Elena Stepanova added a comment - 2018-02-05 00:29 I think it's one of multiple mapping/matching/comparison rules for UTF8, I'll leave it to bar to point at the precise rule (or to say there isn't one). MySQL [test]> set names utf8mb4; Query OK, 0 rows affected (0.00 sec)   MySQL [test]> select 'AKKUŞSAN' = 'AKKUSSAN' ; + --------------------------+ | 'AKKUŞSAN' = 'AKKUSSAN' | + --------------------------+ | 1 | + --------------------------+ 1 row in set (0.00 sec)

Alexander Barkov added a comment - 2018-02-05 07:01 - edited utf8mb4_general_ci is accent insensitive, so this is the expected behaviour. Please use utf8mb4_thai_520_w2 if you need accent sensitive comparison. This script demonstrates that Ş != S when using utf8mb4_thai_520_w2 : CREATE OR REPLACE TABLE t1 ( t varchar (50) NOT NULL , UNIQUE INDEX UK_table1_t (t) ) CHARACTER SET utf8mb4 COLLATE utf8mb4_thai_520_w2; INSERT INTO t1 (t) VALUES ( 'AKKUŞSAN' ); INSERT INTO t1 (t) VALUES ( 'AKKUSSAN' ); SELECT COUNT ( DISTINCT t) FROM t1; +-------------------+ | COUNT(DISTINCT t) | +-------------------+ | 2 | +-------------------+

Alexander Barkov added a comment - 2018-02-05 07:15 Another option is to use utf8mb4_turkish_ci . It implements the following rules: C < ç = Ç G < ğ = Ğ H < ı = I O < ö = Ö S < ş = Ş U < ü = Ü It's accent sensitive only for these specified letters, and accent insensitive for all other letters. Should work for Turkish (and most likely for Azerbaijani)