LEARN

Why an AI Cites One Page and Not Another

Distinct reservoirs outflow at varying rates, each passing through fine-mesh funnels to converge into selected streams, symbolizing source type determining base citation probability and writing style determining selection within the same type
Source type determines the base probability of being cited; writing style determines who is selected among similar sources — two things to invest in separately

IN ONE SENTENCE

Source type sets the base probability of being cited; how you write decides who is chosen among similar sources — two separate investments.

Public data describes the distribution of source types; empirical research describes the effect of how you write. Together they explain most of 'why them and not us': source type sets the base probability, and writing decides who wins among similar sources.

OUR POSITION

Do not treat the list of most-cited domains as a to-do list. Their share comes from years of community and content accumulation that posting cannot replicate. What is replicable is the writing, plus genuine participation on those platforms.

01

Distribution of source types (it moves — date it)

Semrush's June 2025 analysis of 150,000+ citations across 5,000 keywords: Reddit 40.1%, Wikipedia 26.3%, YouTube 23.5%.

Reporting in early 2026 put YouTube ahead of Reddit as the leading social citation source, with sharp divergence across platforms — the same source can carry very different weight depending on the engine. Any 'most-cited sources' list is time-sensitive.

The Chinese ecosystem is structured entirely differently, split across ByteDance properties, WeChat, Zhihu, CSDN and Baijiahao. The English list does not apply there.

02

The effect of how you write (there is evidence)

The Princeton paper (KDD 2024, GEO-bench, 10,000 queries) tested three methods — adding statistics, adding cited sources, adding quotations — and measured a 30–40% relative lift on visibility metrics.

What they share is checkability: specific numbers, stated sources, sentences that can be lifted directly. Adjective-heavy prose and unsourced assertions do not help.

03

Widely repeated claims without evidence

'Publish llms.txt and you get indexed': Ahrefs' study of 137,000 domains found 97% of those files received zero requests in May 2026, and Google says it does not use them.

'Rewrite content for AI / chunk it for AI': Google's guidance lists both as unnecessary.

'Structured data lifts AI visibility': the same guidance calls it unnecessary for AI search — it is a floor, not a lever.

Data behind this page

Reddit 40.1% · Wikipedia 26.3% · YouTube 23.5%

The three most-cited source types

Source：Semrush, 150,000+ citations across 5,000 keywords，2025-06

under 5%

Ceiling on any single domain's citation share on one platform

Source：Evertune, 200M prompts，2026

30–40%

Relative visibility lift from adding statistics / citations / quotations

Source：Princeton GEO paper, KDD 2024, GEO-bench 10,000 queries，2024

Sources

  1. [1]GEO: Generative Engine Optimization．Aggarwal et al., KDD 2024．2024

Updated 2026-08-10