Methodology: how we measure
What we ask, which assistants, how we count and how precise it is.
Questions
We ask the assistants your customers' questions: what they need and where. The full review asks dozens of questions, the free check 5.
In one review each question goes to several AI assistants, 3–5 times to each, on different days, because an assistant answers differently each time.
Assistants
“Named in N of M” counts the answers of:
- ChatGPT
- Perplexity
- DeepSeek
The free check asks ChatGPT and Perplexity. Each question is asked in a fresh conversation, with no history or personal settings.
“Named in N of M”
A question counts when at least one of the assistants named your company or cited your site. N is how many questions did, M how many were asked.
The grade
A short caption under the number: good when you are named in 70% of the questions or more, average from 40%, weak so far below 40%.
The site
Five checks of what the assistants' crawlers can read on your site. A check that cannot be run does not count.
Accuracy
An assistant can answer the same question differently each time, so one check is a preliminary estimate with a margin of about ±15% of the questions.
A particular person's answer may differ: their chat history, region and the day all play a part.
What we do not promise
A place in the answer: who to name and in what order is the assistant's own choice. We measure how often you are named and remove what keeps the assistants from reading you.
Our crawler
GaluroBot reads open pages only and goes by its own name — it never passes itself off as anyone else's crawler. It submits no forms and does not go anywhere that needs a password.