Why AI Cites Others Instead of Your Website

Many companies search for themselves in AI for the first time and get a surprise: AI does mention you — but it cites someone else's article, not your website. So the owner asks: was the website a waste?

The conclusion first: AI hasn't made a mistake, and your website isn't useless — this is how AI is designed. It trusts "what others say about you" more than "what you say about yourself". The website is the foundation of the whole source system; third-party content is the reason AI is willing to cite you. The two don't replace each other.

One boundary though: this doesn't mean "post a lot externally and AI will cite you". What you post, where you post it, and whether it matches your website matter far more than how much you post.

First, a counter-intuitive fact

A University of Toronto study (arXiv:2509.08919, preprint, not peer-reviewed) measured AI's source preferences: when multiple sources are available, AI's preference for third-party sources can reach 53% to 95%, well above a company's own self-description.

Scope note: this is a relative-preference range under experimental conditions, not a fixed ratio for every industry or question, and it doesn't mean "over 53% of citations come from third parties". It points to a direction, not a precise scale reading.

But it points clearly at one thing: when answering "what is this company like?", AI would rather find what others say than copy your self-introduction.

Why doesn't AI just trust you directly?

It works like people dealing with people: you saying you're good is a position; someone else saying you're good is evidence.

AI's entire value rests on giving a credible answer. If it copied corporate marketing lines, it would be no different from a company brochure, and users would stop treating it as a reference.

So AI makes a structural choice: it prioritises information with independent sources that can be cross-verified. This isn't aimed at anyone; it's a necessary move to preserve its own credibility.

Conversely, this explains a common sight: companies that talk themselves up on their website but can't be found in AI — because content that only they say carries the lowest weight in AI's evaluation.

What does AI actually read? One easily misread number

Another number worth knowing. arXiv:2410.22349 (preprint, not peer-reviewed) measured how complete AI's citations are: the sources actually listed make up only 20% to 25% of the material it genuinely read.

Which means: not being listed doesn't mean not being read. A lot of content feeds into AI's judgement without getting that visible slot.

This also explains a common puzzle: you clearly saw yourself mentioned in some third-party article, yet asking AI about it turns up nothing. Very likely that article was read, just not selected into the final answer.

Scope note: this figure describes how complete the citation display is; it cannot be reversed into "not cited means no effect".

KDD 2024 tested it: adding evidence does help

The KDD 2024 GEO paper (DOI: 10.1145/3637528.3671900, peer-reviewed conference paper) tested one variable: what happens when you add source citations, quotations and statistics to content.

The result wasn't one-sided: pages originally ranked 5th gained up to 115.1% in visibility, while pages originally ranked 1st actually dropped 30.3%. The same paper found something else: keyword stuffing is ineffective in AI answers, and can even hurt.

Scope note: these percentages are relative changes under lab conditions within a limited query set, not a promise of real-world results; the effect is negative for already-strong pages, which shows it isn't a cure-all.

But the direction is clear: having sources is itself counted into credibility by AI. And sourcing is exactly what third-party content has by nature and self-description lacks by nature.

So what does each do — your website and third parties?

Look at the two separately and the tension goes away.

Website = the master document. When AI verifies who this company is, what it does, where it is and how to reach it, it ultimately comes back to your own pages. Your website is the one source you fully control, and the only master copy that guarantees consistent facts. If the website is unclear, AI can't even do the verification step.

Third parties = evidence. Content on industry media, professional communities and Q&A platforms is what AI uses to judge whether a company's claims are backed by anyone. You don't control it, but you can build it.

The two must agree. This is where things most easily go wrong — your website says A and a third-party article says B; cross-verification turns up the contradiction and lowers trust in both. Many companies trip here: not for lack of content, but because they've muddled their own story.

Three common questions

1. Should we submit content at volume?

Not volume — alignment. Three things matter: publishing on platforms your target AI genuinely reads, matching the questions your target customers ask, and keeping the facts consistent with your website. Piling up near-identical content gets treated as low-quality duplication — the same outcome as spamming backlinks in the SEO era.

2. Does content I post on Zhihu or a WeChat official account count as third-party?

Call it semi-third-party. It carries one extra layer of platform endorsement over your own site, but the byline is still yours. So the writing has to change: not a promotional piece, but an objective answer to an industry question. A simple test: delete the brand name — is the article still worth reading? If it's nothing but advertising, it's a low-value source for AI.

3. Can you buy AI recommendations?

Google's official documentation is blunt: it does not accept payment to improve rankings — the original line is "If anyone tells you otherwise, they're wrong." Paid placement can appear in search ad slots, but that's an ad slot, not a citation inside an AI answer. The two must not be conflated. Any claim of "pay to guarantee AI recommendations" deserves suspicion.

In closing

To sum up: AI citing third parties isn't a rejection of you — it's a request for evidence. What you need isn't to keep insisting how good you are, but to make "others saying you're good" genuinely exist where AI can find it.

The order: first make your website clear (the master document), then leave objective content in the right places (evidence), and finally make sure both say the same thing (consistency). The step most often skipped is the third — and it's exactly where AI is most sensitive.

A small test to start with: ask AI "how is company X?" and look at which sources it cites and whether they match your website. Where they don't line up is what to fix first.

In AI answers, are you cited most from your website or from third parties? Tell me in the comments.

Yihang Network, based in Shenzhen, previously worked in cross-border B2B and Alibaba.com-related services, and expanded into GEO from 2024. Our approach: make the website master document solid first, then build third-party content to match the source structure of your target AI.

Data and scope note: The University of Toronto study (arXiv:2509.08919) and the citation-completeness study (arXiv:2410.22349) cited here are both preprints, not peer-reviewed; the KDD 2024 paper (DOI: 10.1145/3637528.3671900) is a peer-reviewed conference paper, and its percentages are relative changes under lab conditions within a limited query set; the Google material comes from its official documentation. None of these findings represent real-market outcomes or any single company's results. The methods described raise the probability of being retrieved and cited; they do not guarantee a company will appear in AI answers, and no timeline or business result is promised.

1. University of Toronto, arXiv:2509.08919 (preprint, not peer-reviewed)
2. Citation-completeness study, arXiv:2410.22349 (preprint, not peer-reviewed)
3. Aggarwal et al., "GEO: Generative Engine Optimization", KDD 2024 (DOI: 10.1145/3637528.3671900)
4. Google Search Central, "AI features and your website"