A mid-market e-commerce brand can rank on page one of Google for its category and still never get named when someone asks ChatGPT or Gemini for a recommendation in that same category. This is not rare. By one widely cited estimate, roughly 78% of e-commerce brands are absent from AI recommendations in their own category, despite having functioning websites and reasonable search rankings. The gap is not a content problem. It is usually a crawl, schema, or entity problem, stacked in that order.
Key Takeaways
- Ranking in Google search and being cited in an AI answer are two separate systems with different inputs, and a brand can win one while losing the other entirely.
- Absence from AI recommendations usually traces to one of three layers: the model cannot crawl the content, the content has no structured data to parse, or the model cannot resolve the brand as a distinct entity.
- The three layers should be diagnosed in order, crawl before schema before entity, because a fix at a later layer cannot compensate for a failure at an earlier one.
Why Ranking in Search Does Not Mean Being Recommended by AI
Search ranking rewards relevance and link signals accumulated over years. An AI answer rewards something narrower: the model’s confidence that a specific source can be cited as the fact behind a claim. Those are not the same test, and a brand can pass one while failing the other.
This is the distinction between Share of Voice and Share of Citation. Share of Voice is how often a brand gets mentioned across the web generally, blog posts, forum threads, press coverage. Share of Citation is how often an AI model actually pulls that brand in and names it as a source for an answer. A brand with a decade of content marketing and a strong domain can have high Share of Voice and near-zero Share of Citation, because the model never found a version of the brand’s facts it trusted enough to cite.
The five major models do not evaluate sources the same way. ChatGPT pulls from across the open web, Gemini cross-references Search, YouTube and Scholar, Claude leans on high-authority publications and documentation, Perplexity cites its sources inline, and Grok reads the real-time social web. A brand invisible to one of these may be well represented in another, which is why a single-engine check gives a false read on the problem.
Can the Model Even Crawl You?
The first layer to check is the most basic one: can the model’s retrieval system reach the page at all. Product pages built on heavy client-side rendering, pages gated behind interaction, or pages blocked by an overly broad robots.txt rule are invisible to a model doing live retrieval, regardless of how well the copy is written.
This matters more for e-commerce than most verticals, because category and product pages are often generated dynamically, paginated aggressively, or hidden behind faceted navigation that produces thousands of near-duplicate URLs. A model trying to find the canonical page for “best running shoes for flat feet” may simply never encounter the page that answers it, not because the content is weak, but because the crawl path never resolves to a clean, citable URL.
This is retrieval presence, one of the two layers of AEO. The other, training-data presence, is earned slowly through wide coverage over time and is harder to influence directly. Crawl access is the one lever a brand can pull immediately, through log file review, rendering audits, and checking what a model-facing crawler actually receives versus what a browser renders.
growth in AI brand presence for NewDay USA, tracked weekly across five engines over six months.See the case study
Does Your Schema Turn Content Into Facts a Model Can Cite?
Once a page is reachable, the next question is whether the model can parse it into discrete, citable facts. Prose that reads well to a human buyer can still be structurally illegible to a model looking for a price, a return policy, a material specification, or a certification.
Structured data, Product, Offer, Review, FAQPage, and Organization schema, is what turns a paragraph into a fact a model can extract with confidence. An e-commerce page that states its return window in a sentence buried in the fourth paragraph gives a model far less to work with than the same fact marked up explicitly. Models favor content they can lift cleanly, because clean extraction lowers the risk of misrepresenting the source.
This is also where digital PR and citation building intersect with schema. A fact stated on the brand’s own site carries less independent weight than the same fact corroborated by a third party, a review site, a trade publication, a comparison article. Models increasingly weigh agreement across independent sources, not a single assertion from the brand itself.
Does the Model Know What You Actually Are?
The third layer is entity clarity, and it is the one most e-commerce brands skip entirely. An entity is how a model resolves a name to a distinct, known thing with attributes, a category, a founding date, a relationship to competitors, a set of products. If a brand name is ambiguous, shared with another business, or thinly documented across independent sources, the model cannot confidently attach facts to it, even when those facts are crawlable and well-structured.
Trust, in this context, is not a score like domain authority. It is a graph of agreement across independent sources describing the same entity consistently. A brand mentioned inconsistently, different category descriptions on different sites, conflicting founding details, no clear relationship to its own subsidiary brands, gives the model nothing stable to anchor a citation to.
Citations compound: once a model cites a brand, the next citation becomes more likely, which means the brands absent today tend to stay absent unless something structural changes.
This compounding effect is why the 78% figure is so sticky. Brands already being cited keep getting cited, because each citation reinforces the model’s confidence in that source. Brands outside the loop stay outside it, because there is no mechanism pulling them in without a deliberate fix to entity clarity.
A Diagnostic Sequence: Crawl, Schema, Entity, in That Order
Running all three checks at once produces noise. The sequence matters because each layer gates the next, fixing schema on a page the model cannot crawl changes nothing, and clarifying entity data has no effect if the underlying pages are never retrieved.
- Check crawl access first: pull rendered HTML as a model-facing crawler would see it, compare against what a browser shows, and confirm category and product pages resolve to stable, indexable URLs rather than infinite faceted variations.
- Check schema second: audit whether price, availability, reviews, and policy information exist in structured markup, not just prose, and confirm the markup validates without errors across the core page templates.
- Check entity clarity third: search the brand name across independent sources and look for inconsistency in category description, ownership, or product scope, since this is what a model uses to decide how confidently to cite the brand at all.
A brand that fails at crawl access will show flat zero presence across every engine. A brand that passes crawl but fails schema will show sporadic, low-confidence mentions without specific facts attached. A brand that passes both but fails entity clarity will show presence in training-data-heavy engines like Claude but near-absence in retrieval-heavy, real-time answers. The pattern of failure tells you which layer to fix first.
What to Do Next
Start with a crawl audit this week, not a content audit. Pull server logs for known AI crawler user agents, compare rendered output to raw HTML, and flag any product or category template that depends on client-side JavaScript to display core facts. That single check will tell you whether the problem is even reachable by the fixes that come after it.
Once crawl is confirmed clean, run schema validation against the core page templates, not a sample of ten pages, since e-commerce catalogs typically have variation across subcategories that a small sample will miss. Then check entity consistency across five or six independent third-party sources describing the brand, and note where the description, category, or ownership details diverge. Fix in that order: crawl, then schema, then entity. Fixing out of sequence wastes the work.
FAQs
Why would a brand rank well on Google but still be absent from AI recommendations?
Search ranking and AI citation rely on different signals. Google ranking rewards relevance and accumulated link equity, while AI citation rewards whether a model can confidently extract and attribute a fact to a specific source. A brand can satisfy the first system without satisfying the second.
What is the difference between Share of Voice and Share of Citation?
Share of Voice measures how often a brand is mentioned across the web overall. Share of Citation measures how often an AI model actually names that brand as the source behind an answer. A brand can have strong Share of Voice and near-zero Share of Citation.
Which layer should an e-commerce brand check first, crawl, schema, or entity?
Crawl access first, because it gates everything else. A page a model cannot reach will not benefit from improved schema or clearer entity data, so fixing those layers before confirming crawl access wastes effort.
Why does entity clarity matter if the content and schema are already correct?
A model needs to resolve a brand name to a distinct, well-documented entity before it will cite that brand confidently. Inconsistent descriptions of the brand across independent sources make the model less certain which facts belong to which business, even when individual pages are well structured.
