How IVIA measures
Each measurement is one question put to one AI engine, with live search, and the answer it gave. A typical run is 45 measurements: fifteen questions, three engines. The index is what comes out of adding them up.
Four steps. The first takes minutes; the rest run on their own.
- 01
Describe the brand
Sector, services, competitors, markets, domain. This is where the questions come from — questions a real customer would ask, not keywords.
- 02
Audit the site
Before asking, we check whether the crawlers can read it: robots.txt, structured data, FAQ, titles. There is no point optimising the answer if the door is shut.
- 03
Ask the three engines
Every question goes to ChatGPT, Perplexity and Gemini with live search. The literal answer is stored, and that is the evidence.
- 04
Get the index and the report
A 0–100 number, who takes the spot, which sources were cited and what to fix first. As a PDF, under your brand.
How scoring works
Each answer is worth zero to three points, depending on whether the brand comes first, is listed among the options, is merely mentioned in passing, or does not appear at all. IVIA is the percentage of points earned against points available. Appearing in every answer but never first gives a very different number from appearing in half and leading — and that difference is what matters to whoever has to decide where to work.
What IVIA does and the international platforms don't.
Technical audit included
They all measure whether a brand gets cited. None checks whether the site can be read by the crawlers feeding those answers. A site blocking GPTBot will never be cited — and you find that out by looking at the site, not at the citations.
Built for European markets
Questions, brand detection and reports are built for European markets and languages, not adapted from a US template. Measuring a European market with an American tool means measuring it with the wrong questions.
Real white-label
Reports carry the agency's logo, colours and domain. The end client receives the agency's document — not ours with a logo in the corner.
The literal answer is kept
Every measurement stores the exact text the engine returned. When a client asks “how do you know that?”, the answer is to open it and show them — not a number nobody can check.