Why AI answers are decided off-page
A language model answers based on what it has read about you everywhere, not only what you write on your own site. It weighs mentions, links and context from across the web to judge whether your brand is a credible source to bring forward. On-page work makes you citable; off-page work makes you chosen.
It is the same logic classic search has always rested on, that other people's trust weighs heavier than your own claim. The difference in AI search is that the signals are read by a machine that summarises instead of listing, so being visible in several credible places matters more than ranking in one.
Mentions, not just links
For a model to connect your brand with a topic it needs to meet you in the right context, again and again. A mention without a link counts: when industry media, independent experts and relevant sites write about you, the entity the model later anchors the answer to gets built. The link reinforces it, but the association itself is the signal.
The practical work is therefore about earning a place in the right conversations, being referenced where your audience and the models already read. It is patient and editorial, not a volume exercise.
Relevance over volume
What makes a model trust you is where and how you are mentioned, not how many links you have gathered. A reference from a source your audience already trusts weighs heavier than a large number from the wrong places, and that relevance is what digital PR builds. How that link strategy works in practice, and why "buying links" is the wrong question to start from, I go through in a separate post.
Most of it is old SEO hygiene
Under the new acronyms, AEO and GEO, sit requirements that have been around for a long time. Your site has to be fast, crawlable and clearly structured to be citable at all, and the sender has to be identifiable: named experts, a clear organisation and consistent information across the web. The off-page signals only gain their weight once the foundation on the site holds.
Measure against the business, not against mentions
Visibility is a means. The question is always what it should achieve for the business, not how many times you appear in an AI answer. I track which sources the models cite, but the measure is the same as in every other kind of search: leads and pipeline, not visibility itself.
A brand is not cited because it asks to be, but because others are already talking about it.
If you want to know where you are already mentioned and where the gaps are, the AI-search work starts with exactly that mapping.