Set up
Engines and collection
Which AI engines Botlum asks, how often, and what happens when an engine does not answer.
Which engines
Every prompt is asked on four engines:
- ChatGPT;
- Gemini;
- Google AI Overviews: the AI summary above Google's results;
- Google AI Mode: Google's conversational search.
Perplexity and Microsoft Copilot are on the way. Copilot first needs a way to ask from a chosen country. Without it, its answers could not be attributed to a market.
Answers as a person sees them
Botlum collects the answers the engines show to people, in the country of the prompt, not the answers of a developer API. The two can differ, and what people see is what your buyers read.
Daily collection
Each active prompt is asked on every engine once a day. A new prompt is asked right away the first time, so you do not wait until tomorrow for the first answers. Answers arrive through the day. The dashboard shows how complete each day is.
When an engine does not answer
Not every question gets an answer every time. Botlum keeps the difference visible:
- No answer: the engine was asked and gave none. Google, for example, shows an AI Overview only for some searches.
- Failed: Botlum could not get the answer, even after retrying.
Coverage on the dashboard is the share of scheduled questions that got a result. Visibility is counted only over answers that were received and read, so a missing answer never lowers your visibility.
How answers are read
Every answer is read twice. First, exactly: Botlum looks for your brand's and competitors' names and domains in the text and in the cited sources. This decides mentioned, cited and named first. Then a language model reads the answer for what exact matching cannot tell: the sentiment toward each brand, whether it is recommended or only listed, and which other brands the answer names.
Why answers change
AI engines do not give the same answer twice. The same question asked a minute apart can name different brands. That is why Botlum asks every day and reports shares over many answers rather than a single result. It is also why the dashboard says when a change could be noise.