Luxury was never about being seen. It's about being wanted.
The first benchmark of how five AI models perceive, surface, and narrate ten luxury fragrance Maisons. 2,400 answers, on two axes most teams never separate: how visible a Maison is in AI, and how desirable AI makes it sound.
Scope of the edition.
- 5
- AI models
- 10
- Maisons
- 2,400
- Answers
- 6
- Dimensions
- 2
- Axes
- 70.8
- Highest Olfactory Authority of all ten, scored by a Maison founded in 2009.
- 32 pts
- Gap between how AI reveres Hermès and how it surfaces it where buying begins.
- 34%
- Share of the most-cited Maison's sources it does not control.
- 0% / 97.5%
- Google AI Overviews trigger rate, French vs English.
Visibility is not desirability.
Visibility tells you how often AI mentions you. Desirability tells you how AI describes you when it does. The first is reach; the second is reverence. Most digital teams measure the first, almost no team measures the second, and the distance between them is where the work lies.
In March we measured whether AI sees the great Maisons. This index measures whether it wants them. Same vertical, second axis, harder question.
Where ten Maisons stand on both axes.
The horizontal axis reads AI Presence — how often, prominently, and gift-relevantly each Maison surfaces. The vertical axis reads Cultural Desirability — heritage depth, olfactory authority, cultural resonance. Dividers sit at the sector medians.
Map — accessible table
| Maison | Group | AI Presence | Cultural Desirability | Zone |
|---|---|---|---|---|
| Chanel | Independent | 78.3 | 72.9 | Cultural Icons |
| Dior | LVMH | 68.2 | 68.2 | Cultural Icons |
| Guerlain | LVMH | 68.1 | 72.8 | Cultural Icons |
| Maison Francis Kurkdjian | LVMH | 64.6 | 60.6 | Cultural Icons |
| Tom Ford | Estée Lauder | 60.7 | 41.9 | AI-Native Stars |
| YSL Beauté | L'Oréal | 56.4 | 49.4 | Awaiting AI Maturity |
| Hermès | Hermès International | 53.0 | 66.9 | Heritage-Strong / AI-Emerging |
| Diptyque | Manzanita Capital | 48.4 | 43.6 | Awaiting AI Maturity |
| Byredo | Puig | 47.3 | 33.9 | Awaiting AI Maturity |
| Le Labo | Estée Lauder | 46.1 | 33.7 | Awaiting AI Maturity |
How to read the scores.
Each dimension scores 0–100, model-averaged across five engines. It is not a percentage, not sales, not a market share. It composes four signals — presence (mention), prominence (first-mention position), tone (valence), depth (richness) — plus a category layer for the three desirability dimensions (ingredient & perfumer vocabulary, cultural references). Gaps under one point sit within run-to-run noise.
AI Presence
How often, how prominently, how gift-relevantly a Maison surfaces when AI is asked the questions clients ask.
Product Discovery
AI is asked to recommend perfumes — does the Maison surface?
| # | Maison | Score / 100 |
|---|---|---|
| 1 | Chanel | 71.4 |
| 2 | Guerlain | 64.5 |
| 3 | Dior | 55.4 |
| 4 | Maison Francis Kurkdjian | 51.6 |
| 5 | Tom Ford | 50.5 |
| 6 | YSL Beauté | 42.9 |
| 7 | Byredo | 36.2 |
| 8 | Hermès | 34.6 |
| 9 | Diptyque | 33.9 |
| 10 | Le Labo | 29.2 |
Brand Knowledge
AI is asked to describe the Maison — does it know enough to answer?
| # | Maison | Score / 100 |
|---|---|---|
| 1 | Dior | 89.3 |
| 2 | Chanel | 89.2 |
| 3 | Hermès | 88.5 |
| 4 | Guerlain | 87.8 |
| 5 | YSL Beauté | 87.0 |
| 6 | Diptyque | 83.9 |
| 7 | Tom Ford | 82.4 |
| 8 | Le Labo | 81.2 |
| 9 | Maison Francis Kurkdjian | 80.4 |
| 10 | Byredo | 80.0 |
Gifting Occasion
AI is asked for gifts (Mother's Day, anniversary, wedding) — does the Maison appear?
| # | Maison | Score / 100 |
|---|---|---|
| 1 | Chanel | 74.4 |
| 2 | Maison Francis Kurkdjian | 61.7 |
| 3 | Dior | 59.9 |
| 4 | Guerlain | 51.8 |
| 5 | Tom Ford | 49.1 |
| 6 | YSL Beauté | 39.2 |
| 7 | Hermès | 36.0 |
| 8 | Le Labo | 27.9 |
| 9 | Diptyque | 27.3 |
| 10 | Byredo | 25.8 |
Cultural Desirability
How AI describes a Maison once it surfaces — its history, its perfumers, its cultural place.
Heritage Depth
AI is asked about lineage, founders, archives — what depth does it carry?
| # | Maison | Score / 100 |
|---|---|---|
| 1 | Guerlain | 89.5 |
| 2 | Chanel | 88.9 |
| 3 | Hermès | 83.5 |
| 4 | Dior | 66.9 |
| 5 | Maison Francis Kurkdjian | 59.4 |
| 6 | Diptyque | 39.0 |
| 7 | YSL Beauté | 30.0 |
| 8 | Le Labo | 1.3 |
| 9 | Byredo | 1.3 |
| 10 | Tom Ford | 0.0 |
Olfactory Authority
AI is asked about scents, ingredients, perfumers — how technically credible?
| # | Maison | Score / 100 |
|---|---|---|
| 1 | Maison Francis Kurkdjian | 70.8 |
| 2 | Hermès | 70.3 |
| 3 | Tom Ford | 69.5 |
| 4 | Dior | 67.0 |
| 5 | Guerlain | 63.6 |
| 6 | Le Labo | 61.0 |
| 7 | Chanel | 60.3 |
| 8 | YSL Beauté | 58.4 |
| 9 | Diptyque | 46.3 |
| 10 | Byredo | 43.9 |
Cultural Resonance
AI is asked about icons, references, place in culture — how iconic does it sound?
| # | Maison | Score / 100 |
|---|---|---|
| 1 | Dior | 70.6 |
| 2 | Chanel | 69.6 |
| 3 | Guerlain | 65.5 |
| 4 | YSL Beauté | 59.8 |
| 5 | Byredo | 56.4 |
| 6 | Tom Ford | 56.2 |
| 7 | Maison Francis Kurkdjian | 51.5 |
| 8 | Hermès | 46.9 |
| 9 | Diptyque | 45.6 |
| 10 | Le Labo | 38.8 |
Inline note: gaps under one point are within run-to-run noise. Brand Knowledge: Dior 89.3 and Chanel 89.2 are a tie.
What the data says.
Chanel: dominant, yet dependent.
Chanel sits at the top of every Presence axis and most of the Desirability ones — a Maison AI clearly knows how to recommend, describe, and place in culture. But on Google AI Overviews, only 16% of its grounding sources are owned, against 34% from third parties (retailers, marketplaces, social-UGC). The signal AI receives about Chanel is mostly authored by everyone except Chanel.
The most desired heritage Maison in AI is also the one that leans most on platforms and social voices it does not control.
The niche that became an icon (Maison Francis Kurkdjian).
Founded in 2009, MFK ranks first in Olfactory Authority (70.8) ahead of houses with a century or more of patrimony such as Hermès, Dior, and Guerlain. It also enters the Cultural Icons quadrant alongside Chanel, Dior, and Guerlain. In the eyes of AI, a singular olfactory signature carried over seventeen years reads on par with heritage.
In the eyes of AI, seventeen years of singular olfactory signature can stand level with a hundred years of patrimony.
Hermès: revered, and under-surfaced.
Hermès scores 70.3 on Olfactory Authority and 83.5 on Heritage Depth — the third-most-respected fragrance Maison AI knows. But on Product Discovery (34.6) and Gifting Occasion (36.0) it sits in the bottom half. When AI is asked what to buy, what to gift, where to start, Hermès is rarely the answer — even though, when asked what it thinks of Hermès, AI reveres it.
A Maison can be profoundly admired by AI and still rarely recommended by it.
Google's AI Overviews stay silent in French.
Across maison-directed prompts, Google AI Overviews triggered 97.5% of the time in English and 0% in French. The single AI surface most users will see in 2026 still does not, in practice, exist for a French-speaking client asking about a luxury fragrance. The implication for Maisons addressing a French market: the work is still entirely on ChatGPT, Claude, Perplexity, Gemini — Google's AI layer is not yet a battleground there.
English 97.5% trigger versus French 0%, across maison-directed prompts.
Where the models diverge, the divergence is itself a signal.
Standard deviation across the five models tells you how legible a Maison is. Tom Ford reads almost identically across models (sd 2.1) — a stable, predictable narrative, even if not flattering. Hermès (11.8) and Le Labo (11.2) are read very differently depending on the model — a wide standard deviation says the narrative has gaps the models fill differently.
Stability of reading is the leading indicator that a narrative has been authored. Volatility says the gaps are still being filled by guesses.
Who's writing the story AI tells about you.
Google AI Overviews is the only model layer that publishes its grounding sources. We classify every cited domain into three buckets: owned (brand site + verified social), earned (press, encyclopaedic, editorial), third-party (retailers, marketplaces, social-UGC, forums, aggregators).
Third-party share above 50% is critical. Between 30 and 50% is a watchpoint. Below 30% is healthy.
| Maison | Owned · Earned · Third-party | 3rd-party % |
|---|---|---|
| Guerlain | 25% | |
| Diptyque | 23% | |
| YSL Beauté | 22% | |
| Le Labo | 28% | |
| Maison Francis Kurkdjian | 35% | |
| Chanel | 34% | |
| Dior | 17% | |
| Hermès | 24% | |
| Byredo | 32% | |
| Tom Ford | 28% |
Only three Maisons exceed the 30% watch threshold on third-party share: Maison Francis Kurkdjian (35%), Chanel (34%), and Byredo (32%). For the most desired heritage Maison in the index, that is the headline.
Who AI cites first, by territory.
Top-3 most-cited Maisons in each of the six dimensions. The Chanel/Dior/Guerlain core is the structural majority. The third seat rotates: Tom Ford on Discovery, MFK on Gifting, Hermès on Heritage. That rotation is the part to watch.
| Territory | 1st | 2nd | 3rd |
|---|---|---|---|
| Product Discovery | Chanel | Dior | Tom Ford |
| Brand Knowledge | Dior | Chanel | Guerlain |
| Gifting Occasion | Chanel | Dior | Maison Francis Kurkdjian |
| Heritage Depth | Guerlain | Chanel | Hermès |
| Olfactory Authority | Chanel | Dior | Guerlain |
| Cultural Resonance | Chanel | Guerlain | Dior |
What this changes for a digital team.
Desirability is earned, not inherited.
A century of patrimony does not, by itself, translate into AI desirability. Maison Francis Kurkdjian outranks Hermès on Olfactory Authority. What AI rewards is legibility — a singular, repeatedly described signature — more than antiquity.
The oldest house is not the strongest on the map. The most legible one is.
Reverence is not recommendation.
Hermès teaches the lesson. A Maison can be deeply respected by AI and still be rarely surfaced where the question is what to buy, what to gift, where to start. Presence and Desirability are independent — a high score on one does not protect the other.
Source control is narrative control.
When 34% of Chanel's grounding sources are third-party, AI's read of Chanel is shaped 34% by voices Chanel does not author. Consistency without owned anchor is authority a house can lose. The remedy is not more press; it is more legible, owned, on-domain content the models can pull from directly.
Consistency without an owned anchor is authority a house can lose.
Divergence is the opening.
Where models diverge most on a Maison, the Maison has the most upside — the room to author the reading is widest. Low standard deviation means the narrative is set. High standard deviation means it isn't yet, and whoever writes the next clear narrative wins.
How the index is built.
- Scope
- Ten luxury fragrance Maisons: six mega-brands (Chanel, Dior, Guerlain, Hermès, YSL Beauté, Tom Ford Beauty) and four niche houses (Diptyque, Le Labo, Maison Francis Kurkdjian, Byredo). The first edition is scoped to fragrance because it is the category where AI-mediated discovery is already mainstream and where the brand vocabulary is dense enough to measure narrative.
- Models
- Five engines, each on a fixed version: Claude (sonnet-4-6), ChatGPT (gpt-5), Perplexity (sonar-pro), Gemini (2.5-pro), Google AI Overviews. Engines are run at temperature 0 where exposed and at default otherwise. Each model receives every prompt; responses are stored verbatim.
- Prompts
- 120 prompts (twenty per dimension), authored in English and translated to French — twenty are maison-specific, the rest are brand-agnostic to measure spontaneous surfacing. Generation is algorithmic so the set is reproducible. The full prompt list is in the companion report.
- Scoring
- Four signals are extracted from each response: presence (mention), prominence (first-mention position), tone (valence), depth (richness). The three Presence dimensions weight presence 0.40, prominence 0.30, tone 0.15, depth 0.15. The three Desirability dimensions read the same four plus a category layer (ingredient & perfumer vocabulary, cultural references). Each Maison's score on a dimension is the average across the five models.
- Quadrant dividers
- The map's dividers are the sector medians on each axis (Presence 58.5, Cultural Desirability 55.0) — they describe the centre of the current population, not an absolute benchmark. As more Maisons are added the medians shift, which is the point: the map is relative.
- Source authority
- Computed on Google AI Overviews only — the sole model layer that publishes its grounding sources. Owned = brand domains and verified official social. Earned = press, editorial, encyclopaedic. Third-party = retailers, marketplaces, social-UGC, forums, aggregators. Each cited domain is classified once and the share is computed across all triggered AI Overviews for the Maison.
- Factual accuracy review
- We sampled responses for factual contradictions against authored sources (the Maison's own site, encyclopaedic references, primary press). Clear contradictions are rare (~1%) and below the threshold at which they would move headline findings.
- Limits, stated plainly
- Single annotation pass at temperature 0 — the same response would not produce exactly the same score on a re-run, but the variance is below the dimension-level gaps that drive the rankings. Model versions are listed; engines evolve. Prompt generation is algorithmic and may under-sample edge cases. Category vocabulary reflects our classification, not an industry consensus. None of these limits move the headline findings; they delimit the precision claim.
First edition · June 2026. Built to become the reference for how AI values luxury. We update it as the models evolve.