Mentionary

How Mentionary measures AI visibility

Mentionary writes the unbranded questions your buyers ask, puts each one to ChatGPT, Gemini and Claude, and counts how often your brand is named. This page explains each step and where the method has limits.

What is measured

The unit of measurement is the buyer question: a question a prospective customer asks an AI assistant while choosing a product, written without your brand name. For each question and each engine, Mentionary records whether your brand is named, which competitors are named and which sources are cited.

How the questions are written

  1. Mentionary reads the homepage of the site you give it and extracts the value proposition, the offerings and the likely buyer personas.
  2. You confirm or edit the value proposition and the personas to track.
  3. It generates questions those personas would ask about your market. Each question belongs to a persona and a category, and you can switch any question on or off.

The number of questions you can track depends on your plan: 10 on Free, 25 on Starter, 100 on Pro and 500 on Agency.

How the engines are asked

Every active question is sent to ChatGPT, Gemini and Claude through each provider’s API with web search enabled, so the answers draw on current pages. One question asked to one engine is one answer check. Google AI Overviews is built and running in pilot on selected workspaces, and Perplexity is next.

What counts as a mention

Each answer is analysed to decide whether your brand is named and which competitors are named. The sources an engine cites are recorded separately.

The numbers you get

  • Mention rate, overall and per engine.
  • Competitors named, by how often each appears in the answers.
  • The sources the engines cite most.
  • A per-question breakdown of where you were and were not named.
  • Prioritised fixes, and content gaps against competitors.

Limits of the method

  • Answers vary. The same question can produce a different answer on another day. A rate over many questions is more stable than any one answer, and repeated runs are more stable still.
  • API answers are not identical to answers in a consumer app, which can add personalisation and memory.
  • Small question sets give coarse rates. On 10 questions, one answer moves the rate by 10 points. Larger plans track more questions for this reason.
  • The questions are generated from your site. If your homepage describes your market poorly, edit the personas and questions before running the analysis.

Repeating the measurement

Pro and Agency re-run the analysis every week on the same questions, which turns single readings into a trend. On Starter you re-run it when you choose. See pricing for the allowances.

Method questions

Mention rate is the number of answers that name your brand divided by the number of buyer questions asked, per engine. If ChatGPT names you in 9 of 24 answers, your mention rate on ChatGPT is 37.5%.

An engine asked about a brand by name will describe that brand. That shows the engine knows you exist, not that it recommends you. Buyers describe a need, so Mentionary measures unbranded questions.

Not exactly. Mentionary asks each provider's model through its API with web search enabled. A consumer app can add personalisation, memory and its own formatting, so an individual answer may differ. This is one reason Mentionary reports a rate across many questions and not a single answer.

Your buyers are already asking AI. Find out who it tells them to pick.

Add Mentionary to Claude and ask one question. Your first analysis across ChatGPT, Gemini and Claude is free, inside your assistant or in the web app.

Add to Claude
Create free account

Free plan: 10 buyer questions, 3 engines, 2 competitors. No card.