@@ -20,6 +20,8 @@
 )
 DOCSTRING_INDICATORS = ('"""', "'''")
 COMMENT_INDICATORS = (*DOCSTRING_INDICATORS, "'", '"', "#")
+# Every legal Python string prefix, lower-cased; `u` cannot be combined with another.
+STRING_PREFIXES = frozenset(("r", "u", "f", "b", "fr", "rf", "br", "rb"))
 CODE_SORT_COMMENTS = (
     "# isort: list",
     "# isort: dict",
@@ -34,6 +36,17 @@
 SKIP_IMPORT_COMMENTS = ("isort:skip", "isort: skip")
 
 
+def _strip_string_prefix(line: str) -> str:
+    """Return `line` without a leading Python string prefix (`r`, `b`, `f`, `u`
+    or a legal combination of them), so quote-based heuristics see the quote itself."""
+    for prefix_length in (2, 1):
+        if line[:prefix_length].lower() in STRING_PREFIXES and line[
+            prefix_length : prefix_length + 1
+        ] in ("'", '"'):
+            return line[prefix_length:]
+    return line
+
+
 def _has_skip_comment(import_statement: str) -> bool:
     """Return whether an import statement carries a per-line ``isort: skip`` directive."""
     return any(comment in import_statement for comment in SKIP_IMPORT_COMMENTS)
@@ -427,7 +440,7 @@ def process(
                 and not in_top_comment
                 and not was_in_quote
                 and not import_section
-                and not line.lstrip().startswith(COMMENT_INDICATORS)
+                and not _strip_string_prefix(line.lstrip()).startswith(COMMENT_INDICATORS)
                 and not (line.rstrip().endswith(DOCSTRING_INDICATORS) and "=" not in line)
             ):
                 add_line_separator = line_separator or "\n"
 
@@ -2460,3 +2460,29 @@ def test_isort_does_not_drop_aliased_import_when_plain_name_has_a_comment():
     expected = "from . import bar\nfrom . import one  # NOQA\nfrom . import one as zzz\n"
     assert relative_sorted == expected
     assert isort.code(relative_sorted) == relative_sorted
+
+
+def test_add_import_keeps_a_prefixed_module_docstring_first_issue_1893():
+    """``add_imports`` must not move an import above a module docstring that carries a
+    string prefix (``r``, ``b``, ``f``, ``u`` or a legal combination of them), as reported
+    in issue #1893: https://github.com/pycqa/isort/issues/1893
+    """
+    # The exact input from the report: the added import was emitted a second time, above
+    # the docstring.
+    reported = 'r"""module docstring\n"""\nfrom __future__ import annotations\n'
+    assert isort.code(reported, add_imports=["from __future__ import annotations"]) == reported
+
+    # The import still has to be added, below the docstring -- where an unprefixed
+    # docstring already puts it.
+    with_code = 'r"""module docstring\n"""\n\nx = 1\n'
+    assert (
+        isort.code(with_code, add_imports=["import a"])
+        == 'r"""module docstring\n"""\n\nimport a\n\nx = 1\n'
+    )
+
+    # Every legal prefix, in both quote flavours and both cases.
+    for prefix in ("r", "R", "b", "B", "f", "F", "u", "U", "rb", "bR", "Fr", "rf"):
+        for quote in ('"""', "'''"):
+            docstring = f"{prefix}{quote}module docstring\n{quote}\n"
+            source = docstring + "import a\n"
+            assert isort.code(source, add_imports=["import a"]) == source, prefix + quote