Track your brand in ChatGPT
Volexi asks your tracked questions in ChatGPT on a schedule, stores every answer word for word, and records which brands were named, in what order, and which URLs were cited. Collection runs through the official OpenAI API, so the numbers do not break when the chat interface changes.
How does ChatGPT decide which brands to name?
It expands your question into sub-queries, retrieves candidate sources for each, and writes one answer from what survived. The brands named are the ones present in that retrieved material.
This is why a page can rank first in Google and still not be cited by ChatGPT for the same query. Ranking and retrieval are related but not the same mechanism, and the retrieved set frequently leans on third-party comparisons rather than vendor pages.
Why not just ask ChatGPT yourself?
Because one answer is one sample. Generation is probabilistic, so the same question asked twice in an afternoon can name different brands in a different order.
A visibility number only means something as a rate across many prompts, repeated over time. Checking once and concluding you have a problem, or that you do not, is reading a single draw from a wide distribution.
What Volexi records for every ChatGPT answer
Stored word for word, so any number on your dashboard can be traced back to the text it came from.
Yours and your tracked competitors, with the rank of the first mention, because being named first differs from being named fifth.
Classified by source class, content type, format and angle, which is what makes it possible to say why a competitor won rather than only that they did.
Being named is not automatically good. Mentions are scored positive, neutral or negative.
Common questions
No. Collection runs through the official OpenAI API. Scraped chat interfaces break whenever the product changes, and the resulting gaps look like visibility drops.
Yes. Every answer is stored verbatim and linked from the metrics computed on it.
Yes, and the weight is published rather than hidden. See the metric formulas page for the exact figure.
Every plan includes every engine at one price, with six enabled per scan. See pricing or the published formulas for how each engine is weighted.