Listed Is Not Recommended: The Distinction That Decides Your AI Budget
By Firoz Azees
In our Dubai panel one agency's pages were cited 4 times while it was named once. Another was named 3 times and cited zero. Those are two different outcomes, bought two different ways, and confusing them is the most expensive mistake in AI visibility.
7 min readAsk an AI assistant a buying question and two separate things can happen to your brand. The engine can use your page as a source, footnoting you beneath the answer. Or the engine can name you inside the answer as the thing to choose. Most AI visibility reporting collapses these into one number. They are different events, they cost different money, and buying the wrong one is the most common waste in this category.
The short version
We ran 8 Dubai buyer questions through Google AI Mode and saved every answer as a receipt. In our panel the two outcomes came apart completely. Push Group's pages were cited in 50% of answers while the brand was named in 12.5%. Digital Nexa was the mirror image: named in 37.5% of answers with a 0% citation share. One supplied the engine's words. The other got the engine's endorsement. Neither achieved both, and no dashboard that reports a single "AI visibility score" would have shown you the difference.
What each event means
Cited
The engine pulled from your page and listed your URL as a source. Your content answered the question. Your brand may not appear in the answer text at all, and a reader skimming the answer may never register your name.
Named
The engine wrote your brand into the answer. The reader sees you as part of the recommendation. Your page may not be cited anywhere in the sources, which means the engine knows of you without reading you today.
Recommended
The strongest state: named inside a sentence that endorses. In the 62-query panel we run on ourselves this is rare enough to track separately: 1 question named us, 2 cited us, and 59 returned us absent, so our own Share of Recommendation currently reads 0.0.
What we measured
| Agency | Named (of 8) | Cited (of 8) | Named % | Cited % | Pattern |
|---|---|---|---|---|---|
| Push Group | 1 | 4 | 12.5% | 50% | Source, not endorsement |
| Digital Nexa | 3 | 0 | 37.5% | 0% | Endorsement, not source |
| AIIMS Group | 0 | 3 | 0% | 37.5% | Source only |
| Houses of Growth | 0 | 3 | 0% | 37.5% | Source only |
| SEO Sherpa | 1 | 0 | 12.5% | 0% | Endorsement only |
| Bright Ideas | 1 | 2 | 12.5% | 25% | Both, weakly |
| Ivanooo | 0 | 0 | 0% | 0% | Neither |
Read the first two rows together. If you had bought "AI visibility" from a dashboard reporting one blended number, both agencies would have looked similar. In reality they occupy opposite positions, and the work required to move each one is different.
Why the two come apart
Citation is a retrieval problem
The engine cites what it can parse, trust and reuse. That is structure, freshness, clean markup, a direct answer near the top of the page, and sources it can check. The Princeton GEO research found that passages carrying citations and statistics lifted their visibility in generated answers by up to 40% in controlled tests. Retrieval responds to engineering.
Recommendation is a distinctiveness problem
The engine names what it can tell apart. If your positioning reads as the category average, the model has no reason to prefer your name over any other, and it will cite your useful page while recommending someone with a sharper identity. This is a Voice problem, and it is why most AI answers sound generic: the models default to the mean unless something in the evidence gives them a reason not to.
The practical consequence
Structure work moves citation. Identity work moves recommendation. A retainer that only ships schema and FAQ blocks can raise your citation share for years without ever getting you named.
How to tell which one you are missing
- Ask your 8 buyer questions in a clean browser. Save the full answer and the source list under it.
- Count your name in the answer text. That is your named score.
- Count your domain in the source list. That is your cited score.
- Compare the two. High cited and low named means an identity problem. High named and low cited means a retrieval problem. Both at zero means you are not yet in the consideration set at all.
- Repeat after 60 days. A single reading cannot tell you whether you are moving, and positions in this category churn hard enough that one date proves nothing.
What each gap costs to close
| Your pattern | The actual problem | The work that moves it |
|---|---|---|
| Cited, not named | Nothing distinguishes you | Positioning, proof assets, a claim only you can make |
| Named, not cited | Pages are not retrievable | Structure, answer blocks, freshness, schema |
| Neither | Not in the consideration set | Entity clarity first, then both of the above |
| Both, weakly | Spread thin | Concentrate on the questions closest to money |
Questions buyers ask
Which is worth more, cited or named? Named, for most businesses, because the reader sees it. Cited still matters: it feeds the engine's picture of you over time and is usually the easier of the two to earn first.
Can I be recommended without being cited? Yes, and Digital Nexa demonstrated it with a 37.5% named share and zero citations. The engine knew the brand from its wider footprint without using its pages that day.
Does a high citation share eventually produce recommendations? Not automatically. The two respond to different inputs. Citation compounds your presence in the evidence; recommendation still requires a reason to prefer you.
How do the tools report this? Most report a single blended visibility figure. Of the 31 GEO products we track, the category norm is share-of-answer reporting, which flattens the distinction this page exists to make. Ask any vendor to split the two before you buy.
Why is Ivanooo at zero on both? Because we publish our own numbers rather than only our clients'. As Firoz Azees puts it: "a measured zero is a work order; a flattering estimate is a story you tell yourself." Our zero tells us exactly which of the two problems we have, which is the same diagnosis we would run for you.
Is this specific to Google AI Mode? The distinction holds across engines; the counts do not transfer. Each engine builds its answer differently, so measure the engines your buyers use rather than assuming one reading covers all.
What is the fastest thing to fix? Retrieval, usually. Structure and answer blocks are engineering work with a short feedback loop. Identity takes longer and is worth more, and the limits of what this discipline can do are worth reading before you set a timeline.
At Ivanooo, Firoz Azees runs Distinctiveness Engineering for the AI-answer era: measuring who the engines name, cite and recommend, then engineering the gap between listed and chosen. Start with a free AI visibility check and find out which of the two you are missing.