Features

Seven things we measure, and nothing we cannot

Every assistant answers your buyers' questions differently, and every one of them is generating that answer fresh. These are the measurements that tell you where you stand and what to change — each tracked continuously across all six engines, on every plan.

Start 7-day free trialRead the methodology

Card required, nothing charged for 7 days

What the product does

All of them are included on every plan. Plans differ only in how many questions and brands you track.

Paste your domain and we propose the whole set

The hard part of a tool like this is the blank page: which competitors, which questions. You should not have to guess at either.

  1. 1

    Give us the domain

    Nothing else. No tag to install, no analytics to connect, no call to book.

  2. 2

    We read your site

    And write back the category a buyer would use for you, in their words rather than yours.

  3. 3

    We propose the set

    Up to six competitors you genuinely lose deals to, and eight to twelve questions a buyer asks before they know your name.

  4. 4

    You edit and confirm

    Every suggestion is a suggestion. Nothing is tracked until you say so, and you can change all of it later.

Each proposed question arrives with its real monthly search volume beside it, so you approve a set sorted by how many people ask it rather than by how plausible it sounded in a meeting. That is prompt tracking, and it is the choice every other number on this page depends on.

What we deliberately do not sell

This list is shorter than our competitors' feature pages on purpose.

We enumerated every feature page published by the four biggest tools in this category and found thirty-seven distinct concepts, only two of which appeared on more than one rival. There is no settled vocabulary here yet, so a long feature list is mostly a vendor naming its own internals. We would rather name a handful of things you can verify. The methodology page sets out how each one is collected, and what we refuse to infer from it.

  • No technical site audits. We do not check whether your pages are crawlable or your schema is valid. Plenty of tools do that well already.
  • No content generation. We tell you where you are absent. Writing the page that fixes it is your job, and we are not going to pretend a model can do it unattended.
  • No historical backfill. Measurement starts the day you start. Nobody can retrieve what ChatGPT said about you last March, and any tool that offers to is guessing.
  • No blended multi-market number. You choose the market; a workspace measures that one. Two countries averaged into a single figure describes neither of them.

Once it is running, six things we watch for

Named, because 'we will email you when something changes' tells you nothing about whether it is worth switching on.

  • Visibility drop

    Your rate fell against its recent baseline, on one engine or across all of them.

  • Visibility rise

    It went up. Worth knowing why while you can still find out.

  • Engine blackout

    One engine stopped naming you altogether while the others carried on.

  • Competitor overtook

    A rival passed you on a question you had been holding.

  • Cited, not named

    An answer used your page as a source and put someone else in the sentence.

  • Negative framing

    How you are described shifted, not merely whether you were mentioned.

Each finding is recorded once and emailed once, deduplicated by a database constraint rather than a best-effort check, so a drop that persists for three weeks is one email and not twenty-one. How monitoring works.

Common questions about what we measure

The things people check before they trust a number.

What does AI Mention Tracker actually measure?
Seven things, and only seven. Which prompts your buyers ask, how often each assistant names you, what share of the answers you hold against the competitors beside you, how you are described when named, which pages get cited, what changed since yesterday, and the full text of every answer behind all of it. Everything here is something the product does today.
Which AI assistants do you check?
Six: ChatGPT, Google AI Overviews, Google AI Mode, Perplexity, Claude and Gemini. Every plan gets all six. We do not hold an engine back for a higher tier, and we do not claim engines we cannot collect from.
Do I need a different plan for different features?
No. Every feature on this page is on every plan, and so are the allowances that usually get metered elsewhere: 25 competitors in each brand and 25 teammates, on the cheapest plan as much as the dearest. Plans differ only in how many questions you track and how many brands you spread them across, because those are the two things that cost us money to run.
How often is everything refreshed?
Continuously. Five engines run six days a week and Claude twice, on every question in your set, asked from the market your workspace is set to. Assistants generate a fresh answer each time rather than reading from an index, so a single check is close to meaningless. Every number we show is a rate across repeated checks with the number of checks behind it.
What do you deliberately not do?
We do not audit your site's crawlability, write content for you, track shopping or ad placements, analyse AI agent traffic to your servers, or backfill history from before you signed up. Several of our competitors sell those. We would rather have six pages that are true than thirty that are not.

See where you stand on every one

Add your brand and the questions your buyers ask, and get the first day of data in a few minutes. Card required, nothing charged for 7 days.

Start 7-day free trialSee pricing