As others have mentioned, it’s possible the analysis was done through straight string matching; would know for sure once the code is open sourced.