The comment from @daan above is flat-out wrong.
insert_id() returns the ID placed into an auto_increment field of the last
insert query that the connection executing the insert_id() query performed.
It does not return the largest ID in the table, it doesn't return the id created by some OTHER connection, it doesn't return the id created by some OTHER user.
It is literally the last ID YOU created by the LAST insert YOU performed.
That means it is 100% reliable to perform an insert, get the created ID via
last_insert_id() and then use that ID in other queries to create parent/child records.
But note that
insert_id() only ever returns ONE id value. If you do a multi-value insert, you still only get the ID from the last value set, e.g.
INSERT INTO sometable (x, y) VALUES (1,2), (2,3)
still performs two inserts internally, and you'd only get the id for the
2,3 tuple, not the
As well, there's no memory for all previous queries:
INSERT ... // query #1 INSERT ... // query #2 SET id = last_insert_id();
will only get the ID created by query #2, and the ID from query #1 is "lost".