The four numbers in your report: mentioned, recommended, cited and wrong

Each number is a count of answers. The report also gives it as a share of the answers read, rounded to a whole percent. We never turn a count into a score or a dollar figure.

Mentioned

The answer mentions your company by name.

A name that is also a common word counts only when it is written as a brand. “Fleet” counts. “Your device fleet” does not. If you sell under more than one name, they count as one company.

Recommended

The answer calls your company the best, the top or its recommended pick.

Being listed first does not count, because the order of an AI list changes every time you ask.

Cited

The answer lists your site as a source.

ChatGPT, Claude and Google AI Overviews mark the pages an answer cites. Perplexity lists every page its search returned without marking which ones the answer used. For Perplexity a listed page is all we can confirm. We use the word “listed” for every tool to match.

Wrong

The answer says something false about your company.

Software compares each sentence that mentions you with what you told us you sell. When we have confirmed facts from your own site, it checks the sentence against those too. It flags only clear contradictions and can miss a false sentence. Before your report is sent, a person at bisica reads every sentence it flagged.

Competitors

Each competitor you name is counted by the same rules. We also add up to five companies that AI mentions often in your answers. Your report says which ones we added.

Reading two numbers side by side

If two shares are less than 10 points apart, we can’t say one is higher. Answers to the same question swing that much between asks.

Failed and empty answers

Answers that failed or came back empty don’t count anywhere. Your report says how many there were. Google shows no AI Overview for some questions. Those questions drop out of the Google AI Overviews numbers only.

The day we asked

A count describes the answers on one day. Your report states that day, the model each tool used and how many answers were read. It can’t tell you what one person sees in their own app, for the reasons in Why answers differ.