---
title: AI crawls your product pages. It cites your blog.
description: GolOps research — 337,000 citations cross-referenced with 11.4M AI-crawler visits across 882 brands. Blog content earns 20.2% of citations on 14.4% of crawls; product pages do the reverse. A page-type efficiency matrix across 13 types.
date: 2026-03-03T00:00:00Z
lastmod: 2026-06-02T00:00:00Z
published: true
categories: [research, llm]
author: golops
---

AI crawlers comb your catalog obsessively. Then, when the model assembles a recommendation, it points somewhere else. Between what AI reads and what AI recommends sits a systematic gap, and it changes where companies should invest in content.

GolOps measured that gap. 337,000 unique URL-brand pairs from live AI responses, cross-referenced with 11.4M crawler visits across 882 brands. Every URL is mapped to one of 13 page types. We track a single metric: citation efficiency, the ratio of citation share to crawl share. A 1.0× score means a page type is cited exactly as often as it is read. Above 1.0×, the content outperforms its crawl traffic. Below it, AI reads the pages but does not recommend them.

| Metric | Value |
|---|---|
| Citations analyzed | 337,000 |
| Crawler visits tracked | 11,400,000 |
| Brands in sample | 882 |
| Page types classified | 13 (58% of URLs) |

## Key findings

**0.5× — product pages.** The catalog takes 13% of all crawler visits but only 6.3% of citations. AI reads product pages twice as often as it recommends them. The most expensive zone of lost choice in the sample.

**1.4× — blog and editorial content.** 14.4% of crawls against 20.2% of citations. One in five AI citations lands on a blog post — more than any other page type. It is also the stickiest format: once AI cites a post, it returns to it across an average of 8 different queries. An independent [Search Engine Land](https://searchengineland.com/how-to-get-cited-by-ai-seo-insights-from-8000-ai-citations-455284) analysis of 8,000 AI citations reaches the same conclusion: blogs and news dominate, while vendor product pages are cited in under 3% of cases.

**4.8× — review platforms.** The highest efficiency in the entire study. G2, Capterra, Trustpilot are almost never crawled, yet they are regularly cited as an authority signal.

**+18 points — sentiment on owned content.** When AI cites your domain, the average response sentiment is 76 out of 100 versus 58 for third-party sources. Yet only 1.8% of all citations point to your own domain.

## The efficiency matrix

All 13 classifiable page types, ranked by citation efficiency. Crawl share is how many crawler visits land on the type. Citation share is how many AI citations point to it.

| Page type | Crawl share | Citation share | Efficiency |
|---|---|---|---|
| Review / directory sites | 0.3% | 1.2% | 4.8× |
| About / contact | 0.4% | 1.7% | 4.6× |
| Resource / report | 0.4% | 1.4% | 3.8× |
| Service / use case | 1.4% | 2.6% | 1.9× |
| Blog / editorial | 14.4% | 20.2% | 1.4× |
| Documentation | 0.7% | 0.8% | 1.1× |
| Integration | 0.4% | 0.4% | 1.0× |
| Homepage | 10.7% | 7.9% | 0.7× |
| Legal / terms | 0.4% | 0.2% | 0.6× |
| Comparison | 0.4% | 0.2% | 0.6× |
| Product pages | 13.0% | 6.3% | 0.5× |
| Category / browse | 1.9% | 0.9% | 0.5× |
| Help / FAQ | 1.9% | 0.4% | 0.2× |

The five overperformers have something in common: they provide context, opinion, or proof of authority rather than product information. AI engines use them as evidence for a recommendation, not as the recommendation itself.

## Three types that tell the whole story

Three pages explain citation efficiency better than any chart. One converts crawler attention into recommendations at 1.4×. The second loses half its traffic. The third is nearly invisible.

**The workhorse — blog. 1.4×.** 20.2% of citations on 14.4% of crawls, an average of 8.0 queries per post. One in five AI citations lands on a blog post, and no other type comes close. Authoritative, data-rich editorial earns both the first citation and the repeat appearances: AI returns to it again and again.

**The laggard — product pages. 0.5×.** 6.3% of citations on 13.0% of crawls, 7.2 queries on average. Crawlers are obsessed with the catalog: 13% of all visits go there. But half as many citations come back. AI wants to recommend products; without context and use-case framing it has nothing to back them with.

**The ghost — help / FAQ. 0.2×.** 0.4% of citations on 1.9% of crawls, 6.1 queries. The lowest efficiency in the entire study. AI reads the FAQ for context but never surfaces it as an answer. For user experience and support-cost reduction, yes. For a place in the choice, no.

## Each crawler has its own reading style

"AI-optimized" is not one strategy but four. The distribution of visits across page types diverges sharply between the four crawlers; we broke down bot behavior in detail in [the anatomy of 600K crawler visits](/en/publications/ai-crawler-discovery).

| Crawler | Visits | Product pages | Other |
|---|---|---|---|
| OpenAI (GPTBot) | 461,000 | 83.9% | category 12.3% |
| Meta AI | 77,000 | 95.8% | other 2.0% |
| OpenAI Search | 21,000 | 77.1% | blog 4.2% |
| Anthropic (ClaudeBot) | 2,000 | 15.9% | homepage 9.8%, blog 7.2% |

GPTBot scans catalogs: 84% of its visits are product pages. Meta's crawler goes to the extreme at 96%. OpenAI Search is lighter but still product-focused at 77%. ClaudeBot is the outlier: the only one that reads broadly across types, from homepage and blog to categories. Optimizing for ChatGPT means product-page structure. Optimizing for Claude means content diversity. There is no single "AI-optimized" strategy, only one tuned to a specific engine.

## Owned content versus third-party sources

Only **1.8%** of all AI citations point to your own domain. The rest are third-party sources about your category. But that small slice matters.

When AI cites your content, the average response sentiment is **76 out of 100**, a **+18-point** lift over third-party citations (58 out of 100). Owned citations increase visibility and also change how AI describes you. A separate fact: Wikipedia alone captures **9.6%** of all third-party citations. For most brands the competitor in the citation field isn't another brand, it's the encyclopedia.

The stickiest citation is the homepage. Once AI starts citing it, it returns an average of 35 times. When AI does point to your domain, the distribution across types looks like this:

| Page type | Share of owned-domain citations |
|---|---|
| Blog / editorial | 19.3% |
| Product pages | 7.9% |
| Homepage | 6.6% |
| Service / use case | 3.7% |
| About / contact | 2.8% |

## Sleepers and dead zones

Beyond the three main types, the matrix holds outliers: pages with surprisingly high or low efficiency that most content strategies overlook.

**1.9× — service and use-case pages.** The quiet leader. Pages that describe what the product solves, not what it does, are cited nearly twice as often as they are crawled.

**4.8× — review and directory sites.** The highest efficiency in the study. G2, Capterra, and Trustpilot are almost never crawled, yet they are frequently cited as an authority signal.

**4.6× — about pages.** AI cites "who you are" far more than expected. The about page works for more than humans: to the model it is an important trust signal.

**3.8× — resource and report pages.** Whitepapers, templates, and research reports earn citations far beyond their crawl share.

**0.6× — comparison pages.** Surprisingly low despite high SEO value. AI appears to avoid brand-built comparisons, preferring third-party sources for impartiality.

**0.6× — legal pages.** Dead weight for citation. AI reads them but won't recommend them. Pure compliance, zero value for the choice.

## Methodology

What underpins the numbers:

- **337,362 unique URL-brand pairs** — citations extracted from the AI responses of several providers across 882 brands, collected through the GolOps citation-snapshot pipeline.
- **11.4M crawler visits** — tracked AI-system bot requests to brand sites over the observation window.
- **Domain-aware URL classifier** — path-pattern matching combined with domain override sets. 58% of URLs mapped to 13 page types; 42% remain in an "other" bucket.
- **Brand-normalization** — each brand contributes equally to the efficiency scores regardless of citation volume.

Two limitations, stated plainly. 42% of URLs resist classification: many corporate pages have generic path structures that defeat URL-based classification, so the true distribution across types may be higher. And this is a snapshot: citation patterns shift as models update, so the figures show a current trend, not a permanent ranking.

## Reading this at the decision level

You pour budget into product pages, AI crawls them diligently, and it cites the blog. The numbers are blunt: product runs at 0.5×, blog at 1.4×. The crawler reads your catalog; the recommendation gets assembled from someone else's. Crawling is not citing (per [Cloudflare](https://blog.cloudflare.com/ai-search-crawl-refer-ratio-on-radar/), Anthropic makes roughly 70,900 page requests per referral), and until you measure efficiency by page type the two processes are indistinguishable: you see bot traffic on the catalog and assume you are working on visibility, when half that attention never reaches the buyer's answer. That is the layer GolOps takes under management. We measure your position in the field of choice through the Choice Control Index, decompose it by page type and source, and translate the measurement into a prioritized content plan. The Strategic Pilot delivers the first cycle in 10–12 weeks; the Command Center keeps it running continuously across seven AI systems.

Count it in money. Gartner forecasts 90% of B2B procurement under autonomous AI agents by 2028, and Semrush already shows AI-channel conversion running 4.4× higher than organic search. That means a place on the AI shortlist will soon be worth four times a search click. A product page at 0.5× loses half of that place: for every 100 crawler visits that could convert if cited, you forfeit 50 spots in the choice — and at the 4.4× multiple, each uncited visit costs you four times more than the missed organic click you are used to paying for.

**Before you rewrite your product pages, understand what makes a page worth citing in the first place:**

[**The anatomy of an AI citation. What makes a page worth citing**](/en/publications/anatomy-of-ai-citation)

[Request an index diagnostic →](https://golops.io/en/position) · [Discuss a pilot →](https://golops.io/en/pilot)