By ONmetrics | Research-backed

---

Sarah Chen thought she was doing everything right. Her London, Ontario boutique — a curated home goods store with a thriving Shopify site — ranked on page one for a dozen high-value keywords. Richmond Row foot traffic was strong. Her Google Ads ROAS sat comfortably above 4x. Then March 2025 happened. Organic sessions dropped 34% in four weeks. Her product pages hadn't moved in traditional rankings. They simply... vanished from the new reality. A customer walking into her Masonville pop-up said it plainly: "I asked ChatGPT for the best Canadian-made ceramic vases and your store never came up." Sarah's product pages were invisible to AI search. Not penalized. Not outranked. Just never retrieved. Here's what nobody tells you: your product pages can rank beautifully in traditional search and still be completely absent from AI-generated answers. The game has changed beneath your feet, and most e-commerce store owners in London — and everywhere else — haven't noticed yet.

---

Why AI Search Changes Everything for Product Pages

Let's be direct. The old model was simple: rank on page one, get clicks. The new model? You need to be cited. There's a critical difference. A citation means an AI engine selects your product page as a source, references it in its generated answer, and — sometimes — links to you. Ranking is positional. Citation is selective. And the selector isn't a human scanning blue links. It's a language model deciding which sources best satisfy a user's intent.

The numbers tell a stark story. AI Overviews now appear for approximately 12.95% of US queries (Semrush 2025). That might sound modest. But consider this: the share of AI Overviews triggered by informational queries plummeted from 89% to 57% over the same period (Semrush 2025). Commercial queries — the kind where someone is actively looking to buy — now trigger AI Overviews at rapidly growing rates. Your product pages are no longer competing only with other product pages. They're competing with every piece of content an AI engine might cite instead of you.

Think about what that means for your London e-commerce store. A potential customer types "best winter boots for Ontario weather" into Google. The AI Overview generates an answer. Users click results only 8% of the time when an AI Overview is present, compared to 15% without one (Pew Research 2025). Nearly half your potential click traffic evaporates before anyone sees your carefully optimized meta description. But — and this is crucial — Google's own data indicates that AI Overview clicks are "higher quality" (Google Developers 2025). Fewer clicks. More valuable clicks. The funnel is narrower but deeper.

So what's the dollar impact? If your London, Ontario store currently receives 10,000 organic sessions per month with a 2.5% conversion rate and an average order value of $85, that's $21,250 in monthly organic revenue. A 34% traffic drop — like Sarah experienced — shrinks that to $14,025. That's $7,225 gone. Per month. From a shift you didn't even know was happening.

---

How AI Engines Actually Find Your Products

Understanding how AI search engines work isn't academic. It's tactical. Here's what nobody tells you: each AI platform retrieves and cites sources differently, and optimizing for one doesn't automatically optimize for all of them.

The 3-Step Retrieval Pipeline

Every major AI search engine — Google AI Mode, ChatGPT with Search, Perplexity, Claude — follows roughly the same three-step pipeline:

1. Query Understanding. The engine parses what the user actually means. "Best vegan leather jacket" isn't just about leather alternatives; the engine infers intent around ethics, durability, price, and style. Your product page needs to explicitly address these inferred sub-intents to be retrieved.

2. Retrieval. The engine searches its index for candidate sources. This is where most product pages fail. They're indexed but not retrievable for the specific constellation of concepts the query demands. A product page that says "Vegan Leather Moto Jacket — Black, Size S-XXL" is far less retrievable than one that says "Our vegan leather moto jacket is crafted from plant-based materials for customers who want the look of real leather without animal-derived materials."

3. Synthesis. The engine assembles an answer from multiple sources and decides which to cite. This is where authority, clarity, and structure matter enormously. Fragmented, thin, or redundant content gets absorbed without citation. Your product becomes background noise.

Platform-by-Platform Behavior

Here's where it gets real. Each platform has distinct citation patterns you need to understand:

• ChatGPT aggregates Reddit threads, review sites, and forum discussions heavily. If your product has no presence on Reddit or independent review sites, ChatGPT likely won't cite you — no matter how perfect your product page is. This is a hard pill to swallow for store owners who've invested thousands in on-page optimization while ignoring off-site conversations.

• Claude prefers authoritative, well-structured sources. Academic references, detailed guides, and meticulously formatted product documentation get cited more frequently. If you're selling artisanal goods from your Western University-adjacent workshop, Claude might actually favour your detailed craftsmanship page over a generic Amazon listing.

• Google AI Mode is the wildcard. Backlinko's 2025 research found that 50% of Google AI Mode citations come from beyond page one of traditional results. That means your product page sitting at position 14 can still be cited. This fundamentally changes the ROI calculation of your SEO efforts. Position 14 used to be worth approximately $0 in click revenue. Now it might generate citations that drive high-quality traffic.

---

Schema Markup: Your Product Page's AI Resume

Most product pages fail at this. Schema markup is the structured data vocabulary that tells machines exactly what your page contains. In the AI search era, schema isn't optional — it's your product page's resume. And right now, most resumes are either missing, incomplete, or lying.

Product Snippet vs. Merchant Listing Markup

Google recognizes two distinct product schema types. Product Snippet markup describes a single product on a page. Merchant Listing markup includes offer details like price, availability, and shipping. Here's the contrarian take: most e-commerce platforms default to Product Snippet markup, which is the weaker of the two for AI retrieval. Merchant Listing schema gives AI engines richer, more structured signals about what you sell, what it costs, and whether it's available. Shopify's built-in schema? It generates basic Product Snippet markup. You'll need a third-party app like JSON-LD for SEO or Schema Pro to upgrade to Merchant Listing markup. Both cost $15-30/month. Worth every penny when a single AI citation can drive a $200+ sale.

FAQ Schema

FAQ schema is criminally underused on product pages. When someone asks an AI engine "Is this jacket waterproof?", a product page with FAQ schema that explicitly contains that question and answer is dramatically more retrievable. Install an FAQ section on every product page. Write the questions your customers actually ask — not marketing fluff. "Can I machine wash this?" "Does this run true to size?" "Is this shipped from Canada?" Each FAQ pair is a retrieval hook for AI engines.

Review/Rating Schema

Review and aggregate rating schema signals social proof to AI engines. A product page with 47 reviews and a 4.6-star average, properly marked up, is more likely to be cited than an identical page without rating schema. WooCommerce stores: Yotpo and Loox generate review schema automatically, but verify the output with Google's Rich Results Test. We've seen Loox output malformed schema on variant-heavy products. Trustpilot's widget? It often renders reviews in JavaScript that AI crawlers can't parse, meaning your 200+ reviews might be invisible to AI engines even if they look great to human visitors.

Schema Must Match Visible Content

This is non-negotiable. If your schema says a product costs $79 but your page displays $89, Google will disregard the markup. If your schema claims 4.8 stars but your visible reviews average 4.2, you'll lose structured data privileges. AI engines cross-reference schema against page content. Mismatches don't just hurt your structured data — they erode trust signals that influence citation decisions.

---

Content That Gets Cited, Not Just Indexed

Here's the uncomfortable truth: most product page copy is written for conversion, not citation. "Premium quality." "Shop now." "Limited stock." None of this helps an AI engine decide you're the best answer to a user's question. You need content that's fragment-friendly, context-inclusive, and naturally language-optimized.

Fragment-Friendly Structure

AI engines don't read your page top to bottom. They extract fragments. Each paragraph, each list item, each FAQ answer is a potential citation fragment. Structure your content so any single section can stand alone as a meaningful answer. If your product description for a handmade wooden cutting board reads "Our artisan cutting boards are handcrafted in London, Ontario from locally sourced walnut and cherry wood, finished with food-safe mineral oil, and available in three sizes perfect for everyday meal prep or entertaining guests" — that single sentence is a citation-ready fragment. It answers what, where, from what, and for what purpose. Contrast that with "Beautiful handmade cutting board. Shop the collection." The second version says nothing an AI engine can use.

Context-Inclusive Copy

Never assume the AI engine knows your brand. Embed your brand name, location, and differentiators directly into product descriptions. "The Conestoga Woodshop walnut cutting board, handcrafted in London, Ontario" is infinitely more citable than "our walnut cutting board." AI engines synthesize answers from multiple sources simultaneously. If your brand name and context aren't in the copy, they won't appear in the synthesis.

Natural Language Optimization

Keyword stuffing is dead. Has been for years. But a new form of optimization has emerged: natural language coverage. This means writing product copy in the full sentences and questions your customers actually use. Research from Aggarwal et al. (KDD 2024) demonstrated that GEO — Generative Engine Optimization — boosts visibility by approximately 40% compared to traditional SEO alone. The key technique? Quotation-friendly phrasing. Write sentences that AI engines can quote verbatim as answers. "This sleeping bag is rated to -15°C and tested in Canadian winter conditions" is a quotable answer. "Cold weather sleeping bag -15°C" is not.

Dedicated Pages for Each Use Case

If you sell a cast iron skillet that's used for camping, stovetop cooking, and baking, create three dedicated landing pages — one for each use case. Why? Because AI engines retrieve pages based on specific intent. A query like "best cast iron skillet for campfire cooking" will retrieve a page specifically about campfire cooking before it retrieves a generic skillet product page. Yes, this means more pages. Yes, this means more work. But each page is a separate retrieval opportunity, and in the AI search era, retrieval opportunity equals revenue opportunity.

---

Technical Accessibility for AI Crawlers

Your product pages can be perfectly written and beautifully structured, but if AI crawlers can't access them, none of it matters. And most e-commerce stores are accidentally blocking the very bots they need.

Allow AI Bots in robots.txt

Check your robots.txt file right now. Many platforms — especially Shopify and WooCommerce with security plugins — block unknown crawlers by default. This means ChatGPT-User, ClaudeBot, PerplexityBot, Bytespider, and Google-Extended may all be locked out. Add explicit allow directives for each. Here's what the entry should look like:

``` User-agent: ChatGPT-User Allow: /

User-agent: ClaudeBot Allow: /

User-agent: PerplexityBot Allow: /

User-agent: Google-Extended Allow: / ```

Yes, this means these bots will consume your content. That's the point. You want them to. If you're worried about server load, set Crawl-delay directives instead of blocking entirely.

Whitelist in Your WAF

Cloudflare, Sucuri, and other WAF providers often flag AI crawlers as suspicious traffic. Check your WAF logs. If you see ClaudeBot or PerplexityBot being challenged or blocked, whitelist them. A WAF that blocks AI crawlers is like a bouncer at your Richmond Row storefront turning away every customer who heard about you from a friend. The referral source is different, but the value is real.

Semantic HTML

Use proper heading hierarchy. H1 for product name. H2 for key features. H3 for specifications and FAQ items. Tables for comparison data. Ordered lists for step-by-step usage instructions. AI crawlers parse semantic HTML far more effectively than div-soup markup. If your theme uses