ZLABS
Revised 13 Sep 2026 · 3 min read

How AI Search Decides Who It Recommends

How a question becomes the engine's own searches, why the shortlist is set from the model's memory before any page is fetched, and the measures of what a brand is up against.

ShareLinkedInEmailText

Your question becomes the engine’s own searches.

One buyer question turns into many searches the model writes for itself, and the shortlist of brands is largely set before a single page is fetched.

Your questionSub-queries8+, self-writtenLive web searchWho getsrecommendedThe model’s memoryits shortlist, learned in trainingrewritesthen runsvalidatesseeds the shortlist
Plate 1 — The fan-out. The shortlist is decided at the memory step, before any page is fetched. Live search mostly confirms the names the model already trusts.

You earn your way onto the shortlist.

The shortlist is built from training association, so it rewards genuine standing in the category. That is the anti-spam design, and it is why the position is defensible: the way in is to become genuinely known. Being named in the model’s own query is worth far more than being one of the hundreds of pages it reads.

What you’re up against.

These four numbers are the LBOX AEO Index: our own measurement of the category, run on our own data. They describe the difficulty of the game and a brand’s starting line, snapshot as of 13 Sep 2026.

The prior
3.1×

AI engines recognize a brand far more than they volunteer it.

Recognized on branded questions74.2%
Volunteered on category questions23.9%

When you are named in the question, the engine knows you. When the buyer asks the category with no brand named, that recognition mostly goes quiet. Closing that gap is the work.

LBOX AEO INDEX · n = 4,284 unbranded / 329 branded answers

The shortlist
54%

Five brands take more than half of every category's mentions.

Top 5 brands 54.3%The other ~30 brands

About 35 brands get named in a category, and the top five capture most of the mentions. The shortlist is where the visibility lives.

LBOX AEO INDEX · n = 88 subjects

The sources
15.7%

The sources AI trusts are diffuse and category-specific.

Top 5 domains 15.7%the rest of the long tail

Citations spread across 6,860 domains, led by YouTube, Reddit and brand-owned sites. There is no single directory to buy your way onto; trust is earned across the sources your category actually uses.

LBOX AEO INDEX · n = 28,492 citations

The opening
57%

More than half of every category is still open across engines.

One brand owns it 43%Contested 36%Leaning 21%

Only 43% of category answers are owned by a single brand across all engines. The rest is contested or leaning, which is where the window is open right now.

LBOX AEO INDEX · n = 114 category questions

Plate 2 — The four measures of the challenge. Each is an aggregate over real markets, carrying its sample size. Figures move as the dataset grows; re-read on any frontier model release.

The signal that matters most is coming.

The sharpest signal in AI search is how much more likely you are to be recommended when the model puts you in its own query, versus merely reading your page. LBOX is instrumenting its own measurement of it now, and we will publish our own number the moment the capture lands. Until it is real it stays blank, never borrowed.

Two things decide whether the model recommends you.

Reading the fan-out tells you which of the two you are looking at, so the plan is targeted from day one.

Path A

Earn the association

This is where you become known in the category. A structural build that compounds over time; the association is earned, and once it holds, it lasts. Honest work, worth setting expectations on.

Path B

Strengthen the evidence

The model already considers you and searches you. Here you win by making your page prove the case: specific, addressable, and much faster.

You win AI search by becoming the answer the model already believes in. Now that is measurable.

The LBOX AEO Index is proprietary to LBOX: our own measurement, run on our own data. Snapshot as of 13 Sep 2026, over 28,492 citations, 8,126 brand appearances, 88 subjects and 15 markets. The live instrument runs on the Engine Intelligence surface. Figures move as the dataset grows and should be cited as the LBOX AEO Index.

Published by LBOX.com, a ZLABS platform
ShareLinkedInEmailText
AI search, in depth

Research on LBOX.com

LBOX.com is the ZLABS platform for AI search. Its research library shows how answer engines find, trust and recommend products, with every figure dated and sourced.

Explore the research