diff --git a/klausur-service/backend/grid_editor_api.py b/klausur-service/backend/grid_editor_api.py
index ddc962b..1e4f022 100644
--- a/klausur-service/backend/grid_editor_api.py
+++ b/klausur-service/backend/grid_editor_api.py
@@ -1507,8 +1507,10 @@ async def _build_grid_core(
                 is_artifact = True
             elif len(core) <= 3 and core.isupper() and core.lower() not in _COMMON_SHORT_WORDS:
                 is_artifact = True
-            elif len(core) <= 5 and re.search(r'\d', core) and re.search(r'[A-Za-z]', core):
+            elif (len(core) <= 5 and re.search(r'\d', core) and re.search(r'[A-Za-z]', core)
+                  and not re.match(r'^[pPsS]\.?\d+$', core)):
                 # Mixed digits + letters in short text (e.g. "7 EN", "a=3")
+                # but NOT page references like "p.43", "p50", "S.12"
                 is_artifact = True
             if is_artifact:
                 kept.append(None)  # placeholder
@@ -1717,8 +1719,10 @@ async def _build_grid_core(
     except ImportError:
         pass
 
-    # --- Ensure space before IPA brackets: "word[ipa]" → "word [ipa]" ---
-    _IPA_NOSPACE_RE = re.compile(r'([a-zA-ZäöüÄÖÜß])(\[[^\]]*[ˈˌːɑɒæɛəɜɪɔʊʌðŋθʃʒɹɡɾ][^\]]*\])')
+    # --- Ensure space before IPA/phonetic brackets: "word[ipa]" → "word [ipa]" ---
+    # Matches any [bracket] directly after a letter, as long as the bracket
+    # content doesn't look like a normal text annotation (e.g. "[adj]", "[noun]").
+    _IPA_NOSPACE_RE = re.compile(r'([a-zA-ZäöüÄÖÜß])(\[[^\]]{2,}\])')
     for z in zones_data:
         for cell in z.get("cells", []):
             text = cell.get("text", "")