Three Sites Built 215,128 Pages to Win AI Citations, Not Google Rankings. That's the Growth Play Now: Optimize for Being Quoted by AI, Not Just Ranked by Search.

The Play

A report set out to answer a simple question: when an AI answer engine recommends "best software" in a category, where does it actually get that answer from?

The finding: mostly not from where you'd expect.

Testing 380 software categories and tracking 1,807 distinct products, the report logged 7,534 total citations across 2,055 distinct domains. And 59.8% of those cited sources rank worse than #100,000 on Tranco — with 23.4% falling outside the top million websites entirely.

Three sites — worldmetrics.org, gitnux.org, wifitalents.com — accounted for a huge share of that footprint: 215,128 best-software pages between them (70,731 / 71,684 / 72,713 respectively), all registered "between December 2023 and May 2024." That's a young, purpose-built content operation, not an organically grown authority site.

Why this works as a growth tactic

Search engine optimization trained a generation of marketers to chase domain authority and backlinks. AI answer engines don't necessarily weight sources the same way. The report tested this directly against perplexity/sonar and perplexity/sonar-pro through OpenRouter, and confirmed "both models report the URLs they retrieved" — so citation behavior can be observed and reverse-engineered, not just guessed at.

What the winning sites had in common wasn't backlinks or brand. It was:

• Volume — over 70,000 pages per site, covering categories exhaustively
• Structure — content formatted as comparisons and rankings, the exact shape an AI answer engine needs to synthesize a "best X" response
• Framing built for retrieval — both worldmetrics.org and gitnux.org self-labeled with HTML titles reading "Facts & Grounding Page," language aimed directly at how retrieval systems select trustworthy-looking sources

Legitimate, high-authority sources still earn real share — g2.com (291 citations, 3.86%), reddit.com (261 citations, 3.46%), gartner.com (158 citations, 2.10%), and guideflow.com (194 citations, 2.57%, ranked #3 overall). But they're competing in the same pool as mass-produced pages that outrank them by volume and structure alone.

How to run this play

1. Pick the categories where you want to be the AI-recommended answer, and build comparison pages shaped the way retrieval systems consume them: named products, clear stats, direct comparisons.

2. Scale coverage instead of polishing one page — the report's sites won partly by covering categories exhaustively, not by having one perfect article.

3. Test your visibility directly: query the AI answer engines your buyers use with "best [category]" and see who gets cited. If it's not you, that's your gap.

4. Make sure you're correctly and completely listed on the legitimate high-citation sources in your space (G2, Reddit threads, Gartner) — those still carry real weight in the citation pool.

5. Re-test periodically. Citation patterns shift as retrieval systems change what they trust, so this isn't a one-time build.

Bottom line

The growth lesson isn't "flood the internet with pages." It's that AI answer engines have their own selection logic, separate from search rankings, and it can be reverse-engineered and tested directly. The sites that figure that out first — and build content shaped for it — win the citation, even against sources with far more traditional authority.

Sources:

https://trellner.com/reports/manufactured-sources-behind-ai-recommendations/