Five URLs are answering your buyers' AI questions

We counted every source cited in 184 AI responses for one local services category. Five directory pages and one three-year-old Facebook post were doing most of the answering. The brand we audited was on none of them.

The audit and what we counted

We ran a full-panel AI visibility audit for a brand in a locally-competitive services category. The panel covered 46 prompts, each asked of four AI engines that use live web search: ChatGPT (OpenAI Responses API), Gemini, Perplexity, and Claude. That produced 184 AI responses total.

We extracted every URL cited in those responses and counted how many times each unique URL appeared across the full panel.

What the citation data showed

Source	Citation type	Times cited
Aggregator site A	ISP directory listing	48
Aggregator site B	ISP reviews by city	34
Provider A local page	Competitor coverage page	32
Aggregator site C	ISP availability checker	29
Aggregator site D	ISP comparison by city	25
Aggregator site E	Local internet options guide	23
Facebook community post	Local group discussion, 3 years old	23

The five aggregator pages in the table account for the majority of citations in competitor-winning responses. The brand being audited was absent from all five of them.

A Facebook post tied with a major aggregator

The Facebook entry in that table is worth pausing on. A post in a local community group from three years ago was cited by AI engines as a source 23 times across 184 responses. That is the same number of citations as a national ISP comparison site.

This is not a quirk. AI engines with web search access do not filter sources by publication prestige or domain authority. They find pages where the specific question gets answered, and they cite what they find. A community discussion where locals share ISP recommendations is, from the AI engine's perspective, a relevant source for "which ISP is good in this area."

What this means for your strategy

Most brands thinking about AI visibility focus on their own website. Better content, structured data, llms.txt, more authoritative pages. These are real improvements. But in categories where aggregators and directories are the primary citation sources, fixing your own site is secondary to getting on those aggregators.

When a handful of directory pages are answering most of the buyer-intent queries in your category, and your brand is not on them, your website copy does not decide those answers. You are simply not in the running.

The brand in this audit had a functioning website with some local press coverage. The dominant competitor's advantage was not a better website. It was structured listings on the aggregator infrastructure that AI engines had absorbed and treat as authoritative for the category.

Citation concentration by category

The degree of concentration varies by category. In locally-competitive service categories (ISPs, home services, legal, medical), aggregator concentration is very high. A handful of directory sites get cited for a large majority of buyer-intent queries because they have systematically indexed local providers.

In B2B software, the citation structure is different. Review platforms, comparison sites, and category pages on G2, Capterra, and software-specific directories play the aggregator role. A brand absent from those platforms is invisible in the same way.

The principle is the same across categories: identify which 5-10 URLs are being cited for buyer-intent queries in your space, verify whether you appear on them, and prioritize getting listed or covered there before investing in other visibility tactics.

Implications for content strategy

A common advice pattern in content marketing is to publish more. More blog posts, more landing pages, more long-form content. This can improve organic search ranking. Its effect on AI citation is more limited if the citation infrastructure for your category is already concentrated in a few directories.

For this category, the engines were not pulling from blog content at all. They were pulling from directory and review pages. More posts on the brand's own site would not have put it into the answers, because the answers were not being built from site content in the first place.

Understanding which sources AI engines actually cite for your query types tells you where to put the work. That is different from assuming visibility follows the same logic as organic search.

How we identify citation sources

Every AI response in our audits is parsed for cited URLs. We extract the domain and the specific page, count appearances across the full panel, and segment by whether the brand was mentioned or absent in that response.

The output is a table of sources with two key columns: citations in responses where the competitor won, and citations in responses where the target brand appeared. The gap between those two columns is the citation source gap, and each row is an action item.