When set to ON, existing statistics are automatically updated when the statistics 
become out-of-date because the data in the tables has changed. 

...

Therefore, SQL Server automatically updates this statistical information periodically 
as the data in the tables changes. The sampling is random across data pages, and 
taken from the table or the smallest nonclustered index on the columns needed by 
the statistics. After a data page has been read from disk, all the rows on the data 
page are used to update the statistical information. The frequency at which the 
statistical information is updated is determined by the volume of data in the 
column or index and the amount of changing data. For example, the statistics for a 
table containing  10 , 000  rows may need updating when  1 , 000  index values have 
changed because  1 , 000  values may represent a significant percentage of the 
table. However, for a table containing  10  million index entries,  1 , 000  changing 
index values is less significant, and so the statistics may not be automatically 
updated. SQL Server, however, always ensures that a minimum number of rows are 
sampled; tables that are smaller than  8  megabytes (MB) are always fully scanned 
to gather statistics.

The cost of this automatic statistical update is minimized by sampling the data, 
rather than analyzing all of it. Under some circumstances, statistical sampling will 
not be able to accurately characterize the data in a table. You can control the 
amount of data that is sampled during manual statistics updates on a table-by-
table basis by using the SAMPLE and FULLSCAN clauses of the UPDATE 
STATISTICS statement.