Most stores pour their energy into the homepage — the hero image, the headline, the polish. But AI search does not look there. On June 15, 2026 we asked 30 real questions a 6-to-8-figure ecommerce operator would type to four AI engines, then took every single URL those engines cited — 871 of them — and analyzed what kind of page AI actually pointed to. The answer is unambiguous and a little uncomfortable: AI search cites deep, specific content pages, almost never homepages. Every number below comes from that run.
Methodology
- Run date
- June 15, 2026 (a single snapshot)
- Queries
- 30 real questions a 6-to-8-figure ecommerce operator would type
- AI engines
- Four engines in live-web-search mode, producing 120 query-engine answers
- What we analyzed
- Every cited URL — 871 in total — by folder depth and by page-type keywords in the URL path
- Depth
- How many folders deep the cited URL sits below the domain root
- Page type
- Inferred from keywords in the URL path (blog, product, "vs", guide, etc.)
Honest limitations. This is a single snapshot from one day, not a trend yet — treat it as a baseline to re-measure against. Page type is inferred from URL-path keywords, a heuristic: a single URL can match more than one signal (a "/blog/best-vs-..." URL counts as both blog and comparison), so the page-type figures below are signals, not a clean partition, and they do not sum to 100 percent. The full set is 871 cited URLs across 120 query-engine answers.
Only 1% of AI Citations Were Homepages
Start with the headline number, because it reframes everything else. Of the 871 cited URLs, only 8 were homepages — the bare root of a domain. That is 1 percent. The other 863, or 99 percent, were deep internal pages. When an AI engine answers an operator's question, it almost never points at the front door of a site. It points at a specific page inside.
This is the opposite of how most stores invest. The homepage gets the design budget and the agency hours; the deep content pages get neglected. But the homepage is a brochure — a general statement of who you are. AI search is looking for a specific answer to a specific question, and that answer lives on an internal page, not the front of the store.
AI search does not cite homepages. 99 percent of the 871 cited URLs were deep internal pages — specific content, not the front door of a site.
How Deep AI Goes
It is not just "internal" — it is specific. Breaking the 871 cited URLs down by how many folders deep they sit below the domain root, the single most common shape was a page two levels deep, which accounted for 70 percent of all citations. Homepages were 1 percent, one-level-deep pages were 16 percent, and the deepest pages — three or more levels down — were 12 percent. The center of gravity sits exactly where specific articles, comparisons, and guides tend to live: a couple of folders inside the site, not at the root.
| URL depth | What lives here | Share of cited URLs |
|---|---|---|
| Homepage (root) | The front door of the site | 1% |
| 1 level deep | A top section or hub page | 16% |
| 2 levels deep | Specific articles, comparisons, guides | 70% |
| 3+ levels deep | Deeply nested specific pages | 12% |
Read together, depth and the homepage finding tell one story: AI search rewards the page that directly answers the question, and that page is rarely near the surface of a site. It is buried where real, specific content lives.
What Kind of Page Gets Cited
Knowing the cited pages are deep is one thing; knowing what they are is the actionable part. We inferred page type from keywords in each URL's path — a rough but useful signal. Because a single URL can match more than one keyword, these are signals, not a clean split, and they do not add up to 100 percent. With that caveat, the most-cited kind of page was clear: 37 percent of cited URLs looked like blog, article, or post pages. Product and shop URLs were 18 percent, comparison or "best-of" or "versus" URLs were 16 percent, and guide, how-to, or tutorial URLs were 8 percent.
| Page-type signal | Share of cited URLs | Role |
|---|---|---|
| Blog / article / post URLs | 37% | Content |
| Product / shop / store URLs | 18% | Product |
| Comparison / "vs" / "best-" / alternative URLs | 16% | Content |
| Guide / how-to / tutorial URLs | 8% | Content |
The pattern is consistent with the depth finding. Blog, comparison, and guide pages — all of them deep, specific content — together dominate the cited set, while product pages take a meaningful but smaller slice. The homepage does not even register as a page type, because it is almost never cited at all.
What This Means for Your Store
Everything below is grounded only in the run above — no assumptions, just what the numbers imply.
- Stop optimizing the homepage for AI search. Only 1 percent of citations were homepages. The design budget you pour into the front door earns essentially nothing in AI answers. Polishing the homepage is not how you get cited.
- Build deep, specific pages. 99 percent of citations were internal pages, and 70 percent sat two folders deep — exactly where specific articles, comparisons, and guides live. The page that directly answers a question is the page that gets cited.
- Lead with articles, comparisons, and guides. Blog/article URLs were the single most-cited type at 37 percent, with comparison URLs at 16 percent and guide URLs at 8 percent. These content shapes are over-represented among citations — make them.
- A library beats a brochure. A brochure site is one polished homepage. A library is dozens of specific pages, each answering one real question. The data says the library is what AI search cites — breadth and depth of specific content, not surface polish.
- Re-measure before you conclude a trend. This is one snapshot, and page type is a URL-path heuristic. Treat these figures as a baseline and run the same analysis again later to see what moved.
For the step-by-step version of "become a citable source," read getting your store cited by AI search, and to see which domains rose to the top of the same run, read the AI Citation Index.