BUG: inconsistency in dtype of replace()

pandas/core/internals/blocks.py

@@ -661,8 +661,13 @@ def replace(
             return [blk]
 
         regex = should_use_regex(regex, to_replace)
+        mask = missing.mask_missing(values, to_replace)
 
         if regex:
+            if self._can_hold_element(value):
+                blk = self if inplace else self.copy()
+                putmask_inplace(blk.values, mask, value)
+                return blk.convert(numeric=False, copy=False)
             return self._replace_regex(to_replace, value, inplace=inplace)
 
         if not self._can_hold_element(to_replace):
@@ -672,7 +677,6 @@ def replace(
             #  replace_list instead of replace.
             return [self] if inplace else [self.copy()]
 
-        mask = missing.mask_missing(values, to_replace)
         if not mask.any():
             # Note: we get here with test_replace_extension_other incorrectly
             #  bc _can_hold_element is incorrect.

pandas/tests/series/methods/test_replace.py

@@ -512,3 +512,10 @@ def test_pandas_replace_na(self):
         result = ser.replace(regex_mapping, regex=True)
         exp = pd.Series(["CC", "CC", "CC-REPL", "DD", "CC", "", pd.NA], dtype="string")
         tm.assert_series_equal(result, exp)
+
+    def test_replace_regex_dtype(self):
+        # GH-48644
+        s = pd.Series(["0"])
+        exp = s.replace(to_replace="0", value=1, regex=False).dtype
+        res = s.replace(to_replace="0", value=1, regex=True).dtype
+        assert exp == res