How we read your house
Wealthy clients now ask AI which house to trust. The engine gives one answer, not ten links. We find out what that answer says about you, prove it, and help you change it.
In one sentence
We ask the AI engines what your clients ask, keep every answer as sealed evidence, score what each answer does to your standing, and give you the plan to fix it.
The work, in six steps
- 1
We ask what your clients ask
Five questions a wealthy family types before choosing a house. Who is best, what does it cost, who is behind it, who else should we consider, is there any risk.
- 2
We ask every engine
The same questions go to ChatGPT, Gemini, Claude and Perplexity, with web search on, as your clients use them. We repeat them across days and markets.
- 3
We seal every answer
Each answer is stored word for word, stamped with its time and a unique fingerprint. Nothing can be edited afterwards, by us or anyone.
- 4
Two judges read each answer
Two independent AI judges decide what the answer does to your house. When they disagree, we keep the more severe verdict.
- 5
One score, one tier
The verdicts roll up into your Narrative Visibility Score, from 0.00 to 1.00, weighted by how many people use each engine, and a tier: Stable, Monitor, At-risk or Critical.
- 6
One plan to fix it
Each finding becomes a clear action. Under a Mandate, our agents do the work and we measure the effect the following month.
Weighted by what your clients use
Not every engine counts the same. In Europe, about four in five AI-assistant conversations happen on ChatGPT, then Gemini, Copilot, Perplexity and Claude. Your score follows those shares, so a poor answer on ChatGPT weighs far more than one anywhere else.
- ChatGPT79.3%The assistant in most European pockets. A weak answer here costs the most.
- Gemini10.8%Built into Android phones, Chrome and Google’s own apps.
- Copilot3.6%Preinstalled on Windows and in office browsers. Added to our readings next.
- Perplexity3.5%The research engine of professionals and academics.
- Claude2.8%Favoured by advisers and analysts.
Share of the European chatbot market, October 2026, reviewed every month. Copilot is not in our readings yet, so today’s score counts the four engines we read. Qwen, for families in Asia, is read on request and carries no weight in Europe.
The five questions
- Q1DiscoveryWhich firms would you recommend? Asked three ways: in Switzerland, in your city, by a family abroad.
- Q2ValueAre this house’s fees transparent, and is it worth what it charges?
- Q3ProvenanceWho is behind this house, and what is its record?
- Q4AlternativesAre there better alternatives to this house?
- Q5RiskAre there risks or controversies to know about?
The five verdicts
- RecommendedYour house is named and endorsed.
- NamedYour house is named, neutrally.
- DiscountedYour house is named, but doubted, misdescribed or made to look expensive.
- SubstitutedThe engine sends the client to a competitor instead.
- AbsentYour house is not mentioned at all.
What your house gets
You see what clients see
The exact words an engine gives a prospect about your house, today.
You can prove it
Every finding traces back to a sealed original. It holds up in a partner meeting.
You know what to do next
A short, ranked action plan, each step tied to the score it should move.
You see it move
The same score every month, so you know whether the work is working.
You fix what matters first
Scores are weighted by how many people use each engine, so the plan starts where most of your clients are.
How we improve your standing
A reading is only useful if it changes something. Under a Mandate, our agents do this work and we show its effect in the next monthly reading.
Correct the facts engines rely on
Engines repeat what they find in registries, directories, press and your own site. We find the sources behind each wrong answer and correct them.
Make your house easy to read
We restructure the facts on your site, such as ownership, fees, regulator and team, so an engine can quote them without guessing.
Separate you from your namesakes
When an engine confuses you with another company of the same name, we anchor your identity in the sources it trusts.
Earn the citations that count
We identify the few publications and rankings engines cite for your category, and prepare the material that earns your place there.
Draft, never publish alone
Agents prepare every text. A partner of your choice approves it. Nothing goes out in your name without that approval.
How we tune the agents to your house
No two houses are read the same way. The agents start from Method 1.2, then we fit them to you.
Your questions, not generic ones
In a private briefing we learn how your families actually choose. Their real questions are added to the five core ones.
Your peers and your markets
We set the competitors you lose mandates to, and the markets and languages your clients ask in: Geneva, Zurich, London, Dubai.
Your facts on file
We record your true fees, ownership and record. The judges then catch every answer that misstates them.
A monthly calibration
Each month we re-read a sample of verdicts by hand, adjust the judges, and add the questions your industry is starting to ask.
Fresh questions, same yardstick
Each of the five questions is held in several wordings that rotate, and each answer records the wording it got, so scores stay comparable. Every week we scan Swiss finance headlines for new themes and propose new wordings; a partner approves each one before it is used.
Kept current as engines change
Engines release new versions often. Each month we check the versions your clients use, update the usage weights and rerun the same questions, so every month compares like with like.
Your house, your industry, every conversation
Each client receives a personal dashboard, built around their own peers and markets. It is updated with every monthly reading, and you can explore every answer behind it.
Your standing
Your score and tier per engine, month by month, with the reason for every move.
The conversations
Every question and answer, word for word, filterable by engine, market, language and verdict.
Your industry
How your peers read on the same questions, anonymised, so you see where the category is going.
The sources
The websites and publications each engine cites about you, and which of them hurt you.
The plan
Each action, the finding it answers, who approved it, and the effect it had.
Discretion first
Your industry runs on trust and introductions. So does the way we work.
By introduction
We take on a small number of houses each season, mostly through people who know us.
Under confidentiality
A confidentiality agreement is signed before any reading of your house is shared.
Never named in public
Public benchmarks show House A, House B. Your name appears only in your own private reading.
Shared with whom you choose
Your dashboard and Dossier are seen only by the partners you invite.
Why you can trust the reading
We take no sponsorships, have no engine partnerships and sell no placement. Findings are sealed before we contact any house. Google AI Overviews is not read, because it has no official channel. The readings themselves are run by AI agents, and eyeoverai is built and operated on NanoCorp; every verdict is still checked against its sealed original.
Read the white paperSee the benchmarkA private reading of your house
We read your house on every engine and show you the answers families receive today. A confidentiality agreement comes first. We reply within one working day.