AI Search · 6 min
AI search answers a question in three stages: it interprets what was asked, retrieves material it considers relevant and trustworthy, then composes an answer from that material and attributes some of it. Each stage is a different opportunity to be included or excluded.
How it works
The system rewrites the question into something it can search for, often several sub-queries at once. A question like 'who can help us get found on AI' becomes queries about AI search optimisation, agencies and services. This is why matching a single phrase matters far less than covering the concept the question is really about.
The system searches an index for passages that answer the sub-queries. If your content is not in the index, is not reachable, or exists only after JavaScript runs, you are not a candidate. Nothing later in the process can recover from that.
Among passages that could answer, the system prefers ones that answer completely and unambiguously, from sources it can identify and has reason to trust. A passage that only makes sense in the context of the surrounding page is weaker here than one written to stand alone.
A model reads the selected passages and writes something new, naming some sources. Attribution favours material that is specific and clearly owned, which is why entity clarity matters: a system that cannot tell who published a claim has little reason to credit anyone for it.
Be reachable and renderable. Be identifiable as a specific organisation. Write answers that survive being lifted out of context. Nobody can guarantee inclusion, because retrieval is not controllable, but each of those removes a reason to be excluded.
There is no rank position to report. The available evidence is whether you appear for a fixed set of buyer questions when they are asked repeatedly over time, alongside branded search volume, direct arrivals and the quality of enquiries that mention having found you through an assistant.
Most often because you were never a retrieval candidate: content not indexed, not reachable, or rendered only in the browser. The next most common reason is that the answer exists on your site but not in a form that can be lifted out of context.
Not in the organic answer itself. Some surfaces carry advertising alongside, but citation within a generated answer is not purchasable, and any supplier offering it is describing something else.
AI search answers a question in three stages: it interprets what was asked, retrieves material it considers relevant and trustworthy, then composes an answer from that material and attributes some of it. Each stage is a different opportunity to be included or excluded.
Connected
You have read what we think. If you want to know what it means for your case specifically, describe it and we will tell you which parts of how ai search works actually apply — and which do not.
