The usual advice for getting cited by AI engines is to publish more. One publisher in our dataset followed that advice about as far as it goes, reaching roughly 1,549 pages. AI engines have ever cited 78 of them. The other 95% earn nothing at all, and the pattern in the 5% that work is specific enough to copy.
What we measured
Koalr runs real buyer prompts against live AI engines on a schedule and records every source each engine returns. This piece draws on 300,318 citations across 23 brands and six citation-returning engines, plus a page-level crawl of one high-volume publisher in that dataset.
That publisher is not named here. The lesson is in the shape of the data, not in who it belongs to, and it applies to anyone weighing whether to add another fifty pages.
Does publishing more pages get you cited?
On this evidence, no. Not on its own.
The publisher in question has built a large catalogue: around 1,549 pages in its sitemap, spread across two content sections and several languages. Across a 30-day window of live prompts, engines cited 78 distinct URLs from that domain.
Citations then concentrate hard inside that 78:
| Pages | Share of that domain's citations |
|---|---|
| Top 5 URLs | 49.6% |
| Top 10 URLs | 68.7% |
| Top 14 URLs | 78.7% |
| Top 30 URLs | 92.7% |
Five pages do half the work. Fourteen do nearly four fifths. The remaining ~1,470 pages, the overwhelming majority of the investment, return effectively nothing.
If you are planning a content programme around page count, that is the number to sit with. The programme is not failing at the margin. It is succeeding in a very small, identifiable subset and failing everywhere else.
Two sections, same domain, 39 times the yield
This is the part that changes what you do on Monday.
The publisher runs two separate content sections. Same domain, same category, same period. One is a conventional blog with bylines and dates. The other is a larger library of guide-style pages with no author, no visible date and no article schema.
| Section | Pages | Citations | Citations per page |
|---|---|---|---|
| Blog, with author and date | 596 | 1,419 | 2.38 |
| Library, no author or date | 702 | 43 | 0.06 |
The larger section earns less than 3% of what the smaller one does, per page. Seven hundred pages bought 43 citations.
Be careful about the causal claim: the unauthored section is also weaker content, and the two differ in more than metadata. But the direction is stark, it holds on a single domain at a single moment, and the cheap part of the difference is the part most people skip. An author, a genuine date and article schema cost minutes per page.
What did the pages that worked have in common?
Not range. The opposite of range.
The winning pages were not thirteen different topics. They were the same buyer question, answered thirteen times in different phrasings: ten platforms, eight platforms, twelve tools, one aimed at agencies, one naming specific engines. A buyer asking any phrasing of "which tool should I use for this" meets a page built for that exact phrasing.
That is why the concentration is so extreme. The publisher did not win a category by covering it broadly. It won a narrow, high-intent question exhaustively, and everything else it published was noise around that.
The platform-wide numbers point the same way. Across all 23 brands, pages whose title contains "best" earned 10.6 citations per URL, the highest of any title pattern we tested, over 2,810 URLs. Listicles as a page type earned 9.0 citations per URL. Buyers ask engines to choose for them, and engines reach for pages built to choose.
So what should you publish?
Fewer pages, aimed harder.
A saner content plan for AI citations
- Pick the two or three questions your buyers actually ask before choosing, not the fifty topics adjacent to your category.
- Cover each of those questions in several genuine phrasings rather than covering fifty topics once.
- Give every page an author, a real publication date and article schema. It is minutes of work per page.
- Keep slugs to one to three words. Across our data those earn 9.6 citations per URL against 5.6 for slugs of ten or more.
- Audit at 60 days. Pages with zero citations are a liability on the domain, not a neutral. Rewrite them or remove them.
- Do not mass-produce alternatives pages. At 5.1 citations per URL they were the weakest archetype we measured.
See where your brand shows up in AI search
Enter your website and get your AI Visibility Score across 7 engines. First score in under 5 minutes, no setup.
Methodology
Citations were collected by running real buyer prompts against live AI engines on a recurring schedule through to 2026-07-31, recording every source URL returned in each answer. No prompts were simulated and no citations were inferred from rankings.
The single-publisher figures come from combining those citation records with a crawl of that domain's public sitemap, so the denominator is pages the publisher itself lists, and the numerator is URLs engines actually returned in the window.
Page types were assigned from URL structure using fixed patterns rather than the publisher's own labelling.
Limitations
- One publisher is one publisher. The 5% figure describes a single domain in a single category over 30 days. Treat it as an illustration of how hard citations concentrate, not as a universal rate.
- The 39x section gap is correlational. The two sections differ in metadata and in content quality. The metadata is the cheap, controllable part, which is why we highlight it, but this is not a clean experiment.
- A 30-day window under-counts older pages. A page cited rarely but steadily may show as uncited here.
- Engines are unevenly represented in the platform-wide figures, so cross-engine comparisons are expressed as shares of each engine's own citations.
- Twenty-three brands skew to B2B software and marketing. Other categories may behave differently.
Related reading
For which page types earn citations against which earn brand mentions, see blogs get cited, homepages get recommended. For which specific pages engines pull from when they mention a brand, see which pages AI engines cite.
Frequently asked questions
Not on its own. On one domain we track, only 78 of roughly 1,549 published pages have ever been cited by an AI engine, about 5%, and five URLs account for 49.6% of that domain's citations. Volume without targeting produces pages that are never retrieved.
