Why AI search is not recommending your business
Written by
Odin Lund
Category
AI search
Date
Share this article
Reason one: you are absent from the first query
The most common and the most expensive. The model names two or three options at the first vague question, and everything the customer asks afterwards is asked about that shortlist.
In our legal data set the businesses named in the first-query answers were, with few exceptions, the businesses named later in the journey. Businesses absent from the first answer rarely re-entered.
Test for it directly: run the vaguest question a customer would start with, not the one that already contains your category and your city. If you are missing there, strong decision-stage numbers are a comfort rather than a signal.
Reason two: your facts disagree with each other
A model that finds two versions of a fact about you does not pick one. It hedges, or it drops the claim and names a business it can describe confidently.
Founding date, office locations, what you actually do, which services you offer — these are commonly stated two or three different ways across a website, a directory listing and a review profile, and nobody notices, because no human reads them side by side.
The audit is mechanical: list every claim the models make about you, trace each one to the source it came from, and compare those sources against each other.
Reason three: you are missing from the sources they cite
The third reason is that you are consistent and correct on your own site, and your own site is not mostly what the model reads.
Across 4,222 answers, the third-party sources carrying the client produced 18,742 retrievals, against 4,319 from the client’s own website and 1,296 from the nearest competitor’s. A small number of domains carried most of it.
If those domains do not mention you, no amount of work on your own pages will put you in the answer.
How to tell which one is yours
Run the prompt set and read the citation record rather than the scores.
Absent from first-query answers but present later: reason one. Named, but described inconsistently or hedged: reason two. Correct everywhere on your own site and still absent: look at which domains the models retrieved and whether you appear on them, which is reason three.
Most businesses have two of the three. Fix the facts first, because a conflict poisons everything downstream of it, then the first-query content, then the earned sources.
Share Blog
