Detect and convert bad encoded UTF-8 data in MySQL

convert-latin1-utf8.sql

-- Convert
SELECT CONVERT(CAST(CONVERT('São Paulo' USING latin1) AS BINARY) USING utf8) -- São Paulo

-- Our lovely friends celebrate ÀÁÂÃÄÅÆÇÈÉÊËÌÍÎÏÐÑÒÓÔÕÖ×ØÙÚÛÜÝÞßàáâãäåæçèéêëìíîïðñòóôõö \o/

-- Poor man's bad encoding detection (in my case there was LATIN1 data mixed with UTF8 in the database and I just wanted to find when things started to get mixed up)
-- When trying to convert, normally the amount of characters decrease...
SELECT field
FROM table
WHERE LENGTH(CONVERT(CAST(CONVERT(field USING latin1) AS BINARY) USING utf8)) < LENGTH(field) 
-- And if an specific character fails, MySQL replaces it by "????"
AND (
  -- So the output shouldn't have the interrogation
  LOCATE('?', CONVERT(CAST(CONVERT(field USING latin1) AS BINARY) USING utf8)) = 0
  -- Or the original text should have it as well
  OR LOCATE('?', field) <> 0
)