Why it matters
A brand's AI visibility isn't mostly about which page search happens to find
If the winner were driven mainly by whichever page search happens to surface that time, a brand could move its odds largely by getting picked up on whichever handful of review sites tend to get cited. This study instead found the model's own pick largely independent of the specific source: in 6 of 8 categories the winner barely moved across 20 real, live-search repeats spanning two independent rounds, even while the exact cited domains moved far more. That points influence toward what the model already associates with a brand, the subject of most of the rest of this series, rather than toward chasing whichever page search happens to cite in a given moment. Meal kits, the one category where source and winner moved together in both rounds, is the exception worth someone's attention, not the rule this study found.
Update: this page's 90% winner-stability number came from repeating the identical question, a test-retest measure. A LinkedIn reader pushed back that real stability should be measured across differently worded questions with the same intent, not literal repeats. Continued in
Reword the Question Three Ways. The Winner Mostly Stays the Same., which tests that objection directly on the same 8 categories and finds the gap between same-phrasing and different-phrasing pairs was not statistically significant in either of two independent rounds.
Update (2026-09-24): a separate LinkedIn thread raised a third layer, does a brand's rank in the list (1st mentioned, 2nd, 3rd...) hold as steady as the winner itself, or does citation-level churn leak into ordering too? A secondary analysis of this study's own response text, 400 responses across 16 category-rounds here and in Winner Stability Intents, found position drift averaging just 0.25 ranks from a brand's own median position, and pairwise rank correlation of 0.82 between responses in the same category. Every one of the 16 category-rounds cleared a cluster-permutation test against random ordering (p<0.001 each). No new search calls, extracted directly from the response text already collected for this study and for Winner Stability Intents.
Update (2026-09-24): a LinkedIn reader also asked whether this 90% number would hold up in categories with no settled incumbent, since her own data shows sharp variation by category maturity. We tested it directly: 8 fresh categories, 4 established (wireless earbuds, running shoes, drip coffee makers, laptop backpacks) and 4 young (AI wearable pendants, creatine gummies, smart rings, e-bike conversion kits), same repeated-question method, 2 independent rounds. Established categories averaged 88.5% winner stability, young categories 59.2%, a gap that held directionally in both rounds but only reached the edge of significance combined (p=0.052, n=4 categories per group). The split was not purely about age: one young category (smart rings) was 100% stable because it already has a dominant incumbent, while the least stable categories were the ones with no clear leader yet, young or not.
Update (2026-09-25): a Semrush study found that 62% of citations across the wider web never lead to a named brand mention, what they call a ghost citation. We tested the equivalent question on this study's own response text: of the citations that appear inside a single search-enabled answer here, how many actually convert into a named, specific recommendation within that same answer, versus getting cited and then dropped? Across 160 responses and 1,576 citations, 49.9% converted to a named pick and 50.1% did not, close to the same order of magnitude as the wider-web number despite measuring something narrower and different, citation-to-name conversion within one answer rather than across the open web. It varied by category, from 31.7% ghost in headphones to 58.9% ghost in mechanical keyboards. One reassuring check on this study's own method: the winning brand itself was classified as a named citation 98.1% of the time, so the extraction that drives every number on this page is not quietly resting on ghost-cited winners.
