Method

How we measure

How we measure

Every week we ask the AI tools your prospective clients use the questions those clients actually ask, and record which firms are recommended. This page sets out exactly how, so you can check it.

The engines

The engines

We track four: ChatGPT (with web search), Google Gemini (with Google Search), Perplexity, and Google’s AI Overviews. Each is queried through its official API, with web search switched on, as a client would use it.

The questions

The questions

120 questions across six categories, 20 in each: property buying agents, household staff agencies, education and school placement consultants, private GPs, family and private client law, and wealth management. Each is written the way a client would ask it, and none names a firm. The wording is fixed. If a question ever changes, it becomes a new question and the old one is retired, so results stay comparable week to week.

How often, and from where

How often, and from where

Every Monday, each question goes to each engine twice. We store the full answer and every source it cites. Where an engine lets us set a location, the request is set to London, United Kingdom, and we record the location sent with every run.

What counts as being named

What counts as being named

A firm counts as named when an answer recommends it by name. We match names against a list of around 950 UK private client firms, including the different ways each one is written. A firm that only appears as a cited source is recorded, but not counted as a recommendation.

How we score

How we score

Share of answers: of all the answers in a category over the last four weeks, the percentage that name a firm. Using a rolling four weeks smooths out the normal variation between runs.

The limits

The limits

AI answers vary from one run to the next, so small differences between firms are not meaningful. The engines change their models without notice, which can move results for reasons unrelated to any firm. Our questions are a fixed sample, not every question clients ask. We measure what the engines say, not why they say it.

Independence

Independence

Firms we work with, and firms we are in discussion with, are marked wherever results are published. That status is never used in scoring, and no one can pay to change the questions or the results.

Find out what AI tells your clients about you.

Find out what AI tells your clients about you.

Find out what AI tells your clients about you.

Subscribe to the Briefing

Your information is kept confidential

Recommendable

Subscribe to the Briefing

Your information is kept confidential

Recommendable

Subscribe to the Briefing

Your information is kept confidential

Recommendable