AI search visibility · open beta

See where your brand appears in AI answers.

Track mentions and citations in ChatGPT and Gemini, see which competitors get recommended instead, and turn every gap into the next article, category or landing page to create.

Free during the beta · no card · ChatGPT and Gemini available now

The assistants your buyers actually ask

  • ChatGPTavailable
  • Geminiavailable
  • Claudecoming next
  • Perplexitycoming next

Available means validated end to end: request contract, response normalisation and citation handling. Claude and Perplexity arrive at no change in price.

What it measures

Everything you need to win the answer, not just the ranking

Six capabilities, and every one of them is something Rankmat does today.

Tracked prompts

The questions your buyers actually ask an assistant, asked on a schedule.

  • Write your own or import a set in bulk
  • Group by topic, intent and funnel stage
  • Every prompt keeps its own history

AI visibility over time

Whether you are named, and which way that is moving.

  • Mention coverage, share of voice and prominence
  • Only measured days are drawn — no invented points
  • Compare any period against the one before it

Citations and sources

Assistants answer from sources. These are the ones deciding your category.

  • Every page an answer cited, and whose it is
  • Your own domain separated from everyone else's
  • Retrieved results never counted as citations

Competitors

Who the models recommend when they do not recommend you.

  • Share of voice counted over the same answers
  • Per prompt, per platform, side by side
  • Track up to fifteen rivals per project

Opportunities

Each gap turned into one specific action, with the evidence behind it.

  • Create an article, category, landing page or comparison
  • Optimise a page, or improve citation coverage
  • Every recommendation links to the answers it came from

Content ideas

The work grouped the way you would plan it.

  • Articles, categories and landing pages to create
  • Existing pages to optimise
  • Named prompts and the rival currently winning them

The actual product

Charts you can act on, from answers you can read

These are the components the application renders, drawn here with sample data.

Watch visibility move, honestly

Every point is a measurement that happened. Days nobody measured stay empty rather than being drawn as a straight line through them — which is the single most common way an AI-visibility chart misleads.

Mention coverage over time

Sample data
0%25%50%75%100%4 Jul12 Jul20 Jul28 Jul1 Aug57.1%46.2%

Only the marked dates were measured; the lines join them directly.

  • Meridian Analytics
  • Northbeam Insights
Show the numbers behind this chart
Mention coverage over time, as a table
DateMeridian AnalyticsNorthbeam Insights
4 Jul24.9%61.2%
8 Jul28.2%58.4%
12 Jul26.0%59.9%
16 Jul44.1%55.0%
20 Jul46.0%52.7%
24 Jul41.1%54.1%
28 Jul52.4%49.8%
1 Aug57.1%46.2%

Share of voice, this measurement

Sample data
Meridian Analytics (you)57.1%Northbeam Insights46.2%Quantly38.4%Stackfeed21.7%Toolcompare12.9%

See who is recommended instead of you

Your brand and every tracked rival counted over the same set of answers, so the comparison is arithmetic rather than assertion. Open any bar to read the answers behind it.

How it works

Three steps to your first measurement

  1. 1

    Add your brand and competitors

    Your domain, the rivals you want measured against, and the country and language you sell in. About a minute.

  2. 2

    Choose prompts and platforms

    Track the questions that matter to you, on ChatGPT and Gemini. Nothing runs until you say so.

  3. 3

    Measure, then act

    Rankmat asks the models, records every answer, and turns the gaps into work you can hand to a writer.

Why the numbers hold up

Every figure opens onto the answer it came from

An AI-visibility number nobody can check is a number nobody should act on. Each one here walks back to the verbatim response, the brands found in it and the pages it cited.

  • Verbatim answers

    The model's own words, stored and readable — not a summary of them.

  • Cited, not retrieved

    A page the answer cited is counted. A page merely retrieved is not, and is never folded into a citation figure.

  • Repetitions shown

    Ask a prompt five times and the spread is reported, so you can tell a real drop from an inconsistent model.

  • Absences explained

    A zero says the models never named you. A blank says nobody has measured yet. The two are never the same.

Pricing

Priced on what you track

No per-scan charges and no provider bill to reason about. Free while we are in beta.

Starter

One brand, weekly tracking, the full evidence trail.

€49/ month

billed monthly

Projects
1
Tracked prompts
50 per project
AI platforms
2
Measurement
Weekly
Join the beta

Free while in beta · no card

Growth

Most chosen

Daily scans across five projects, with impact measurement.

€99/ month

billed monthly

Projects
5
Tracked prompts
250 per project
AI platforms
2
Measurement
Daily
Join the beta

Free while in beta · no card

Agency

Twenty-five clients, shared reporting and client access.

€299/ month

billed monthly

Projects
25
Tracked prompts
1000 per project
AI platforms
2
Measurement
Daily
Join the beta

Free while in beta · no card

Compare every limit

Questions people ask first

Which AI platforms do you measure?
ChatGPT and Gemini today — both with a validated request contract, response normalisation and citation handling. Claude and Perplexity are next; they are not selectable until each has been through the same validation, because a measurement nobody has verified is worse than no measurement.
How is AI visibility calculated?
Eligible answers that name your brand, divided by all eligible answers. Every figure on every page opens onto the answers behind it — the verbatim response, the brands found in it, and the pages it cited.
Do you discover prompts for me?
No. Rankmat measures the prompts you track, and says so wherever a number appears. Discovering the wider set of questions people ask is a different product with a different data source, and claiming it without one would make every denominator on the page a guess.
Can I track more than one country?
Yes — as separate projects. One project is one brand, in one country, in one language, so Decathlon CZ and Decathlon SK are two projects compared side by side. Markets inside a single project are on the roadmap.
What does a project cost to run?
Nothing during the beta, and no card is required. Provider costs are ours to manage, not a number you should have to budget against — you are billed on what you track, never per scan.
Do you need access to my website?
No. Rankmat needs your domain so it can tell your pages from everyone else's in a citation list. There is no crawler, no tag and no analytics integration.

Find out what AI says about you this week.

One project, your prompts, your competitors. A measurement takes about a minute.

Join the beta