Who actually gets cited by ChatGPT, and why

From watching which pages ChatGPT cites when it answers B2B SaaS queries with its web_search_preview tool, the patterns are less random than we expected and more uncomfortable for anyone who's been optimising for Google for the last decade.

#The headline

A small set of domains captures the large majority of citations. The long tail isn't quiet, it's silent. If you're not among that small set for a given query, your odds of being cited at all drop sharply.

That sounds like Google rankings, but the leaderboard is not the Google leaderboard. Three patterns:

#Pattern 1: independent reviews beat vendor pages 4 to 1

For comparison queries ("best CRM for 50 person SaaS"), independent review sites (G2, Capterra, but also smaller niche outlets like SaaSworthy, GetApp) account for the bulk of citations, with vendor pages taking a much smaller share. The rest is a mix of YC company pages, ProductHunt, and a surprising amount of Reddit.

If your product page is your only citation surface, you're losing to the comparison site that wrote one paragraph about you.

#Pattern 2: schema markup correlates strongly with citation count

Cited URLs very often carry @type: Article, Review, or Product structured data with at minimum a name, author (or brand), and dateModified. Pages that rank in Google's top results for the same query but aren't cited by ChatGPT far less often have similar markup.

We're not saying schema causes citation. We are saying that the citation set looks very different from the Google ranking set, and schema is one of the variables that explains the difference.

#Pattern 3: "atomic answer" pages dominate "comprehensive guide" pages

Cited URLs are usually focused pages (roughly 800 to 1500 words) that answer ONE question, far more often than sprawling 4000 word "ultimate guides" that cover the topic exhaustively.

This is a real reversal from the SEO playbook of the last five years. Google rewards depth. ChatGPT rewards extractability. A page that says "Yes, in three sentences" beats a page that says "Here are 12 considerations, in 4000 words" almost every time.

#What this means for SaaS

Three actions worth taking this quarter:

1. Audit your top 10 most-important queries against ChatGPT directly. Not "do you rank in Google?" but "does ChatGPT cite you?" The two answers are uncorrelated more often than you'd expect.
2. Get listed in the niche review sites your customers actually read. Even one good review on a site like SaaSworthy can shift your citation share for an entire query class.
3. Rewrite your "ultimate guide" pages. Or rather, don't rewrite them. Split them. One page per question, each one designed to be the citable answer for that specific question.

The strange thing about AI search is that being invisible feels the same as being unpopular. The metrics that used to tell you that you were on the map (impressions, organic traffic, rank tracking) don't move when citations do, because users who get an answer from ChatGPT don't click through. The first signal you might get that you're losing AI visibility is a drop in pipeline that you can't attribute to anything.

Measure the citations directly. That's the only signal that's honest.