Measured, 27 July 2026

The 200 pages AI cites when you ask which AEO tool to buy

We run a tool that logs every source behind an AI answer. Pointed at our own category, it recorded 1,117 citations across the 200 most-cited pages. The finding is blunt: the entry price to being named by ChatGPT is a page, not an audience. Half the winning domains are software vendors citing their own comparison posts. One of the cited sources is a Reddit thread with six upvotes. And our own name appears in none of the 200.

How this was measured

Buyer questions in the AI-visibility category, run across ChatGPT, Gemini and Perplexity, with every cited URL stored per answer. We then counted citations per page. 200 pages is what our query returns, ranked by citations, so treat it as the top of the list rather than the whole index. The lowest page in the sample was cited 3 times.

The domains

Reddit leads. That much matches what everyone says about AI search. The surprise is the rest of the column: six of the top ten are software vendors, and four of those are companies you have probably never heard of. They are not there on reputation. They are there because they published the page.

The pages that do the work

These are individual blog posts on small vendor sites, each cited more often than most press coverage in this category. Read the titles together and the pattern writes itself.

Every one of them is a list. Not a manifesto, not a feature page: a list of competing tools with prices, published by one of the competitors. The engines want a page that surveys the field, and the only people willing to write that page are the vendors themselves. So the vendors get cited.

The threads

Reddit’s 51 citations are not one viral post. They are a handful of ordinary buying threads, most of them small.

The last row is the one to sit with. Six upvotes, written by someone promoting their own product at number two on the list, and the engines quote it when a buyer asks what to use. Whatever you imagined the bar was, it is lower.

What this means if you are invisible

1. 1. Write the survey page, not the sales page. A page that names ten competitors with real prices gets cited. A page that names only you does not. This is uncomfortable and it is what the data says.
2. 2. Match the question, word for word. The winning URLs read like the questions buyers type: best X alternatives, best tools to monitor Y, X vs Y vs Z. Clever titles lose to literal ones.
3. 3. Go where the argument already happens. Find the threads the engines already cite in your category and be a useful voice in them. One answer in a thread that is already a source beats ten pages nobody retrieves.
4. 4. Count citations, not impressions. None of the pages above would look impressive in an analytics dashboard. They are winning the only metric that decides whether a model repeats your name.

Our own number

Fulcru appears in zero of the 200 pages. Not once, in 1,117 citations, in the category we sell into. We had already published the zero; this is the reason behind it. We wrote pages about ourselves while the companies beating us wrote pages about everyone.

So we changed what we publish, starting with this page and with a comparison that lists every tool above ours where the price says it belongs. We will re-run the same measurement and post the new number, whichever way it goes.