27,000 pages, zero AI citations

Somewhere in New York there is a law firm website with nearly 27,000 pages. Across seven weeks of AI answers to the lawyer questions we track, ChatGPT, Gemini, and Google AI Overviews cited it zero times. The firm publishes more pages than almost anyone in the market and gets nothing back from AI. This article is about why, and about what the sites that do get cited look like instead.

We read the sitemaps

Every serious website publishes a sitemap, a machine-readable index of its own pages. We collected the sitemaps of 2,626 New York law firm websites and could read 2,026 of them (517 sites publish no sitemap, and the rest were empty or unreachable). That gives us a content census of the market: the median firm site has 103 pages, 1,526 of 2,026 sites have a blog or news section, and 1,798 of 2,026 have practice area pages. We then matched each site against seven weeks of AI answers to the questions New Yorkers type when they need a lawyer, the same panel behindour ranking study.

The first cut looks like a clean win for publishing more:

A firm with a 500-page site is about seven times as likely to be cited as a firm with a 20-page site. If the story ended there, the advice would write itself: publish more pages. But look at the top bucket. Sites with more than 1,000 pages get cited less often than sites a third their size. The biggest publishers in the market underperform, and the 27,000-page site from the opening paragraph is the extreme case.

The giants that AI ignores

So we read the sitemaps of the largest never-cited sites page by page. They have practice area sections, blogs, and thousands of attorney-written articles. What they lack is any meaningful coverage of the problems in our tracked searches: car accidents, arrests, divorce, immigration, disputes with an employer or a landlord. These are corporate defense giants. Their pages advise companies.

Once you separate sites by what their pages cover, the size effect from the first chart mostly dissolves:

Among sites whose practice pages address consumer legal problems, roughly half get cited, and a 3,000-page site does no better than a 400-page site. Among sites whose practice pages address corporate work, citation rates stay in single digits at any size. Page count matters at the bottom of the range, where a 15-page brochure site gives an engine almost nothing to cite. Past a real website, the engines respond to subject matter. The cleanest illustration runs the other way: one 20,000-page firm site in our census, built around consultation pages, case writeups, and FAQ pages for criminal charges, divorce, DUI, and immigration, collected 79 AI citations over the window.

The path that skips page 1

Our ranking study showed that 69 to 83 percent of AI citations point at sites outside that day's Google top 10. The census puts a firm-level number on the same pattern. Of the 547 firm sites cited at least once in the window, 354 never appeared in the organic top 10 for any tracked search on any day we collected. That is 65 in 100 of the cited firms, and the pattern holds on every engine:

The 79-citation firm above is one of the 354: it collected all of those citations without a single day on page 1. Ranking still works when you have it, and firms that reached the top 10 were nearly always cited somewhere. The census shows the door most firms actually walk through. Two thirds of the firms AI cites got there on content alone.

What this means for your firm

• A brochure site is close to invisible. Sites under 26 pages were cited in 34 of 520 cases. Among sites with no practice area pages at all, 9 of 228 were cited. An engine assembling an answer needs a page that addresses the question; a homepage with a phone number gives it nothing to quote.
• Write for the problems your clients bring you. The engines cited half of the consumer-focused sites and 14 of 198 of the corporate-leaning ones, at identical page counts. A page that walks through what happens after a DUI arrest in Queens earns citations that a thought-leadership essay does not.
• Volume past a few hundred pages buys little. Citation rates flatten around 300 to 1,000 pages. The 27,000-page site and its peers show where the treadmill leads when the pages cover the wrong questions.
• You can be cited without ranking, and you should check.354 of 547 cited firms never saw page 1. Your rank tracker will not show you any of this. The way to know is to ask the engines the questions your clients ask and watch the answers, which iswhat we measure, every day, per engine.

Usual caveats: one metro, one seven-week window (July 1 to August 21), and the answers to our tracked searches rather than every question anyone could ask. A sitemap is what a site declares about itself, so a site that hides pages from its sitemap is undercounted. Practice page coverage is measured by keyword matching against page addresses, which reads employer-side employment pages as consumer-adjacent in a handful of cases. All of this is correlation measured carefully, and none of it proves that adding a page causes a citation. What the sites AI already cites have in common isits own article.