diff --git a/.jules/bolt.md b/.jules/bolt.md
index 5762283c6..f778fe2d9 100644
--- a/.jules/bolt.md
+++ b/.jules/bolt.md
@@ -21,3 +21,6 @@
 ## 2024-05-18 - Pre-compile Regex in OCR formatting
 **Learning:** `OcrTextFormatter.formatLines` instantiated `Regex` for bullet removal inline every time it was called when `stripBullets` was true. This causes unnecessary allocation and compilation overhead when parsing OCR lines.
 **Action:** Extract frequently used formatting regexes (such as `bulletRegex`) into `object` or companion object private constants (`private val BULLET_REGEX`) to optimize regex reuse and eliminate unnecessary compilation per text parsing operation.
+## 2024-05-18 - Optimize Diacritic Stripping in Suggest
+**Learning:** In `Suggest.isSameIgnoringAccents`, `val DIACRITIC_REGEX = "\\p{M}+".toRegex()` was being used in combination with `Normalizer.normalize` and `replace` to strip accents. This created string allocations via `Normalizer.normalize` and `Regex.replace` even for plain ASCII strings that lack accents, which is a hot path during typing and suggestion processing.
+**Action:** Replace `Regex.replace` with a manual character loop that avoids `Normalizer.normalize` entirely for pure ASCII strings and strips `Character.NON_SPACING_MARK` without regex allocation when normalizing.
diff --git a/app/src/main/java/helium314/keyboard/latin/Suggest.kt b/app/src/main/java/helium314/keyboard/latin/Suggest.kt
index 7e8328b34..d6700f11e 100644
--- a/app/src/main/java/helium314/keyboard/latin/Suggest.kt
+++ b/app/src/main/java/helium314/keyboard/latin/Suggest.kt
@@ -603,11 +603,30 @@ class Suggest(private val mDictionaryFacilitator: DictionaryFacilitator) {
             return false
         }
 
-        private val DIACRITIC_REGEX = "\\p{M}+".toRegex()
+        private fun stripAccents(s: String): String {
+            var hasAccents = false
+            for (i in s.indices) {
+                if (s[i].code > 127) {
+                    hasAccents = true
+                    break
+                }
+            }
+            if (!hasAccents) return s
+
+            val norm = java.text.Normalizer.normalize(s, java.text.Normalizer.Form.NFD)
+            val clean = java.lang.StringBuilder(norm.length)
+            for (i in norm.indices) {
+                val c = norm[i]
+                if (Character.getType(c) != Character.NON_SPACING_MARK.toInt()) {
+                    clean.append(c)
+                }
+            }
+            return clean.toString()
+        }
+
         private fun isSameIgnoringAccents(a: String, b: String): Boolean {
-            val normA = java.text.Normalizer.normalize(a, java.text.Normalizer.Form.NFD).replace(DIACRITIC_REGEX, "")
-            val normB = java.text.Normalizer.normalize(b, java.text.Normalizer.Form.NFD).replace(DIACRITIC_REGEX, "")
-            return normA.equals(normB, ignoreCase = true)
+            if (a.equals(b, ignoreCase = true)) return true
+            return stripAccents(a).equals(stripAccents(b), ignoreCase = true)
         }
 
         private fun getTransformedSuggestedWordInfoList(