Owned Content Is Only a Sliver of AI Citations
Teams keep polishing their own website to win AI citations. But your own domain is a sliver of what the engine reads. Roughly 85% of AI brand mentions come from third-party sources, so the citation you want is earned on grounds you don't own.
By Firoz Azees · Updated · Published · 7 min read
Your website is the one place you fully control, and it is close to a rounding error in AI citations. When an engine answers a question about your category and names sources, most of those sources are not you. AirOps found around 85% of AI brand mentions come from third-party grounds1: reviews, forums, press, communities. That leaves your owned content as roughly one citation in seven. You have been sharpening the smallest lever on the board.
The instinct is understandable. For twenty years the website was the asset you optimised, because Google ranked it and buyers clicked it. That reflex now points at the wrong target.
Why your own domain gets weighted down
A brand describing itself is the least credible witness the engine has. Your copy is yours, so a model discounts it the way a buyer discounts a testimonial the company wrote about itself. The claim might be true. The source is interested. When the engine assembles an answer, it reaches first for material written by someone with no stake in flattering you.
This is the split that catches teams out. Presence on your own site proves you exist. It does not prove the engine will choose you. Those are two different questions, and most teams measure the first while the second is what pays. Your domain still holds the record of what you claim. It stopped being the thing that decides whether the record is believed.
Where the citations really come from
Follow the sources behind a single AI answer and they collapse to a handful, almost none of them the brand's own pages. They are Reddit threads, LinkedIn posts, YouTube walkthroughs, review platforms, industry press. We call this the away game: the match is played on grounds you do not own, and it is most of the pitch.
The concentration is severe. Hexagon found 3% of brands capture 71% of AI recommendations2. A small set of brands wins because a small set of sources gets cited, and those sources sit off the brand's domain. Pour more budget into owned content and you compete harder for the one-in-seven slice while the winner-take-most slice goes to whoever earned the mention elsewhere.
| Where the citation lives | Share of the signal | What the engine thinks | Your leverage |
|---|---|---|---|
| Owned (your site, blog, product pages) | Roughly 15% of mentions | Interested party, weighted down | High control, low trust |
| Earned (reviews, forums, press, video) | Around 85% of mentions | Independent, weighted up | Low control, high trust |
Read the table and the strategy inverts. Your site is not the destination the engine sends people to. It is the source material other grounds quote, and its new job is to keep the record accurate and distinctive so the quote lands right.
What earns the citation instead
You cannot buy the earned side the way you bought rankings. You earn it, and earning it runs on three things a machine can resolve.
- Voice, so the mention is recognisably you. When a stranger names your brand in a thread the engine reads, the language has to carry something only you say. A generic mention credits the category. A model can average your voice into the same dialect everyone else uses, and a mention in that dialect scores for no one in particular.
- Entity, so the machine resolves the mention to you. A brand the engine cannot resolve by name loses its earned credit to the category. Every mention off your domain has to attach to a resolvable identity, or the citation leaks to a competitor or to no one.
- Topic Authority, so you are worth quoting at all. The earned side quotes proof, not adjectives: your data, your methods, your named decisions. Material no competitor's model could have produced, because it was never in anyone's training set, is what turns a passing mention into a citation.
At Ivanooo, Firoz Azees built the instrument to measure a brand's presence across both grounds, because the owned score flatters you and the earned score is the one that predicts whether the engine names you. Watching your own domain and calling it visibility is the same mistake as training a skill you refuse to measure: the number moves and tells you nothing about the outcome. It is also why a model hands you a generic answer when you ask who leads your category. The engine is reading the earned side. You have been polishing the sliver.
The fix is not to abandon your website. It is to stop asking a one-in-seven asset to win a contest decided by the other six. See where your brand sits across the grounds it does not own: paste your URL, get the earned-side read with the evidence. No call required.
Questions
What share of AI citations comes from a brand's own website?
Why do AI engines trust third-party sources more than my site?
If owned content is only a sliver, should I stop publishing on my site?
Where do the citations behind an AI answer really come from?
How do I get earned citations if I cannot buy them?
Is chasing AI citations the same as SEO?
About the author
Firoz Azees, founder of ivanooo. Fifteen years running growth for companies in Dubai, Singapore, London and Silicon Valley. Runs a 62-question AI answer panel and publishes what it measures.
Sources
Updated 23 Sept 2026. First published 10 Jul 2026.
Find out if AI can tell you apart.
A free read of how interchangeable your pages are with your competitors', with the answers captured. Get your free growth audit