Avisible measures how brands appear when people ask AI: which brands get named, how they are described, and where buyers are sent. Make X Media built all of it, from the measuring platform to the agent system that does much of the building.
What Avisible does
Avisible asks ChatGPT, Gemini and Google’s AI search the questions real buyers ask. It then shows which brands get named, how they are described, and where buyers are sent.
The report follows a buyer through three stages: is the brand named when buyers ask for options, how do answers about it speak of it, and do ready-to-buy answers point to its own site.
What MXM built
- The measuring platform. It captures AI answers and turns them into a report for each brand and market.
- The consulting toolkit behind the first client reports.
- The website, rebuilt in August 2026, and the content on it: 27 articles in English and 18 in Finnish.
- The agent system that does much of the building, described below.
- The working processes: 14 written records of why key decisions were made, 15 step-by-step guides for running the system, 5 reusable templates for setting up its technical environment, and 630 files of automatic tests.
How the work runs
A small system of AI agents, running on Claude, does much of the building. Bob, the lead agent for the platform, plans each task and hands it to builder agents. Jane is the lead agent for the website.
Every piece of work then goes to a separate reviewer agent that did not write it. Four automatic checks have to pass before anything goes live, and automatic scans for accidentally exposed passwords and keys, plus security tests, run on every change.
In all there are seven defined agent roles and dozens of reusable skills. In October 2026, the first full month under this setup, 522 pieces of work were finished and merged. In September there were 72.
In numbers
- AI answers collected in the production database
- 127,000+
- finished, reviewed pieces of work in the platform’s main code
- 703
- planned tasks finished on the build board
- 443 of 790
- pieces of work finished in October 2026, against 72 in September
- 522
- articles on the website: 27 in English and 18 in Finnish
- 45
- of code shared by every page when the rebuilt website launched in August 2026, down from 1.11 MB
- 106 kB
Checked on 30 October 2026.
The timeline
- 20 November 2025: first code of the consulting toolkit.
- 12 May 2026: first code of the new measuring platform.
- August 2026: the website rebuilt.
Clients
Avisible publishes one client case study: an AI visibility audit for Naisten Linja, a Finnish organisation that supports women experiencing violence. Read it on avisible.co.
Avisible is open by invitation. Visit avisible.co.