AI engines converge on which brand to mention well before they converge on where the answer came from. What that gap looks like inside a single category, cut by buyer segment, is not.
The pattern holds for 7 of the 8 buyer segments measured across 54 persona-tagged India beauty search queries. One segment inverts it. For the Gen-Z trend-driven buyer, agreement on the sources cited runs to 0.081 against agreement on the brands mentioned of 0.076.
What the data shows
The measurement covers 54 persona-tagged prompts of 304 measured, an 11-day run in July 2026, drawn from the India Beauty & Personal Care Industry Report. Every figure is an overlap averaged across the engines that answered that prompt, three engines for all but one of the 54.
The reversal, and how thin it is
The Gen-Z row is the only one in the set where the direction flips, and the margin is 0.005 across 10 prompts. That is a reversal, not a rout. It marks where the pattern stops holding, not the size of an effect. Every segment on the chart carries the same caution: support runs from 4 prompts to 11, so these are readings from a narrow set, not settled rates.
What the row points at is a category where trend-driven queries pull answers toward a shared pool of sources while the brands in those answers stay unsettled. Engines answering a trend query reach for the same places and come back with different names.
Everywhere else, the gap holds
The widest gap belongs to the acne-prone or oily skin buyer: brand agreement of 0.361 against source agreement of 0.070. Engines converge hard on which brand to recommend, but rarely on the source behind that recommendation.
It sits alongside a near-tie for second place. The grooming enthusiast's gap of 0.276 rests on the same six-prompt support count as the acne-prone buyer's, close enough that this measurement records both rather than picking a winner by an invented tiebreaker.
What it means for brands
A brand can be the consensus answer to "who" while the "why" behind that answer stays wide open. High brand agreement without matching source agreement means a brand is winning the mention without controlling the citation, the one part of the answer a brand can still shape directly by publishing the kind of content AI engines choose to cite.
For the trend-driven segment the order is different. The sources are close to settled and the brand is not, so the open question there is which name those sources end up carrying.
Methodology
54 persona-tagged prompts of 304 measured, an 11-day run in July 2026, drawn from the India Beauty & Personal Care Industry Report.
Brand agreement and source agreement are set overlaps, computed for every pair of engines answering the same prompt, averaged across those pairs and then across the prompts tagged to each buyer segment. Pairs where neither engine returned a brand, or neither returned a source, are excluded rather than scored as a perfect match.
Engine coverage is scoped per prompt, never a fixed round number. ChatGPT, Google AI Overviews, and Perplexity answered 53 of the 54 prompts; 1 prompt drew answers from every engine this run covered, 8 in total.
The agreement reported here is therefore mostly a three-engine comparison, and the finding holds at that depth: three engines answering one query still reach for different sources.
Per-segment support ranges from 4 prompts to 11. Segment-level figures are directional at that sample size, and the Gen-Z reversal in particular rests on a 0.005 margin.
Zumi is an AI Search Intelligence Platform that measures how AI engines mention brands and cite sources across category questions. Full method and category-level findings: the India Beauty & Personal Care Industry Report.