Playbook
Get mentioned by AI assistants.
Buyers increasingly ask an assistant for a shortlist before they ask a person. If you are not in that answer you are not in the evaluation, and you will never see it happen in your analytics. This is the routine for measuring it and moving it.
The shortlist now happens without you.
The old version of being invisible was ranking on page two. The new version is an assistant naming four vendors in your category and not naming you, and it leaves no trace in your traffic at all. There is no referrer for a conversation you were absent from.
Which makes this the one marketing problem you genuinely cannot detect by looking at your website numbers.
Step 1 — Get a baseline before you change anything.
AI Visibility asks the assistants your buyers use and records what comes back: an overall score, a trend line per engine, and a breakdown by buying stage, so you can see whether you are absent at awareness, at consideration, or at the decision point.
Do this first. Without a baseline you will publish for a month and have no way of knowing whether it did anything, which is how content programmes get cancelled.
Step 2 — Read how the assistants describe you.
The score matters less than the wording. AI Visibility shows the sentences the models produce about you, and that is where the useful discomfort lives.
Common finding: the description is accurate about what you sell and completely wrong about who it is for. That is not a visibility problem, it is a positioning problem, and no amount of publishing volume fixes it.
Step 3 — Find the stage where you disappear.
Strong at decision and absent at awareness means people who already know you find you, and nobody else does. The fix is explanatory content about the problem, not more product pages.
Absent at consideration usually means there is nothing published that compares you to the alternatives a buyer is weighing. Assistants summarise what exists, and if the only comparison of you and your category was written by a competitor, that is the one that gets summarised.
Step 4 — Publish the thing that is missing, then re-measure.
Write for the gap the baseline showed you rather than for a keyword. Assistants synthesise across sources, so a clear, well-structured page that answers a real question does more than a page stuffed with the phrase you want to rank for.
The longer argument, including how to structure a page so a model can actually extract it, is in the AEO playbook.
Then measure again. Monthly, not daily. These numbers move on the timescale of model updates and crawl cycles, and watching them every morning tells you nothing except that you are anxious.
Questions worth asking
How long does this take to move?
Weeks to months. Assistants reflect what has been published and indexed, so you are waiting on the same slow machinery as search, plus the model’s own refresh cycle. Anyone selling a faster number is selling something.
Is this just SEO with a new name?
It overlaps, and it is not the same. Search ranks pages and sends a click. An assistant synthesises an answer and may send nothing at all, so being extractable and being quotable matter more than being ranked.
Which assistants does Adamiro check?
The dashboard reports per engine and includes ChatGPT, Perplexity and Google’s AI Mode, with an all-engines view over the top. Engine coverage changes as these products do.
What if the assistants describe us wrongly?
That is the most useful thing this can tell you, and the fix is usually publishing a clear statement of who you are for rather than trying to correct the model. They summarise the public record, so change the record.
Last reviewed: August 2026