Select Count(*) has to do the fuzzy like search on the whole table just to show rows that match might exist - so it will probably do a full table scan (if start of like is a %) better is to let SQL return as soon as it has found one row that matches - so use exists - e.g.
String.Format("SELECT 1 WHERE EXISTS (SELECT 1 FROM data WHERE oznacenie LIKE '{0}')", data)
That way SQL can stop looking after the first row it finds that matches. Depending on size of table this can make a huge difference in performance. This will return 0 or 1 records to the rowcount which you can get using ExecuteScalar return
