AI visibility platform: measure what AI engines say about your brand

Refractia answers three questions about your brand: whether AI names you when someone asks about your category without mentioning you, how it describes you when it does, and where it gets what it says. Everything is measured against the competitors you declare, in your market’s language and country.

It works for challenger brands and SMBs competing against much larger names, and for enterprises with several product lines and markets. The model is a platform plus a strategy team. Brands that work with Refractia include Pan American Energy, AXION energy, Bancor, Grido, Sendbox, Universal, Bando, Titan Push and Castrol.

Four of the engines are chat assistants that answer from what the model knows. One is AI search, which summarizes what ranks on the web. Refractia reports the two types separately.

1. 01

   Identity & DNA

   Refractia reads your website and builds the brand profile: category, market, the traits you want to be known for and suggested competitors, all of which you can edit.

2. 02

   Questions in your market’s language

   The Prompts Library generates category, comparison and purchase-intent questions that name the country in full. For a US brand that means “What’s the best mattress brand in the United States?”, not “What’s the best mattress brand?”. For a fuel brand in Argentina, the questions are asked in Spanish: “¿Cuál es la mejor nafta en Argentina?”

3. 03

   Five engines

   The questions go to all five engines. Google is queried with the country’s location, so you see what someone in that market sees.

4. 04

   Metrics

   Mentions, share of voice, sentiment, position and cited sources, against each competitor and engine by engine.

5. 05

   Strategy

   The Strategy Room builds a plan, GEO Optimization reviews your pages and the Content Workshop turns the plan into tracked tasks.

A person starts each scan, someone on your team or ours. Refractia is a deep diagnostic, not an always-on brand monitoring dashboard that refreshes itself every day. You see change by comparing scans in Evolution.

In the app, the modules are grouped into three areas: your brand, competitive intelligence and strategic actions.

The platform also includes the Command Center (a one-screen summary of a scan, exportable as a PDF), the Prompts Library (every question with each engine’s answer) and Demand (which topics people ask about and whether your site covers them).

Brand size doesn’t change how the measurement works; it changes how much structure there is to measure. For enterprises, agencies and groups, the platform adds:

• Product lines, for example a consumer-electronics brand with phones and air conditioners measured as two arenas.
• Topics, with your position against each rival in each one.
• Multiple countries, each market measured separately and in its own language.
• Client accounts that hold several brands, with roles (client admin, viewer) and per-client limits set by Refractia.
• Reports in PDF (Command Center, partial report, group report) and Excel export.
• Public figures mode, which switches to a reputation framing for leaders, candidates and executives, including a War Room.

An SMB with one product line and one market needs none of that to start. It needs to know whether AI names it, how AI describes it and which sources carry weight. Plans are quoted, with no public prices: FUNDACIONAL for challenger brands and SMBs, CRECIMIENTO for scaleups and ENTERPRISE for large brands. See solutions by type of brand.

• k of n. Every rate shows its sample. 25% can be 1 of 4 or 250 of 1,000, and those are not the same claim.
• Not measured is not zero. If an engine didn’t answer, that question doesn’t count as “answered and didn’t name you.”
• Chat and search reported separately. What the model knows and what the web ranks are shown apart.
• The country in every question, in the market’s language, so you don’t measure the wrong market.
• Graded evidence. Every technical check carries grade A, B or C; a convention without evidence (C) is shown as an observation and carries no weight in the score.