AI SHARE OF VOICE
A client asks ChatGPT to name a European asset manager. Your firm is either in that answer or it is not. Here is what the data says actually moves it.
This quarter I asked nine AI engines, 2,243 times, in five languages, to recommend a European asset manager. The exercise runs as the AI Asset-Manager Observatory, and the point of it is one number per firm: its AI share of voice, the share of answers that name it at all. What the readings keep showing is that the firms AI names are not the biggest firms. They are the most retrievable ones. That is good news, because retrievable is something a marketing team can actually work on.
I came into asset management last year from B2C performance marketing, where you measure every step or you do not get paid. Looking at fund distribution through that lens, AI recommendations are just a new acquisition channel that nobody has instrumented yet. So here is the honest playbook for getting named by AI, drawn from what the Observatory data rewards, not from generic advice an AI could give you itself.
Get into the sources the engine reads
The single biggest lever is the engine’s mode. When an engine answers from memory, it names the firms it was trained on, and the European incumbents win: Amundi leads, the household names cluster at the top. When the engine searches the live web first, the ranking reshuffles toward whoever is present in the sources it retrieves that day. In Dutch, the recall answer puts BlackRock ahead of Amundi, 69.2% to 61.5%. The recall advantage that big European names enjoy expires the day the engine goes searching.
For a challenger firm, that flip is the opening. You will not out-remember BlackRock, but you can out-publish it in the specific sources an engine pulls when someone asks about, say, European small-cap or Article 9 funds. Structured, current, specific pages on your own domain and in the third-party listings and directories these engines lean on. This is the mechanism behind why the winner changes when the engine searches the web, and it is the difference between being remembered and being retrieved.
Check you are not blocking the crawlers that feed the answer
This one is unglamorous and it is where I see the most self-inflicted damage. If your robots rules or your CDN quietly block the AI crawlers, you are invisible in grounded mode by your own hand, no matter how good the pages are. I looked at how 105 European finance firms set their AI crawler rules, and the settings are all over the place, often decided by IT with no marketing input. Before any content work, confirm the engines are allowed to read you. It is a ten-minute check that gates everything else.
Stop trusting brand size to carry you
The correlation between a firm’s web brand pull and its AI share of voice is 0.35. Weak. Some of the most recognised names underperform their fame in AI answers, and several specialist firms punch far above theirs. I set this out in full in AI share of voice versus real brand pull. The practical read: a big brand budget does not buy you into the shortlist. Retrievable, well-structured evidence does. That reallocation, from brand spend toward the structural work of being citable, is the hard part for large firms and the whole opening for smaller ones.
Be the source AI quotes, not just a firm it mentions
Being named is the floor. Being the source an engine cites when it explains its answer is the ceiling, and it compounds: a citation feeds the next answer. The firms that get cited publish things worth citing, original data, clear method, a real point of view, in a format a machine can lift a passage from. I track which sources AI cites for asset managers, and it is rarely the glossy brochure page. It is the piece that answers the question directly.
The honest part
None of this is a one-off. AI answers move as models retrain and the live web shifts, so getting named is structural work you run every quarter, not a campaign you ship once. The first step is knowing where you stand today: which engines name you, in which mode, and against which competitors. That is what an AI visibility audit measures, the same method as the Observatory pointed at one firm. The method is public on purpose. Running it rigorously, every quarter, is the part most teams do not have spare hands for.
Notes from a European operator
When a client asks an AI to recommend a manager this quarter, do you know whether your name comes up, and whether it holds when the engine stops remembering and starts searching?
Before the levers, know the board you are trying to enter. The quarterly reading of who the machines name covers 104 managers across nine engines and five languages, and every answer is archived. The other half of the picture is what the firms themselves put on the record: fifteen weeks of public statements from European finance.
Sources and method
- The full study and public method: the AI Asset-Manager Observatory, 2,243 logged answers, five languages, Q3 2026.
- The live numbers: the AI citation engine dashboard.