LLMs Citation Deep Content, Not Homepages

The numbers are clear. Only 11.6% of citations come from URLs with a folder depth of one. With each additional level, the percentage grows – 27.5% at two folders, 33.3% at three. Pages with depths of four to ten folders account for another 16.2% of citations combined.

Why AI Models Prefer Specific Content

ChatGPT and similar tools don’t look for general information on main pages. They need precise answers to specific questions. When someone queries “UV protection in sportswear”, the model will favour a detailed blog article or product page over generic text about athletic clothing.

The study results show that successful brands create content precisely for this type of search.

Detailed FAQ sections, technical product specifications, data studies and long-tail content give AI models material they can cite.

Source: Similarweb

Structure Determines Visibility

Walmart and Temu provide a good example of how formatting affects results. Walmart uses a separated “About this item” section on product pages with bullet points and clear characteristics. Temu puts most information into the product title.

The difference in results is significant. Walmart appears in 59% of relevant AI responses. Temu is practically invisible. The reason is simple – LLM models process structured content with clear section labels better.

Source: Similarweb

What This Means For Your E-shop

Create specialised content that answers your customers’ specific questions. Instead of a general “How to choose running shoes” page, make five separate articles for beginners, marathoners, winter running, people with joint problems, and ultralight footwear.

Use clear structures – H2 and H3 headings, bullet points for features, and separate sections for specifications. AI models need to know where to find specific information.

Analyse what questions customers are asking. Long-tail searches are your opportunity – competition is lowest here and citation chances are highest.