An AI answer engine does not rank a list of links. It reads a handful of sources, writes one reply, and decides which brands to name inside it. Understand the four steps it takes to get there and you know exactly where your brand can win or disappear. This is the mental model everything else in AI discovery builds on.
An answer is built, not retrieved
A classic search engine returns a ranked list and lets you choose. An answer engine does the choosing for you. It gathers sources, reads them, writes a single reply, and names a few brands as the answer. The list is gone and a paragraph takes its place.
That one difference changes the whole game. You are no longer competing for a click on page one. You are competing to be one of the few names the model includes when it writes the answer. Miss that moment and a buyer never learns you exist.
The four steps every engine takes
Retrieve. The engine gathers candidate sources for the question, from its training data, its own index, or a live web search run in the background. Different engines lean on different mixes, which is why the same question can produce different answers.
Read. It reads those pages, reviews, and threads and pulls out the facts, claims, and names that look like they answer the question. Content that states its answer plainly is easier to read and quote than content that buries it.
Shortlist. It writes one answer and decides which brands to name and recommend. This shortlist is the thing you want to be on. A brand that is accurate, well sourced, and clearly associated with the topic is more likely to make it.
Cite. It links a few sources as its reasons. Being one of those citations is how you appear as evidence rather than a passing mention, and it is how the answer earns trust with the reader.
Training data and live retrieval both matter
Some of what an engine knows is baked into the model during training. Some is fetched live at the moment you ask. Your visibility depends on both: durable, high-quality source material that models learn from over time, and a current web presence that live retrieval can find today.
This is why AI visibility is not a one-time fix. Models get retrained, the web gets re-crawled, and the sources an engine trusts shift. An answer that names you this month can drop you next month if a competitor earns a stronger source.
The two moments that decide your visibility
Of the four steps, two are where brands are won and lost: the shortlist and the citations. The shortlist decides whether you are named at all. The citations decide whether the answer treats you as evidence.
Everything you do for AI discovery aims at those two moments. Be present and accurate so you make the shortlist, and earn the sources so you become a citation. The rest of this library is about how to do each one.