The AI visibility practicefor luxury houses
Your buyer asks the AI.
Does your name come up?
What watch for my son’s thirtieth. Which jewellery house for a Christmas gift. Your next client puts these questions to ChatGPT, Gemini or Perplexity before putting them to anyone, and whatever the engine answers is the shortlist.
The state of play · AI in the buying journey
The engines were read in . The large names fill the answer. Independent houses are rarely named where the buyer decides. That is the gap the Report measures.
A practice for the house that wants to remain irreplaceable to its clients, and unmistakable when the engines answer.
The Reading
Two pages on how ChatGPT, Gemini, Google’s AI Overview, Perplexity, Google’s AI Mode and Microsoft Copilot describe your house today, evidence attached. Complimentary, and a small number are prepared each month.
01What the reading found, and why now
The answers are hardening.
A pilot ran in on two Swiss houses. Neither was named when a buyer asked what to buy at their own price. Asked about them directly, both engines answered fairly, then advised waiting for the pre-owned market and quoted a discount of 40 to 60% with no source attached. Two houses is not a finding. It is the reason forty-six are now being measured.
The census is running now. The early record is worse than the pilot suggested.
Do the arithmetic on your own house. A maison selling eight hundred watches a year at an average of twenty-five thousand francs turns over twenty million. If one buyer in twenty arrives already told to wait for the pre-owned market, that is a million francs of revenue at full margin, surrendered to a figure with no source behind it. I do not know your number. The engines are quoting one.
And each month the engines settle further on their sources. The reading here is that citation compounds the way search authority did between 2008 and 2015, when the names that took position early held it for a decade.
The house that is the answer in 2026 is likely to be the answer in 2028.
A house’s position in these answers is an asset, and it has begun to behave like one. It compounds over years, it cannot be bought quickly, and most of it sits on pages the house does not own, which is precisely what makes it hard to replicate and hard to lose overnight.
It is also newly visible. Anyone examining a house, a family shareholder, an incoming investor, an acquirer, a distribution partner, can now open an engine and ask what a buyer would ask, and see in ninety seconds whether the house comes up, how it is described, and whether the machine sends the buyer to look for a lower price. That check costs nothing and leaves no trace. A house that has never run it does not know what is being seen.
Goodwill has always rested on what others say about a house when it is not in the room. The room has changed.
The Luxury AI Report™
How AI engines describe the luxury industry
Edition One
Every eligible Swiss house at Geneva Watch Days, measured.
Edition One · Swiss Watches · 46 houses at Geneva Watch Days 2026
The Luxury AI Report is published by ELIA Atelier and measures how AI answer engines describe luxury houses. Edition one, published , records 14,256 answers in the census and 15,330 in total, across 46 Swiss watch houses, nine buyer questions, three engines and three markets.
A census of the houses showing in Geneva this September, none excluded by judgement.
The headline finding
Which brands should I consider? He names no house. He says nothing about resale.
82.5%
of those answers steer him to the second-hand market, or tell him to wait for a lower price
475 of 576 recorded answers · 4 price levels · 3 markets · 3 engines · 16 days · 10 to 25 August 2026
Hong Kong was the intended third market and was withdrawn before sampling, since ChatGPT is not available there. Singapore, the second largest watch market in Asia, took its place.
The five verdicts were named on 20 August 2026, before any figure was computed. The rule set that applies them was fixed on 24 August, versioned and dated, and no verdict is assigned by hand.
Recommended
Put forward by name, in terms that favour the house
Named without endorsement
Present as a matter of record, with no reason to be chosen
Named and discounted
Named, then steered towards pre-owned or a lower price
Substituted
Asked about, and the answer arrives carrying another house
Absent
The question is the house’s own territory, and it is not in the answer
The findings · Edition one
What AI tells wealthy buyers when they want to buy a watch?
Read the findings →
Read the full interview with Europa Star →
The Report publishes at its own address, luxuryaireport.com, with its method and its arithmetic in full.
Measured houses may request their letter before publication.
02What ELIA does
Three rooms. A house may enter through any of them; most enter through the first.
ELIA Atelier publishes the Luxury AI Report, and is the practice that turns its findings into strategy and implementation.
The house’s presence in AI answers, at the register the house is used to.
When a buyer asks the engines for the best in your category, four things decide the outcome:
- Your house is present in the answer.
- What the answer says is correct.
- The answer does not send the buyer looking for it at a lower price.
- Another house does not arrive in the answer in your place.
Most of what decides that answer sits on pages you do not own. An engine that recommends the house pre-owned, or advises waiting for a better price, has answered against you while appearing to answer for you.
That is the difference between a visibility problem and a margin problem, and only one of the two reaches the accounts.
Nobody had asked the engines. The Report asked them 13,365 times.
Every other reading of AI and luxury published this year is an estimate drawn from public signals, and every one of them ranks the same twenty-five names. Nine buyer questions went to three engines across three markets over fifteen days, with a second panel on three further engines, and what came back was recorded for a whole population, none of it excluded by judgement.
The Report is published separately, at its own address, and it sells nothing but its own record. What it gives this practice is the map: which publications, references and platforms decide the answers in a given price band, which of them carry your competitors, and which carry nothing at all. It extends to jewellery, perfumery and other categories on a fixed calendar, so the map widens with each edition.
What follows the measurement. Counsel is standing direction at the principal’s side, one working session each month and a private line answered within the day. Instruments are built on the house’s own archive and owned by the house alone, one at a time.
Before you ask
My clients do not choose watches through ChatGPT.
Some of them do, and the number is not small: 82% of top-tier luxury clients used an AI tool during their most recent purchase, and nearly one in two consulted one before entering a boutique. But the objection misses where the exposure sits. The engine is rarely the place a client decides. It is the place they arrive already informed, having formed a shortlist that may not include the house, and a price expectation the house did not set. Many luxury questions put to these engines name no house at all, and that is the moment a shortlist is built.
The same answers reach people the house never counted as clients: the journalist checking a fact before an interview, the retailer deciding what to stock, the collector’s adviser, the family office assistant, the investor running a first look. None of them will mention that they asked.
The check is free and takes two minutes. Open an engine you have never used, ask what a buyer at your price should consider, and do not type your name. What comes back is what your next client read on Sunday.
Bain & Company with Comité Colbert, June 2026.
My agency already does this.
Some of it, probably. Structured data, clean markup, a site an engine can crawl. The practice does that work too, and the plumbing only gets the house to the floor, because it addresses the part of the answer you control, which is the smaller part. What an engine says about a house is assembled mostly from sources the house does not own, and no amount of work on your own domain reaches them.
Consultants tell me the same thing. Why you?
This practice publishes the Luxury AI Report: forty-six Swiss houses, nine buyer questions, six engines, three markets, fifteen days, and the source map for the category that comes out of it. A consultant brings judgement without measurement, and an agency brings tooling without the industry. Every recommendation made here is traceable to a line in that record, and where the record does not support a recommendation, none is made.
Who does the work?
The reading, the strategy and the writing are done by Alexandre Olive: twenty years inside the luxury industry, thirteen at executive level in high jewellery and watchmaking, ten spent sparring with founders on strategy and positioning. The technical layer is executed by specialists with real years in the discipline, directed from here. The house is not handed to the most junior person available, which is the usual arrangement and the usual result.
How much of it sits outside my own site?
The published research disagrees, and the disagreement is itself the finding. One analysis of twenty-five million cited links puts earned media ahead of everything a house owns. Another, of nearly seven million, reaches the opposite conclusion, because it counts a directory listing as house-managed. Neither measured watchmaking. The Report did, across three markets and fifteen days, and it gives the answer for your category rather than for the internet in general.
What does the work consist of, then?
Establishing the record that contradicts what is wrong, in a form an engine can lift. Reaching the specific pages carrying an outdated claim. Placing substantive material in the sources that decide answers in your price band, which the Report identifies by measurement. Rewriting the house’s own pages so its claims are stated rather than implied, since the engines read the explicit and misread the restrained. And repairing the technical layer where it is broken, because the floor still has to be there. And where an engine is steering buyers to a lower price on a claim that is out of date or unsourced, correcting the record behind that claim is usually the first work done here, because it is the one that reaches the accounts.
Why does the Report matter to me?
Because it shows where the engines look before anything is touched. The practice holds the source map for the category: which publications, references and platforms carry the answers, which carry nothing, and which of your competitors sit inside them. That map is not available anywhere else, and it is what turns guesswork into a brief.
What do you promise?
The work, the measurement of the work, and an honest reading of what changed. Never a position, never a date by which an engine will change its answer. The Report publishes twice a year, so an overclaim would be contradicted by the practice’s own evidence.
The Luxury AI Report publishes on at 07:00 CET.
03Publications
Written from the record.
Research paper 01, second edition, . A working thesis for independent houses and ateliers of ten to one hundred and fifty people.
A note for founders, . How independent houses should read the next fourteen months, and what the engines are already saying about them.
Edition one, Swiss watchmaking. The field record, the arithmetic and the five verdicts, ahead of publication on at 07:00 CET.
04The founder
Tailor made to you, not a commodity.
ELIA Atelier is the AI visibility practice for the luxury industry, founded by Alexandre Olive and based in Genève, Switzerland.
Twenty years inside the luxury industry, thirteen of them at executive level in high jewellery and watchmaking, ten sparring with founders and senior executives across the top end.
ELIA Atelier is one person and a small bench, and it works with a limited number of houses at a time. Every message is read and answered by Alexandre.
05The door
Begin with a written message.
A few lines on your house and where it stands are enough. You receive the Reading, or a straight answer that the work is not for you.