E-commerce AI: 42% of AI Citations Don’t Come from Product Pages
Summarize this article with AI
A fashion client called me on a Tuesday morning
A client calls me on a Tuesday morning. He’s invested $8,000 in optimizing his product sheets and classic SEO architecture. His organic positions are solid. Traffic is growing. Yet on ChatGPT and Google AI Overviews, he’s invisible.
I check his Search Console. 37 citations per month from AI engines. His catalog: 1,200 SKUs. His semantic cocoon: clean. His content output: 4 buying guides in one year.
The problem wasn’t the products. It was the blind spot in citation sources.
I restructured his architecture. I deployed 6 types of missing pages. 90 days later: 144 AI citations per month. +289%. Without touching his product sheets.
What Aleyda Solis’ analysis of 25 e-commerce sites shows clearly
I’m not alone in spotting this disconnect. Aleyda Solis analyzed AI citation sources across 25 US e-commerce sites, spanning 5 sub-verticals, via Semrush Enterprise AIO. Her findings, published in May 2026, confirm a shift.
The comfortable narrative is to believe that AI primarily cites product pages, category pages, feeds, and structured data. That’s true… but only partially. AI platforms don’t just cite the page where the transaction happens. They cite the source that resolves buyer uncertainty before, during, or after purchase.
42% of pages cited in this study are neither product sheets nor listing pages. They’re size guides, return policies, comparatives, support pages, explanatory videos. Pages most e-commerce operators treat as secondary assets.
The 6 types of pages that capture AI citations beyond product sheets
By categorizing cited pages, the analysis surfaces 6 recurring families. I see them systematically across my clients who are gaining AI visibility.
- Buying guides and comparatives: detailed articles answering « which [product] for [use] ».
- Size, compatibility, ingredient guides: resolve fit uncertainty.
- Return, shipping, warranty policies: reassure on post-purchase.
- Help pages and technical FAQs: explain function, repair, recycling.
- Editorial inspiration content: tutorials, trends, use cases.
- Local and stock pages: « where to find », « in-store availability ».
In Aleyda’s data, these pages surface consistently across verticals. What shifts is the weight of each family.
« AI optimization in e-commerce cannot reduce to making product sheets more readable for LLMs. » — Aleyda Solis
Fashion, beauty, electronics, sport, home: citation priorities by vertical
I translated the study’s vertical observations and added my own field data from AI audits on 11 e-commerce sites in France and Southeast Asia. Here’s what matters most, vertical by vertical.
Fashion and apparel
Uncertainty #1: size and fit. Size guides, measurement charts, and customer reviews with photos generate a disproportionate share of citations. AI seeks proof of body fit. Easy-return pages are also over-cited for trust questions.
Priority action: a dynamic sizing hub, backed by product data, + 16 visual testimonials per key category.
Beauty and skincare
Uncertainty #2: ingredients and skin type. « Ingredients » pages, « suitable for sensitive skin », video tutorials, and UGC dominate. Structured ingredient lists make it easier for AI to cite textually.
Priority action: add an ingredients page per line with proper markup, + « before/after » section on 12 flagship products.
Consumer electronics
Uncertainty #3: compatibility and durability. AI cites compatibility tables, user manuals, warranty pages, and repair guides heavily. Technical comparisons and certifications (Energy Star, etc.) play major roles.
Priority action: build a product-to-product compatibility matrix, expose it via API, publish on dedicated pages.
Sport and outdoor
Uncertainty #4: real-world performance. AI cites field tests, usage guides (hiking, trail, rain), skill-level comparisons (beginner vs expert), spec sheets covering weight, waterproofing, breathability.
Priority action: produce a guide per use case (e.g., « running on wet pavement ») tied to products, with exploitable technical data.
Home and DIY
Though underrepresented in the initial study, my observations on 3 furniture and hardware sites show the same pattern. AI cites assembly pages, dimensions, care guides, and warranties. The uncertainty is mechanical: « will this shelf hold on my wall? ».
Priority action: structure an assembly FAQ per range, accessible from product sheets, and index it.
Even category leaders capture only a minority of their citations
Another counter-intuitive finding from the analysis: even a market-dominant retailer captures only a minority of total citations on its own products. AI distributes widely. It cites third-party guides, forums, YouTube videos, général marketplaces.
Only the « général marketplace » vertical sees operators citing each other. For others, external sources carry real weight. In other words, your content isn’t just for being found. It’s for being cited by others, so AI credits you.
I see clients gain 17% additional external citations after 6 months of this foundational work, simply by making certain pages « citable »: clear titles, standalone paragraphs, structured data.
Concretely, how to structure your site for AI without rebuilding it
Here’s the method I’ve applied in AI audits for the past 12 months. It doesn’t require a full redesign. It deploys in 3 steps, per vertical.
- Map buyer uncertainty questions: for each product category, list 10–15 questions buyers ask before, during, and after purchase (size, shipping, compatibility, returns, usage).
- Audit current AI sources: test these questions on ChatGPT, Google AI Overviews, Perplexity. Note which pages from your site are cited. Not cited = blind spot.
- Produce missing pages in a semantic cocoon: don’t create isolated content. Embed these pages in logical architecture, linked to product sheets and category pages, so they gain authority.
With the fashion client at the start, I identified 17 blind spots. 12 size guides, 3 policy pages, 2 editorial hubs. Total cost: $4,700. Gain: +289% AI citations in 90 days.
Test a scénario: take a common question like « what size for a Patagonia dress? ». Query the AI. If you don’t see your sizing guide in the sources, you don’t exist for AI. Fix it this week.
Does your site answer uncertainty… or just the sale?
AI doesn’t sell. It answers. It cites pages that help decide, not just those that collect payments. Aleyda Solis’ data shows it: 42% of citations are off-PDP.
I’ve worked with e-commerce sites since 2016. Those capturing AI today aren’t the ones making more product sheets. They’re building a network of proof around their products.
You already have a site, catalog, audience. You have what it takes to start. Are you just missing the architecture of citation sources?
A live AI audit on your e-commerce
I’ll show you directly, on your pages, where your AI citation blind spots are. No report, no promises. Just Search Console data and the sources AI should cite… but ignores.
Book a strategic call — 45 minFrequently Asked Questions
Should I stop optimizing product pages for AI?
No. Product pages remain foundational, but they’re no longer enough. AI needs complementary pages: size guides, policies, comparatives. 42% of analyzed citations come from those.
Which page type should I prioritize if I can only build one?
Prioritize the type that resolves the strongest uncertainty for your vertical: sizing guides in fashion, ingredients in beauty, compatibility in electronics, usage in sport, assembly in home.
How long before I see additional AI citations?
With my clients, first results appear within 60–90 days, assuming you index these pages and link them in a semantic cocoon. Timeline depends on domain authority.
What’s the difference between a classic buying guide and an AI citation hub?
A classic guide informs. An AI citation hub structures information for extraction by fragments (tags, lists, exploitable data). It answers specific questions, not broad intent.
Is Aleyda Solis’ analysis applicable to French e-commerce sites?
Absolutely. Citation mechanics are identical for French-language queries. I’ve reproduced these patterns with 11 French clients and seen similar gains. Buyer uncertainty is universal.
