diff --git a/llm/cli.py b/llm/cli.py
index ecd8121a7..38f07eeaa 100644
--- a/llm/cli.py
+++ b/llm/cli.py
@@ -3258,11 +3258,14 @@ def get_db():
     if not content:
         if not input or input == "-":
             # Read from stdin
-            input_source = sys.stdin.buffer if binary else sys.stdin
-            content = input_source.read()
+            if binary:
+                content = sys.stdin.buffer.read()
+            else:
+                content = sys.stdin.buffer.read().decode("utf-8")
         else:
             mode = "rb" if binary else "r"
-            with open(input, mode) as f:
+            kwargs = {} if binary else {"encoding": "utf-8"}
+            with open(input, mode, **kwargs) as f:
                 content = f.read()
 
     if not content:
@@ -3570,11 +3573,14 @@ def similar(collection, id, input, content, binary, number, plain, database, pre
         if not content:
             if not input or input == "-":
                 # Read from stdin
-                input_source = sys.stdin.buffer if binary else sys.stdin
-                content = input_source.read()
+                if binary:
+                    content = sys.stdin.buffer.read()
+                else:
+                    content = sys.stdin.buffer.read().decode("utf-8")
             else:
                 mode = "rb" if binary else "r"
-                with open(input, mode) as f:
+                kwargs = {} if binary else {"encoding": "utf-8"}
+                with open(input, mode, **kwargs) as f:
                     content = f.read()
         if not content:
             raise click.ClickException("No content provided")
diff --git a/tests/test_embed_cli.py b/tests/test_embed_cli.py
index 11b67f56c..295ca1c54 100644
--- a/tests/test_embed_cli.py
+++ b/tests/test_embed_cli.py
@@ -347,6 +347,24 @@ def test_embed_multi_file_input(tmpdir, use_stdin, prefix, prepend, filename, co
     assert ids == expected_ids
 
 
+def test_embed_file_utf8_content_preserved(tmpdir, user_path):
+    """Non-ASCII UTF-8 characters in a file should be stored without corruption."""
+    db_path = str(user_path / "embeddings.db")
+    path = tmpdir / "content.txt"
+    # Write raw UTF-8 bytes to guarantee the on-disk encoding
+    path.write_binary("æ ø å".encode("utf-8"))
+    runner = CliRunner()
+    result = runner.invoke(
+        cli,
+        ["embed", "items", "1", "-i", str(path), "-m", "embed-demo", "--store", "-d", db_path],
+        catch_exceptions=False,
+    )
+    assert result.exit_code == 0
+    db = sqlite_utils.Database(db_path)
+    row = list(db["embeddings"].rows)[0]
+    assert row["content"] == "æ ø å"
+
+
 def test_embed_multi_files_binary_store(tmpdir):
     db_path = tmpdir / "embeddings.db"
     args = ["embed-multi", "binfiles", "-d", str(db_path), "-m", "embed-demo"]