice online is someone’s opinion. This guide isn’t. Every recommendation below traces back to Google’s own documentation — Search Central, Merchant Center, and the guidance Google published in 2026 on generative AI features — and we’ve flagged clearly where something is Google’s position versus our interpretation as an agency that does this daily.

That distinction matters more than it used to. In 2026, “ranking” isn’t one thing. Your products can surface in classic blue links, image results, the Shopping tab, Google Lens, Maps, AI Overviews, AI Mode, and inside AI assistants that never send a click at all. Each surface pulls from a slightly different pipe. Get the plumbing right and you show up in all of them; get it wrong and you can rank well in one place while being invisible everywhere else.

1. Start by Understanding Where Your Products Can Actually Appear

Before you touch a single tag, understand the surfaces you’re competing for. Google documents six distinct places ecommerce content can show up, and each has its own entry requirement.

Surface	What gets you in
Google Search	Standard SEO fundamentals; structured data unlocks product rich results
Google Images	Following Google’s image best practices for preparation and indexing
Google Lens	Product details uploaded to Google Merchant Center + opting in to product listings + image best practices
Google Shopping tab	Products uploaded to Google Merchant Center (this is mandatory, not optional)
Business Profile	Claiming your Business Profile and linking it to Merchant Center
Google Maps	Products with inventory location data in Merchant Center (via local inventory ads onboarding)

The single most common strategic mistake we see: treating “SEO” and “Merchant Center” as two separate departments. They’re one system. The Shopping tab is closed to you without a Merchant Center feed no matter how well-optimised your product pages are. And Lens — which is where a huge amount of mobile product discovery now starts — requires both a feed and good image hygiene.

Also worth noting: Google explicitly says shoppers are looking for more than just product pages. Its documentation recommends publishing your company story, special offers, merchant reviews, customer reviews, a browsable product catalogue with descriptions, educational content, live streams, and customer service information including return and shipping policies. That’s a content roadmap handed to you by Google, and most stores ignore it.

2. Share Your Product Data With Google – Both Ways

Google’s guidance here is unambiguous: use structured data on your pages and upload a feed to Merchant Center. Not one or the other.

Structured data on your product pages

This is markup embedded in your HTML that describes what a product is. It makes you eligible for product rich results, improves Google’s accuracy on price, discount and shipping cost, and — this is the underrated part — helps Google verify your Merchant Center feed against your live site. Feed-to-site mismatches are one of the most common causes of Merchant Center disapprovals.

Google’s own framing: “While structured data isn’t required to appear in Google Search results, it can help Google understand your page better and display it as a rich result.”

A Merchant Center feed

Not mandatory for Search, but mandatory for Shopping tab listings and several other surfaces. You have three ways to get data in:

Method	Best for	Update frequency
Automated feeds (built from crawled web content; structured data improves extraction accuracy)	Smaller sites that are updated less frequently	Tied to Google’s crawl of your site
Feed file uploads	Larger sites with frequent changes	Weekly, daily, or hourly
Content API	Stores where stock moves fast	Real-time

Why bother with a feed if your structured data is solid? Three reasons Google gives directly: it increases confidence that Google knows about all of your products (crawling is never guaranteed to be comprehensive), it gives you control over update timing, and it lets you share data that isn’t published on your website at all — store-level inventory being the obvious example.

Keeping the two in sync

When your feed and your site disagree on price or availability, you get disapprovals and, worse, wrong prices shown to shoppers. Google’s recommended fix is to configure Merchant Center to automatically update its copy of your product data from your website content when a discrepancy is detected. Turn this on. It’s one setting and it prevents a category of problem that otherwise eats hours every month.

3. The Structured Data That Actually Moves the Needle

Google names six structured data types as relevant to ecommerce sites. Here’s what each one does and how to implement it properly.

3.1 Product and merchant listings

First, understand the distinction Google draws, because it changes which properties you need:

• Product snippets are for pages where a shopper cannot buy directly — editorial reviews, comparison articles, roundups. These support richer review information including pros and cons.
• Merchant listings are for pages where a shopper can buy directly from you. These support detailed commercial information: shipping, returns, apparel sizing, availability.

Meeting merchant listing requirements generally makes you eligible for product snippets too, so if you’re a store, aim at merchant listings.

Required for merchant listings:

• Product.name
• Product.image — high-resolution, minimum 50K pixels, ideally supplied in 16:9, 4:3 and 1:1 aspect ratios
• Product.offers — an Offer object where you are the direct seller
• Offer.price (must be greater than zero) or priceSpecification.price
• Offer.priceCurrency — three-letter ISO 4217 code

Strongly recommended: availability, url, itemCondition, shippingDetails, hasMerchantReturnPolicy, priceValidUntil, plus brand, sku, gtin/mpn, description, color, material, pattern, size, and aggregateRating.

Two implementation warnings from Google’s documentation. Product rich results only support pages focused on a single product (or its variants) — category and aggregator pages don’t qualify. And your markup should be present in the initial HTML wherever possible; if you’re generating it with client-side JavaScript, you’re adding a rendering dependency and a failure mode you don’t need.

A 2026 update worth knowing: Product.category now accepts both Text and CategoryCode types, aligning with Merchant Center feed specs. Google also added a hasAdultConsideration property to merchant listing and variant documentation in May 2026.

3.2 Product variants (ProductGroup)

If you sell the same item in multiple colours or sizes, variant markup tells Google those pages are siblings, not duplicates. This is one of the highest-leverage fixes available to apparel, footwear and homeware stores, and almost nobody implements it correctly.

The structure: a parent ProductGroup that references child Product variants via hasVariant (or variants pointing back with isVariantOf).

Required: ProductGroup.name, plus the full merchant listing (or product snippet) required properties on every variant Product — Google’s words: “Each variant must have a unique ID in its corresponding structured data markup (for example, using the sku or gtin properties).”

Recommended, but effectively necessary in practice: a shared group identifier via productGroupID on the group or inProductGroupWithID on the variants — if you use both, they must match. Also variesBy (an array naming the dimensions that vary), plus shared brand, description and aggregateRating. Colour and size on each variant are recommended rather than formally required; Google advises making each variant’s description specific and using words that identify the variant, such as colour, size or material.

Two architectures, two rulesets. On a single-page setup where all variants live on one URL, there must be exactly one canonical URL for the whole ProductGroup — normally the base URL with no variant preselected — and each variant must still be directly reachable via a distinct URL using query parameters. On a multi-page setup where each variant has its own page, every page needs full, self-contained markup that doesn’t depend on entities defined elsewhere.

3.3 Organization, return policy and loyalty program

These go on your business information pages (About, Contact, Shipping & Returns), not on individual products. Google recommends nesting merchant return policies and loyalty program details under Organization — a site-wide default that individual Offer-level policies can override when a specific product differs.

Google’s Organization documentation also lists business identifiers worth adding if you have them: vatID, duns, leiCode, naics, and iso6523Code. No properties are strictly mandatory — include what genuinely applies to your business. Your logo must be at least 112x112px, crawlable, and legible on a white background.

For return policies specifically, Google accepts two configurations: either applicableCountry + returnPolicyCategory (+ merchantReturnDays when the category is a finite window), or simply a merchantReturnLink pointing at your policy page. The first gives Google structured detail; the second is the low-effort fallback.

3.4 BreadcrumbList

Straightforward, cheap, and it directly improves how your URL is presented in results. Mark up the full path a shopper took to reach the page — for example Men → Base Layers → the product itself — with each step numbered in order and linked to its own URL. The final step is the current page, so it doesn’t need a link of its own. Most ecommerce platforms will generate this automatically if your category hierarchy is set up properly, which is one more reason to get section 4 right.

3.5 Review and AggregateRating and the rule that disqualifies most stores

Review requires author (max 100 characters), itemReviewed (omit if nested inside the item), and reviewRating containing a ratingValue. AggregateRating requires the item identifier plus ratingValue and at least one of ratingCount or reviewCount. Both benefit from datePublished, bestRating and worstRating.

The rule people miss: if the entity being reviewed controls the reviews about itself, those pages are ineligible for the star review feature. This applies to LocalBusiness and Organization markup where reviews about your own business appear on your own website — including via embedded third-party widgets. Product reviews on your own product pages are fine; “5 stars, 400 reviews of our agency, marked up on our own homepage” is not.

A second rule added to Google’s documentation in July 2026 that directly affects stores running review-incentive programs: “Don’t include fake or undisclosed incentivized reviews on your page or in your structured data markup.” Google names two examples — reviews not based on a genuine experience of the product or service, and reviews written in exchange for a benefit such as money, discounts, vouchers or free products that don’t clearly and prominently disclose the incentivisation. If you offer a discount code for leaving a review, the disclosure has to be visible on the review itself, not buried in a terms page.

Separately, Google publishes detailed guidance on writing genuinely high-quality reviews, which matters if you run buying guides or comparison content. Its checklist: evaluate from the user’s perspective, demonstrate expertise, include visuals or audio evidencing your own experience, share quantitative measurements, explain what sets a product apart from competitors, cover comparable alternatives, discuss both benefits and drawbacks based on original research, describe how a product has evolved across generations, focus on the decision-making factors that actually matter, explain key design choices and their effect on users, link to other useful resources (including multiple retailers), and provide first-hand supporting evidence whenever you call something “best.” Ranked lists need enough useful content to stand on their own.

3.6 VideoObject and LocalBusiness

Add VideoObject on pages that are primarily about an individual video — prerecorded product videos or livestream events. Add LocalBusiness on pages showing physical store locations and opening hours, and pair it with a registered Business Profile plus store locations and store codes registered in Merchant Center.

Practical note: if you’re on a hosted ecommerce platform, Google explicitly suggests it may be easier to “use an integrated platform extension or plugin to add structured data for you” rather than hand-coding markup. (Google doesn’t name platforms; in practice this means the Shopify, WooCommerce, Magento and BigCommerce ecosystems.) That’s sound advice — but audit what the plugin actually outputs. Most default apps emit a bare-minimum Product block and skip shipping, returns and variants entirely, which is exactly where the competitive advantage sits.

We’ve deliberately described what each markup type needs rather than pasting code you’d have to adapt anyway. The properties above are the specification; how they get onto the page depends entirely on your platform, and a plugin setting is usually the right answer over a hand-written template edit. If you want the actual markup written and tested against your templates, that’s a technical SEO job rather than a copy-paste one — and it’s the single most common thing we get hired to fix.

4. Site Structure: Make Every Product Reachable by a Link

Google determines your site’s structure by analysing the links between pages — not by reading your URL patterns. The recommended hierarchy is explicit: menus link to category pages, category pages link to sub-category pages, and sub-category pages link to every product page.

The critical failure mode: if your category pages don’t connect to all your products, Googlebot may simply never find them. Products reachable only through your on-site search function will not be indexed. Google’s wording is direct — “It’s strongly recommended to link to all products that you wish indexed.”

Where direct linking genuinely isn’t feasible (very large catalogues, long-tail SKUs), sitemaps and Merchant Center feeds are the documented fallbacks for discovery.

Two more structural rules:

Use real links. Google says to use <a href> tags for navigation and explicitly warns: “Don’t use JavaScript events on other HTML DOM elements for navigation.” A <div onclick=”…”> is not a link to a crawler.

Internal links signal importance. Google infers relative page importance partly from how many internal links point at a page. If you have hero products or priority categories, link to them from your homepage, blog posts and newsletters. This is free, fully within your control, and consistently underused.

5. URL Structure

Google’s ecommerce URL guidance, condensed:

Do:

• Add descriptive words to URL paths — /product/black-t-shirt-with-a-white-collar, not /product/3243
• Minimise the number of alternative URLs returning the same content
• Standardise text case so URLs are easier to match as identical
• Give every paginated page its own unique URL
• Use ?key=value parameter format — /t-shirt?color=green, not /t-shirt?green
• Combine repeated values into one parameter — ?type=candy,sweet, not ?type=candy&type=sweet
• Keep URLs consistent across internal links, sitemaps and canonical tags
• Add a self-referencing canonical to every indexable page
• Use meaningful anchor text, not “click here”
• Apply noindex to empty category pages

Don’t:

• Internally link to temporary parameters — session IDs, tracking codes, time-based values
• Rely on JavaScript navigation instead of <a href>

For variants specifically, both /t-shirt/green (path segment) and /t-shirt?color=green (query parameter) are acceptable. When you use optional query parameters, designate the base URL — with the parameter omitted — as the canonical version.

6. Faceted Navigation and Crawl Budget

Faceted navigation is the single biggest crawl-efficiency problem in ecommerce. Google names two specific harms: overcrawling (crawlers can’t tell whether a filtered URL is useful without fetching it, so they fetch enormous numbers of useless combinations) and slower discovery of your genuinely new content as a result.

If you don’t want faceted URLs indexed

Option A – robots.txt. Block the filter parameters while keeping item pages crawlable. Google’s own worked example targets user-agent: Googlebot, then disallows the filter parameter patterns — /*?*products=, /*?*color= and /*?*size= — while adding one allow rule for /*?products=all$ so the full unfiltered listing stays crawlable. The principle: disallow the parameter, not the path, and always carve out an exception for the view you do want indexed.

Option B — URL fragments. Use #products=fish instead of ?products=fish. Google generally ignores fragments for crawling, so a fragment-based filter has no crawl impact at all — positive or negative.

If you do want some faceted URLs indexed

There’s real revenue in indexing high-demand facet combinations (“waterproof hiking boots size 10”). If you go this route:

1. Use standard & separators between different parameters. Avoid commas, semicolons and brackets as separators — crawlers struggle to parse them. (This doesn’t contradict the rule in section 5: & separates different keys, while a comma is how you pass multiple values of the same key — ?color=green&size=large&type=candy,sweet.)
2. Keep filter order consistent if you encode filters in the path (/products/fish/green/tiny).
3. Return a genuine HTTP 404 for empty filter combinations, duplicate filters, nonsensical combinations and non-existent pagination. Don’t redirect to a soft error page.

What works less well

Google is candid that rel=”canonical” pointing at the unfiltered version may gradually reduce crawl volume but is less reliable long-term than robots.txt. And rel=”nofollow” on facet links only works if you apply it to every link pointing at those URLs — one missed link anywhere and the door is open again.

7. Pagination, Load More and Infinite Scroll

The core constraint, in Google’s words: crawlers “generally crawl URLs found in the href attribute of <a> elements,” and “Google’s crawlers don’t ‘click’ buttons and generally don’t trigger JavaScript functions that require user actions to update the current page contents.”

Those two sentences determine everything else.

Pattern	Pros	Cons
Pagination	Shows result size and position	Content split across pages; more loads
Load more	One page for all content; can show total results	Struggles with very large result sets
Infinite scroll	Single page, intuitive scrolling	Unclear result size; poor performance at scale

Pagination rules

• Link each page to the next with an <a href> tag so Googlebot can discover the sequence
• Consider linking every page back to page one to signal its importance
• Give each page a unique URL — ?page=2 style query parameters work fine
• Give each page its own self-referencing canonical. Do not canonicalise page 2, 3, 4 back to page 1
• Never use URL fragments (#page=2) for page numbers — Google ignores fragments and may skip links it believes are duplicates
• Consider preload, preconnect or prefetch to speed up navigation between pages

Deprecated markup

Google no longer uses <link rel=”next”> and <link rel=”prev”>. They’re harmless and other search engines may still use them, but they do nothing for Google. If an agency is billing you for implementing them as a Google ranking fix, that’s a red flag.

Sort and filter variations

Block unwanted indexed variations — alternative sort orders like ?order=price — with either noindex robots meta tags or robots.txt patterns.

If you use load more or infinite scroll

Google’s crawlers won’t trigger the button or the scroll event. You must supply discovery another way. Google names two: sitemaps and Merchant Center feeds. A third option we implement regularly — not something Google spells out on that page, so treat it as our recommendation rather than Google’s — is a crawlable paginated fallback serving the same content behind <a href> links. Follow Google’s JavaScript SEO best practices throughout.

8. Content Quality on Product and Category Pages

This is where most stores lose, and it isn’t a technical problem.

Google’s helpful content self-assessment asks whether your content provides original information, reporting, research or analysis; whether it offers a substantial, complete or comprehensive description; whether it delivers insight “beyond the obvious”; whether it merely copies or rewrites other sources without adding substantial value; and whether it’s something a reader would want to bookmark, share or recommend.

Now apply that honestly to a product page running the manufacturer’s stock description, which is running on 400 other sites simultaneously. It fails almost every question.

What actually differentiates a product page:

• Original photography and video, not supplier assets
• Specifications you measured, not specs you copied
• Genuine answers to the pre-purchase questions your support team fields daily
• Sizing and fit guidance based on your actual return data
• Real customer reviews, marked up correctly
• Comparison against the alternatives a shopper is actually weighing
• Clear shipping, returns and warranty information on the page itself

Category pages need the same treatment. A category page that’s a bare product grid with an H1 is a thin page. Add a genuinely useful buying introduction, the two or three decision criteria that matter in that category, and internal links to the relevant sub-categories and buying guides.

On E-E-A-T — Experience, Expertise, Authoritativeness, Trustworthiness — Google states that trust is the most important of the four. For a store, trust is concrete: a real business address, a real phone number, transparent returns, visible policies, secure checkout, named people behind the brand, and no easily-verified factual errors anywhere on the site.

9. Technical Foundations You Can’t Skip

Core Web Vitals. Three metrics, measured at the 75th percentile of page loads and segmented by device:

Metric	Measures	“Good” threshold
Largest Contentful Paint (LCP)	Loading	2.5 seconds or less
Interaction to Next Paint (INP)	Interactivity	200 milliseconds or less
Cumulative Layout Shift (CLS)	Visual stability	0.1 or less

INP replaced First Input Delay as a Core Web Vital in 2024. If your monitoring still reports FID, it’s out of date.

Images. Ecommerce lives or dies on images, and they’re your heaviest asset. Serve modern formats, size them responsively, lazy-load below-the-fold images (never the LCP image), use descriptive filenames and alt text, and supply the multiple aspect ratios that merchant listings expect.

JavaScript. Google renders JavaScript, but rendering is a second pass with a queue. Anything critical — product name, price, availability, canonical, structured data, internal links — should be in the initial HTML response.

Crawler access. Check your robots.txt actually allows Googlebot to fetch your CSS and JS. Blocking them prevents Google from rendering your pages the way users see them.

Search Console. Verify the property. Without it you’re guessing, and you lose access to the reports covered in section 11.

10. Launching a New Ecommerce Site the Right Way

Google documents a specific registration checklist for new stores:

• Verify site ownership with Google
• Request indexing — the URL Inspection tool for a small number of URLs, sitemap submission for large numbers
• Monitor progress in the Page Indexing report
• Establish business details if you have physical locations
• Sign up for Google Merchant Center to enable Shopping tab visibility

Google also outlines four launch strategies, each with a real trade-off:

• Grand reveal — keep everything password-protected, then publish all at once. Nothing leaks, but you get zero pre-launch Google visibility.
• Home page launch — publish a placeholder homepage only. Lets you register the domain and start building backlinks while product data waits.
• Launch without product availability — publish the full site with products marked out-of-stock. Everything gets indexed, nothing can be bought. Use the excluded_destination attribute in your Merchant Center data.
• Soft launch — turn everything on the moment it’s production-ready, save the marketing announcement for later.

One warning from Google worth putting in bold: don’t disable add-to-cart functionality as a way of stopping users from purchasing products. Google may use that functionality for price verification, and breaking it causes problems downstream.

11. Getting Cited by AI: What Google Actually Says vs What the Industry Sells

This is the section most 2026 guides get wrong, so we’re going to be precise about sourcing.

What Google has officially published

Google’s documentation on AI features is unusually blunt: “There are no additional requirements to appear in AI Overviews or AI Mode, nor other special optimizations necessary.” And: “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.”

Google’s dedicated guide to optimizing for generative AI features goes further and names the tactics that don’t work. Under a section headed “LLMS.txt files and other ‘special’ markup,” it states: “You don’t need to create new machine readable files, AI text files, markup, or Markdown to appear in Google Search (including its generative AI capabilities), as Google Search itself doesn’t use them.” It says there’s no requirement to break content into tiny pieces for AI comprehension, no need to write in a specific style for generative AI systems, and no value in chasing every long-tail keyword variation with a separate page. It calls out pursuing inauthentic product “mentions” across the web as ineffective. And it states plainly that “Structured data isn’t required for generative AI search,” while noting that continuing to use it supports rich result eligibility.

The llms.txt question, answered properly

This gets misreported in both directions, so here is the full picture with dates.

Google’s Gary Illyes said at a Search Central Deep Dive event in July 2025 that Google doesn’t support llms.txt and isn’t planning to. Google’s documentation was then updated on 15 June 2026 to add a deliberately neutral position:

“It’s completely fine if you decide to create and maintain LLMS.txt files (or other similar files) for other services or systems that use these files. Doing so will neither harm nor help your site’s visibility or rankings in Google Search, as Google Search ignores them.”

So Google is not telling you to avoid llms.txt. It is telling you that Google Search ignores it entirely, in both directions.

The honest summary: llms.txt is not a Google ranking factor and won’t get you into AI Overviews or AI Mode. If you want to maintain one for other AI services that do read it, Google says that’s completely fine and it costs you nothing. Just don’t let anyone sell it to you as a Google SEO deliverable.

What Google says does work

Google’s AI optimization guide explains the mechanics: AI features use retrieval-augmented generation (RAG) and a query fan-out technique — issuing multiple related searches across subtopics and data sources to build a response. Both depend entirely on Google’s existing indexing and ranking systems. Which means the path into an AI answer runs through conventional search quality, not around it.

Google’s actual recommendations:

• Meet the technical requirements. Your pages must be indexed and eligible for standard Google Search snippets. Content must be crawlable — Google’s generative AI models use publicly accessible, crawlable content.
• Create non-commodity content. Google specifically advises “providing a unique point of view” through first-hand experience or expert insight, and warns against generic tips that go no further than common knowledge.
• Don’t multiply pages per query variation. Creating content primarily to manipulate rankings violates Google’s scaled content abuse policy and is “an ineffective long-term strategy.”
• Keep content primarily text-based, well-organised with clear headers, supplemented by quality images and video where relevant.
• Keep structured data aligned with visible content. Markup that claims things the page doesn’t show is a liability.
• For ecommerce specifically: keep Merchant Center feeds and your Business Profile current. Google names these as the levers that increase product visibility in AI responses.
• Semantic HTML helps — not because AI requires it, but because it aids accessibility tooling, which agentic systems increasingly rely on.
• Reduce latency and duplicate content, and maintain good page experience across devices.

Where the query fan-out actually helps you

Here’s our read as an agency, clearly labelled as interpretation rather than Google policy. Because AI Mode decomposes one shopper question into many sub-searches, the content that gets pulled in isn’t only your product page — it’s whatever page best answers each sub-question. “Best merino base layer for winter hiking” fans out into questions about weight in gsm, odour resistance, sizing, wash care, and price comparisons.

If you have a page that genuinely answers each of those, you have multiple chances to be the retrieved source. If your entire site is product pages with supplier copy, you have one weak chance. That’s the practical, non-gimmicky argument for building real buying guides, comparison pages, care guides and sizing resources — and it’s fully compatible with everything Google says above.

Beyond Google: ChatGPT, Perplexity and other assistants

Google’s rules aren’t the only ones now. Other assistants run their own crawlers, and blocking the wrong one makes you invisible in that ecosystem. Check your robots.txt against this table.

User agent	Operator	What it controls	Should a store allow it?
Googlebot	Google	Google Search, Images, News, Discover, and all Search features including AI Overviews and AI Mode	Yes — essential
Storebot-Google	Google	All Google Shopping surfaces including the Shopping tab	Yes
Google-Extended	Google	Whether your content trains future Gemini models and grounds Gemini Apps / Grounding with Google Search	Business decision — blocking it does not affect Google Search
GoogleOther	Google	Generic crawler for various Google product teams; affects no specific product	Generally yes
Google-InspectionTool	Google	Rich Results Test and URL Inspection in Search Console	Yes — you need your own testing to work
OAI-SearchBot	OpenAI	Surfacing your site in ChatGPT search results	Yes, if you want ChatGPT search visibility
GPTBot	OpenAI	Training OpenAI’s foundation models	Business decision
ChatGPT-User	OpenAI	User-initiated fetches inside ChatGPT; not automatic crawling and doesn’t determine search inclusion	Generally yes
OAI-AdsBot	OpenAI	Validates landing pages for ads in ChatGPT; not used for model training	Yes, if you advertise there

The distinction that trips people up: Google-Extended and GPTBot are about training. Googlebot and OAI-SearchBot are about being found and cited. Blocking training crawlers is a legitimate choice with no Google Search penalty. Blocking OAI-SearchBot removes you from ChatGPT’s search answers entirely — a very different decision.

Agentic commerce and UCP

Google’s own guidance now points to an emerging reality: AI agents may access your site through browser agents that analyse visual renderings such as screenshots, inspect the DOM structure, and interpret the accessibility tree. Practically, that means a checkout that only works with a mouse and perfect JavaScript execution is a checkout an agent can’t complete.

Google has also introduced the Universal Commerce Protocol (UCP), described as “an open standard designed for the future of commerce, empowering you to turn AI interactions into instant sales.” Adopting it enables agentic actions on AI Mode in Google Search and in Gemini, starting with direct buying, via either native checkout integrated with the AI surface or embedded iframe-based checkout. Onboarding runs through Merchant Center and there’s a waitlist plus an open-source repository.

Our take: UCP is early and not every store needs it yet. But if a meaningful share of your category’s discovery is already happening inside AI assistants, it’s worth putting on the 2026 roadmap rather than the 2028 one.

Measuring AI visibility

Search Console now includes a Generative AI performance report covering impressions from AI Overviews and AI Mode (Search Labs experiments are excluded). You can group by page (canonical URL), country, date and device.

Two caveats. The primary metric is impressions — how many times links to your site were shown in a generative AI feature — and the report is still rolling out, so not every property has access, and sites need sufficient AI-feature impressions to appear at all. Standard performance report limits apply, including the 1,000-row cap, and the newest data can be preliminary.

Google also warns explicitly against third-party tools claiming access to “internal” Google metrics: no third-party tool has access to Google’s internal ranking or AI systems. Treat “AI visibility scores” from vendors as directional estimates, not measurement.

12. Your Ecommerce SEO Checklist

Product data

• Merchant listing structured data on every product page with name, image, offers, price, priceCurrency
• Shipping details and return policy in markup
• ProductGroup + variant markup wherever you sell colours or sizes
• Merchant Center feed live, with auto-update from website enabled
• Structured data present in the initial HTML, not injected by client-side JS
• Validated with the Rich Results Test and monitored in Search Console

Site-wide markup

• Organization (or OnlineStore) with logo, contact, address and sameAs
• hasMerchantReturnPolicy at organization level
• hasMemberProgram if you run a loyalty scheme
• BreadcrumbList on category and product pages
• LocalBusiness on store location pages
• VideoObject on video-led pages

Architecture

• Every indexable product reachable via <a href> from a category or sub-category page
• XML sitemaps submitted and current
• Self-referencing canonicals on all indexable pages
• Descriptive, lowercase, stable URLs
• noindex on empty category pages
• Priority products linked from the homepage and content pages

Crawl efficiency

• Faceted navigation handled via robots.txt patterns or URL fragments
• Real 404s for empty and nonsensical filter combinations
• Sort-order parameters blocked from indexing
• No internal links to session IDs or tracking parameters

Pagination

• Sequential <a href> links between pages
• Unique URL and self-referencing canonical per page
• No fragment-based page numbers
• Sitemap or feed fallback if you use load more or infinite scroll

Content

• Original product descriptions, not supplier copy
• Category pages with genuine buying guidance
• Real customer reviews (and no self-serving review markup on your own business)
• Any incentivised reviews clearly and prominently disclosed on the review itself
• Buying guides, comparisons, sizing and care content that answer sub-questions
• Visible shipping, returns, warranty and contact information

Performance

• LCP ≤ 2.5s, INP ≤ 200ms, CLS ≤ 0.1 at the 75th percentile
• Modern image formats, responsive sizing, correct lazy-loading
• CSS and JS crawlable

AI and agents

• robots.txt audited against the crawler table above
• Merchant Center and Business Profile current and accurate
• Generative AI performance report monitored in Search Console
• Checkout usable without perfect mouse interaction
• UCP evaluated for your category
• No budget spent on llms.txt as a Google ranking tactic

Frequently Asked Questions

Do I need structured data to rank in Google Search?

No. Google states that structured data isn’t required to appear in Google Search results. But it helps Google understand your page and makes you eligible for product rich results, merchant listing experiences, and richer presentation in Search and Images. For a store, the eligibility alone is worth the implementation.

Should I create an llms.txt file for AI search?

Not as an SEO tactic for Google. Google’s documentation states plainly that Google Search ignores these files and that maintaining one “will neither harm nor help your site’s visibility or rankings in Google Search.” Its June 2026 update added that it’s completely fine to maintain them for other services or systems that do use them. So it’s a free, neutral choice — just don’t let it displace work that actually matters.

What’s the difference between a product snippet and a merchant listing?

Product snippets are for pages where a shopper can’t buy directly — editorial reviews and comparisons — and support richer review information including pros and cons. Merchant listings are for pages where a shopper can buy directly from you, and support shipping, returns and apparel sizing details. Meeting merchant listing requirements generally makes you eligible for product snippets too.

Can I put review stars on my own homepage for my own business?

No. If the entity being reviewed controls the reviews about itself, those pages are ineligible for the star review feature. That includes reviews served through embedded third-party widgets on your own site. Product reviews on your own product pages are a different case and are fine.

Should I use pagination, load more, or infinite scroll on category pages?

Pagination is the safest for SEO because Google’s crawlers follow <a href> links but don’t click buttons or trigger scroll-based JavaScript. If you use load more or infinite scroll for UX reasons, you must give Google another discovery path — sitemaps, a Merchant Center feed, or a crawlable paginated fallback.

Do I still need rel=”next” and rel=”prev” tags?

Google no longer uses them. Other search engines may, so they’re not harmful, but they will not affect your Google performance.

How do I stop faceted navigation from wasting crawl budget?

If you don’t need filtered URLs indexed, block the parameters in robots.txt or switch your filters to URL fragments, which Google ignores for crawling. If you do want some indexed, use standard & separators, keep filter order consistent, and return genuine 404s for empty or nonsensical combinations.

Should I block GPTBot and Google-Extended?

That’s a business decision about AI training, not an SEO one. Blocking Google-Extended has no effect on Google Search. Blocking GPTBot only affects OpenAI model training. But be careful not to block OAI-SearchBot, which controls whether you appear in ChatGPT search answers, or Googlebot, which controls everything in Google.

How do I know if I’m appearing in AI Overviews?

Use the Generative AI performance report in Google Search Console, which covers impressions from AI Overviews and AI Mode. It’s still rolling out, so not every property has access yet, and the metric available is impressions. Google warns that no third-party tool has access to its internal ranking or AI systems, so treat vendor “AI visibility” scores as estimates.

Do I need both structured data and a Merchant Center feed?

Google recommends both. Structured data drives product rich results in Search and helps verify your feed. A Merchant Center feed is mandatory for the Shopping tab, required for Lens product listings and Maps inventory, gives you control over update frequency, and lets you share data that isn’t published on your site.

One note on FAQ markup, since it comes up constantly: don’t bother adding FAQ structured data to a page like this expecting rich snippets. Google’s documentation states that as of 7 May 2026, FAQ rich results no longer appear in Google Search, with the related tooling and API support removed by August 2026. Eligibility had already been narrowed to well-known authoritative government and health sites before that. A clearly structured FAQ section written in plain HTML headings does the job on its own.

Where This Leaves You

Nothing in this guide is a hack. That’s the point. Google’s own 2026 guidance on AI features says it directly: there is no shortcut, no special file, no magic schema. What there is instead is a long list of unglamorous things that most stores have not done properly — variant markup, return policy markup, crawl-efficient faceted navigation, product pages with original content, category pages that actually help someone decide, and a Merchant Center feed that agrees with the website.

Do those well and you compete everywhere at once: classic results, Images, Lens, Shopping, AI Overviews and AI Mode all draw from the same well.

Need this implemented rather than just documented?

At Niche SEO Agency, ecommerce is where we spend most of our time technical SEO audits, ecommerce content SEO, Shopify SEO, website migration SEO and authority backlink campaigns for stores that need revenue, not vanity metrics.

If you’d like a straight answer on which of the items above are actually costing you money right now, get in touch and we’ll take a look.