TL;DR: There is a failure mode in AI search that does not exist in Google. You can be read constantly and recommended almost never. Across 608 answers we measured in one category, the brands supplying the reading and the brands getting the recommendation overlap far less than anyone expects.
The models are reading us. They are using our pages to construct answers about which agency someone should hire. Then they name somebody else.
Two jobs that feel like one
When an answer engine responds to "which agency should I use for X", it does two things.
It retrieves. It finds passages relevant to the question, from wherever it can. This is what most content strategy is built for, and a clear page answering a specific question in ordinary language is exactly what gets retrieved.
It decides who to name. A different operation with different inputs. It is looking for evidence that a business exists, is real, and is the kind of thing a person would be pleased to be pointed at. That evidence is mostly not on your website, because a website claiming excellence is not evidence of anything.
Optimize hard for the first and not at all for the second and you get our number.
The other end of the distribution
First Page Sage, the most-named brand in the same dataset, shows the inverse. Cited as a source 59 times. Named 147 times.
They are named more than twice as often as their own material is read. That only happens one way: they are being named inside other people's content. Round-ups, comparison pieces, industry lists, forum threads. When a model assembles a shortlist, their name is already sitting in the sources it reads.
Your own pages influence whether you are retrieved. Other people's pages influence whether you are recommended.
Why almost nobody measures this
Most AI visibility tools report one number, some version of "are you appearing". Depending on the tool that can mean cited, mentioned, linked, or a blend, and the blend is the problem.
If a tool tells you that you appear in 8% of answers, you cannot tell whether you are being read and passed over or not being read at all. Those situations need opposite work. The first means your content is fine and your third-party presence is missing. The second means the content is not being retrieved and no amount of outreach helps.
The crude version you can run yourself
Ten prompts, twenty minutes, no tools.
Open a logged-out or temporary chat. In your normal account the assistant has seen you discuss your own business and will bring it up, which is memory rather than visibility.
Ask each question and record two things: was your business named in the answer, and did your domain appear in the citations.
Then compare. A ratio well below one, named far less often than cited, means you have a retrieval gap. A ratio above one means you are being named in other people's content, which is the position you want.
What closes it
Nothing on your own website, which is the annoying part.
Get into the sources that already get cited. In our category those were Reddit at 225 citations across 608 answers, YouTube at 145 and LinkedIn at 134, all of them ahead of every agency website in the field. In yours it will be a different list and it is measurable.
Be in the round-ups. Somebody is already writing "best X for Y" in your category and the model is already reading it.
Publish something worth citing. Original data is the most reliable route into someone else's article, because it gives them a reason to mention you that is about their piece rather than about you.
Then make sure a model checking your legitimacy finds a clear answer. Consistent details across the places it will look.
How long closing it takes
Longer than anything else on the list, which is why it is worth starting first.
Weeks one to six. Nothing visible. Corrections and claims land, pitches go out, community presence starts. The citation figure may move; the named figure almost certainly will not.
Months two to four. First external mentions appear. The named figure starts to move on the narrowest questions first, because those have the fewest competitors and the thinnest existing sources.
Months three to six. Broader questions follow. This is the point at which the gap measurably narrows rather than anecdotally.
The reason it is slow is structural. Closing a retrieval gap depends on other people publishing things, and you can create the reason for that but you cannot set the date. Anything promising to close it faster is either working in an empty category or measuring something else.
Method
39 buyer prompts, five answer engines, 608 answers collected on a rolling schedule between 24 May and 24 August 2026. A mention is the brand named in the answer body, counted once per answer. A citation is the domain appearing as a source. Full method and dataset are published.
What the two situations cost you
Worth being concrete, because the gap is often dismissed as a technicality.
Cited but not named. Your pages are doing unpaid work for a competitor. A model reads your explanation of the problem, uses it to construct the answer, and then names someone else as the solution. Every page you publish makes the answer better without making you the answer. This is the more expensive of the two situations and the one that feels most like failure while looking most like progress.
Named but rarely cited. Less common and much healthier. The models know who you are from third-party sources but are not reading your own material, usually because there is not much of it or it is not in a retrievable shape. This is the faster of the two to fix.
Neither. You are not in the conversation. Frustrating, and the most tractable of the three, because both levers are available and the first one shows movement in weeks.
The reason to measure them separately is that the work is close to opposite. Closing a retrieval gap means other people writing about you. Closing a citation gap means writing better pages yourself. Doing the second when you have the first is the single most common wasted year in this category.
What closing it produces
“We've always grown through referrals - builders who know us pass our name on. That works, but it only reaches people who already know someone in the industry. Within a couple of weeks of the content going live, we had someone contact us directly through the website. That's a channel we didn't have before - and the enquiries have kept coming.
PROCERT is a building certification firm that had grown entirely on referrals. Within a couple of weeks of the work going live, someone who had never met them inquired directly through the website. Fur Magic went from absent to the most-cited brand in their category in three weeks, ahead of long-established competitors with far more online presence.
Both published with the starting number, the end number and the period: PROCERT and Fur Magic.
A worked example of the gap closing
A B2B services firm, cited on twelve of twenty-five tracked questions, named on two.
Month one. Diagnosis and correction. Two association listings claimed, one license record corrected where the business name had drifted. Nothing visible happens.
Month two. One original piece of research published, small, on a question in their category nobody had put numbers to. Three pitches sent. Nothing visible happens.
Month three. Two pitches land, both in industry round-ups. The named figure moves from two to three, on the narrowest question in the set.
Month four. A community thread they contributed to starts appearing as a cited source. Named figure moves to five.
Month five. The research piece gets cited by two other writers, which is the mechanism doing exactly what it is supposed to. Named figure moves to seven.
Month six. Seven of twenty-five, from two.
What it took. One piece of original research, four external placements, and about twenty minutes a week in one community. No increase in publishing volume at all.
Why the first two months look like failure. Because they are indistinguishable from failure until month three. That is the honest cost of this fix and the reason most people abandon it.
See which questions name you
A free AI visibility audit on your own buying questions, across all five answer engines.
Book your audit
Ashur Homa
Built and scaled a digital brand to $100M+ in sales with zero ad spend. Has helped businesses generate millions through AI go-to-market strategy. Leads growth at Omni Eclipse.
Connect on LinkedIn
