AIVIS's Brand Perception Test does one simple but often overlooked thing: it actually runs the prompt you write against configured AI engines (no simulation, no estimate) and shows you the real answer, with cited sources when the engine exposes them.
How to read the result
- Mentioned or not: the simplest data point, but not the only one that matters.
- Position in the answer: are you the first option or the third after two competitors? Position often weighs as much as — sometimes more than — the mention itself.
- Sentiment of the text: is the tone the AI uses positive, neutral or negative? An engine can mention you and still describe you unflatteringly if the sources it read suggest that.
- Cited sources: when the engine exposes them (Perplexity and Gemini do this natively more often than ChatGPT), check whether they cite your own site or a third party talking about you — very different situations.
What a single result does NOT mean
AI answers aren't deterministic: the same prompt run twice can return different results. A single test is a useful snapshot to start with, not a stable measurement — for that you need prompt tracking over time, available to registered users.
A practical tip on which prompts to test
Avoid prompts that already name your brand ("what do you think of {brand}?"): that's not a discovery test. Use prompts a real potential customer would ask without already knowing you ("best {category} in {city}") — it's the only way to know if you'd be discovered organically.