How we measure AI visibility

The Placement Audit asks 50 fixed questions in three AI engines and analyses every answer by the same rules. Below are the questions, the analysis, the metrics and the limits of the measurement.

The question library

The basis is a fixed library of 50 questions. They are written from the perspective of five audiences: fund selector, independent wealth manager, institutional investor, family office and retail investor. The questions cover four phases: discover, compare, verify and brand check.

40 questions do not mention your firm. They show whether the AI names your firm on its own when someone looks for managers in your strategy. The 10 brand-check questions name your firm and show what the AI says about you.

Questions we also derive from your website are shown separately in the report. They are not part of the metrics, so the metrics stay comparable across firms.

The questions are templates into which we insert your strategy, your vehicle and your region. For example, “Which European asset managers are strong in {strategy}?” becomes:

“Which European asset managers are strong in direct lending?”

Engines and queries

In the audit we ask every question in three AI engines of your choice, by default ChatGPT, Perplexity and Google AI Overviews. Each question runs once per engine, with web search switched on. We query ChatGPT and Perplexity through the providers' interfaces and read Google AI Overviews from the Google results page.

Analysis

An analysis model reads every answer and records which providers appear in which order, whether your firm is among them, how positively or negatively it is described and which statements are made about your firm.

If the model fails, we search for your name with a text search. Those answers count towards the mention rate, but not towards position, share of voice or the fact check.

The metrics in the audit

Mention rate: Share of answers to the open questions that mention your firm. Open means the question does not name your firm.
Position: How far up your firm appears when an answer lists several providers. An early place counts for more than a late one.
Answers linking to your website: Share of answers to the open questions that link to your own website at least once.
Placement Score: A score from 0 to 100 based on mention rate, position and the share of answers linking to your website, calculated over the answers to the open questions. The weighting is the same for every firm.
Share of voice: Your share of all mentions of your firm and your peers. Each firm counts at most once per answer. Peers are the competitors you name. If you name none, we take the five firms that appear most often in the answers.
Engine × audience matrix: The mention rate split by engine and audience. It shows, for example, whether your firm is missing from one engine's answers to family office questions.
Factual accuracy: We check every statement about your firm in the brand-check answers against your fact sheet, for example on AUM, team, funds and track record. Each statement counts as correct, wrong or unclear. It is unclear if the fact sheet contains nothing on the point.
Sentiment: How positively, neutrally or negatively the answers describe your firm.

Sources and technology

For the source diagnosis we assign every cited website to a type, such as trade media, database or directory, and fetch up to 30 cited pages once. A text search shows whether your firm and your competitors appear there. This is where the levers come from.

The quick technical check reads robots.txt, llms.txt, the sitemap, the home page and up to five subpages. It shows, for example, whether AI crawlers are blocked or key facts sit only in PDFs.

Limits of the measurement

Each question runs once per engine. Language models do not always answer the same way, and the same question can get a different answer the next day. The audit therefore shows the state at one point in time.
During the optimisation we measure every month. That shows whether figures move for good or only fluctuate.
What a user sees in their app can differ, for example because of their history, their location or a different model version.
Nobody controls what a language model says. Placed by AI guarantees measurement, action and transparency, not placement.

Fund Placement Index

The index is in preparation. It is meant to measure the asset managers and private markets firms in the DACH region every quarter with the same question panel, split by asset class. More about the index

Request a Placement Audit

Questions about the methodology? hello@netboosting.de