THE LEDGER ABD SHANTI
CITATION STRENGTH0%
SOURCE #045 VERIFIED · LIVE SEP 07, 20266 MIN READMANIFESTO PART OF GETTING CITED

Most of Your Citations Happen on Pages You Do Not Own

¶ MANIFESTOTHE ARITHMETIC

01Line the numbers up and look at who holds the pen

The six entries before this one each took a single place that AI answers are built from and looked at it on its own. Lists of the best tools in a category. Reddit threads. Wikipedia. Review platforms. Video transcripts. Pages written in languages other than English. Each one came with a number, and each number sounded like a niche worth a paragraph. Put them next to each other and they stop looking like niches.

Listicles carry roughly 44% of AI citations. Wikipedia and Reddit together account for more than a quarter of what ChatGPT cites. A review about you carries around three times the weight of a page you wrote about yourself. YouTube is the most cited single domain on both ChatGPT and Perplexity. Localised pages win between 85 and 96% of the citations in languages other than English.

Why you cannot add them, and why it does not matter

Those figures come from different studies, different engines and different samples, so they do not sum like a bank statement, and anyone who adds them into one tidy percentage is selling something. They do not need to add up. The pattern is visible without the arithmetic: on every surface in that list, somebody other than you writes the sentence that gets repeated. Your website is a minority shareholder in your own reputation.

// WHO HOLDS THE PENSIX SURFACES
Best of lists
~44% of AI citations
The editor of somebody else's list
control · None
Wikipedia and Reddit
>25% of ChatGPT citations
Volunteers and strangers
control · None
Review platforms
~3x the weight of self description
Your customers
control · Some
YouTube
Most cited single domain
A channel, through its transcript
control · Some
Pages in other languages
85 to 96% of non English citations
Whoever localised first
control · Some
Your own website
The facts, not the verdict
You
control · Full
Different studies, different engines, so do not add the figures up. Count who holds the pen instead.
¶ MANIFESTOWHAT THE MODEL IS DOING

02Why the answer is assembled from other people's pages

A model asked which tool to use is doing something closer to reporting than to reading. It wants the claim, and it wants somebody other than the claimant to have made it. A founder saying the product is the best in its category is a data point about the founder. Twelve unrelated people saying so is a data point about the product.

Self description is discounted by design

None of this is hostile to your website. Your own pages are still where a model learns what you are called, what you cost and what you do, which is why saying the same true thing everywhere still matters as much as it does. What your pages cannot do is vouch for you, and no amount of polish changes that. A website can describe you. Only somebody else can recommend you.

The page you polish is the one that counts least

Most marketing budgets are spent in almost exact inverse proportion to that weighting. The homepage gets redesigned every year. The comparison list for your category, the thread where somebody asked for alternatives, the review profile with eleven reviews from two years ago, those get nothing, and those are the pages a model leans on before it decides who to recommend. Companies spend the most money on the page that carries the least weight.

There is one honest exception. In a category so new that nobody has written the comparison yet, your own explainer can be the only source a model finds, and it will quote you because there is nobody else to quote. That window closes the moment a stranger publishes a list, which is the best argument for becoming the page that list links to for its definitions rather than one more vendor it ranks.

EXTRACTED — THE SENTENCE THIS ENTRY EXISTS FOR
You do not own the pages that decide your reputation. You can own the reasons those pages mention you.
¶ MANIFESTOWHAT CHANGES

03Stop publishing more, start being easier to mention

The instinct after learning this is to publish more about yourself, faster, on more of your own pages. It is the wrong direction. More self description only grows the part of the conversation that gets discounted, and it spends the hours you needed for the part that does not.

Make your facts copyable

Every surface in that table is written by somebody in a hurry. The person adding you to a list, answering a thread or leaving a review will use whatever facts they can find in under a minute. Whoever describes you in a hurry will use the first fact they can find. A plain page with your price, your limits, what you do not do and the date those facts were last checked turns you from a product they half remember into an entry they can write correctly. It also helps on the one surface nobody can write for you at all, the encyclopaedia, where the rule is that you are not allowed to write the page and the facts you feed it are all you control.

Earn mentions before you chase traffic

The practical order changes. The first job is not another blog post, it is being present in the handful of places your category is already discussed, with a description accurate enough to survive being repeated by a stranger. Traffic follows mentions more reliably than mentions follow traffic. That is slower than publishing, harder to measure than ranking, and it is where most of the citations already are.

Tomorrow: the engine with the most users, and why it is the least interested in your news.

// QUICK ANSWERS
>Where do most AI citations actually come from?+
Mostly from pages the brand does not own. Listicles carry roughly 44% of AI citations, Wikipedia and Reddit together account for more than a quarter of ChatGPT's, and a review about you carries around three times the weight of a page you wrote yourself. Those figures come from different studies and cannot be added into one percentage, but on every major surface somebody other than the company writes the sentence that gets quoted.
>Does my own website still matter for AI citations?+
Yes, for facts rather than verdicts. Your own pages are where a model learns your name, price, limits and what you do, and inconsistent facts there make every third party mention harder to trust. What your site cannot do is recommend you, because self description is discounted by design. The recommendation has to come from somebody else.
>What is off site GEO?+
The work of being mentioned accurately on pages you do not control. In practice it means being present on the lists, threads, review profiles, videos and localised pages where your category is already discussed, with facts plain enough that the person writing about you gets them right. It overlaps with digital PR, but the goal is an accurate sentence that survives being repeated rather than a link.
>Should I publish more content to get cited more often?+
Usually not first. More self description grows the share of the conversation that models discount. The higher return order is to make your facts easy to copy, then earn presence where your category is already discussed, and only then publish more on your own site, ideally the explainer that other pages end up citing.
Abd Shanti, author of CITED
VERIFIED HUMAN
Abd Shanti
GEO EXPERT · THE AUTHOR
// CITE THIS ENTRY
Abd Shanti. "Most of Your Citations Happen on Pages You Do Not Own" CITED, Entry 045, Sep 7 2026. unknown.ps/blog/pages-you-do-not-own/
// RELATED ENTRIES
QUOTE COPIED — CITE FREELY