A machine learning model is used to determine the correct frequency of indexing to provide the most accurate results based on several factors, including competition and appearance rate. The indexing frequency varies based on the type of search term. An exception to this are Hourly terms, which are indexed up to 24 times per day per device.