For most of my career the job was easy to describe, even when it was hard to do. Get the Maison found. First on Google, then on every screen a client touched before they walked into a boutique. You learned the plumbing, you fed the machine the right things, and the brand showed up.
That job has changed. People don’t only search anymore. They ask. They open ChatGPT or Gemini and type “what’s a good perfume for my mother” or “is Le Labo worth it,” and what comes back isn’t ten blue links. It’s an answer, written in full sentences, with a point of view. The machine stopped listing. It started forming opinions.
So I wanted to know what those opinions actually are. Not whether a Maison gets mentioned. We already have tools for that. I wanted the harder thing: when AI talks about a Maison, does it make that house sound desirable? Worth wanting? That is a different question from visibility, and nobody was measuring it.
That is what The AI Desirability Index is.
The setup, quickly. I took ten fragrance Maisons, six mega-brands and four niche names, and asked five models the questions a real client asks. Which one should I buy. What makes a good gift. Tell me about this house. I ran it in English and in French, across ChatGPT, Claude, Gemini, Perplexity and Google’s AI Overviews. That came to 2,400 answers, each scored on two things I kept deliberately apart: how present a Maison is in AI, and how desirable AI makes it sound.
I expected the data to mostly confirm what the industry already believes. It didn’t.
Age doesn’t buy desirability
The most striking name in the study is Maison Francis Kurkdjian. The house was founded in 2009. In the eyes of the models it sits level with Maisons carrying a century of patrimony. Heritage, it turns out, is not something AI reads off a founding date. It reads it off how clearly and how often a Maison’s story is told in the places these models learn from. A young Maison that is legible beats an old one that is quiet.
That is good news if you run a younger brand. It is a warning if you have been coasting on history.
Admired isn’t recommended
Look at Hermès. On pure reverence it scores about as high as anyone. The models plainly hold it in enormous regard. Then you ask the question that comes right before a purchase, what should I buy, what makes a good gift, and Hermès slips out of the frame. It is admired and under-recommended at the same time.
This is the part teams get wrong most often. We assume that if a Maison is loved, it will be suggested. AI treats those as two separate functions. You can win the first and lose the second, and most people aren’t even watching the second.
You don’t own your story if you don’t own your sources
Chanel is the most cited Maison in the entire study. It is everywhere the models look. But check where that information comes from and roughly a third of it sits on third-party platforms rather than anything Chanel controls. The most visible Maison in luxury fragrance is, in AI terms, also one of the more dependent ones. Its authority is real. A good part of it is borrowed, and borrowed authority can be taken back.
Source control was the lever almost nobody is pulling. The Maisons that own the ground AI stands on when it describes them are in a much stronger spot than the ones letting everyone else do the narrating.
In French, the machine goes quiet
In French, Google’s AI Overviews barely trigger. The layer closest to mass search, for a large part of the luxury market, is not showing up yet. It won’t stay that way. Right now it’s open ground.
What I would actually do with this
If you run digital or client at a Maison, the practical reading is short.
Stop treating your AI presence as a single number. Presence and desirability move on their own, and you need them on separate lines.
Find out where the models get their information about you, and start owning more of it.
And measure the distance between how admired you are and how often you get recommended, because that gap is where the money quietly leaks out.
A word of transparency. Diptyque, where I ran digital and e-commerce, is one of the ten Maisons in the study. It is measured exactly like the others, and the methodology says so on the record. Better you hear that from me than spot it yourself.
The full index is out now. Read it the way you would read a competitive teardown, not a league table. The number next to your name matters less than the shape of your two scores and the distance between them.
AI is going to keep forming opinions about your Maison whether you take part or not. The only real choice is whether you have a hand in what it learns.
The AI Desirability Index · Fragrance is out now.