diff --git a/src/utils/Cursor.nfc.test.ts b/src/utils/Cursor.nfc.test.ts
new file mode 100644
index 0000000000..776c882952
--- /dev/null
+++ b/src/utils/Cursor.nfc.test.ts
@@ -0,0 +1,31 @@
+import { describe, expect, test } from 'bun:test'
+import { Cursor } from './Cursor.js'
+
+describe('Cursor.insert NFC boundary offset', () => {
+  test('places the cursor correctly when the insert composes with the prefix', () => {
+    // Cursor sits between "e" and "X". Inserting a combining acute accent
+    // composes with the "e" into a single "é", so the normalized text is "éX"
+    // (length 2) and the cursor belongs at offset 1 (before "X"). The old code
+    // measured the insert in isolation and landed at offset 2 (after "X").
+    const c = Cursor.fromText('eX', 80, 1)
+    expect(c.text).toBe('eX')
+
+    const after = c.insert('́')
+    expect(after.text).toBe('éX')
+    expect(after.offset).toBe(1)
+  })
+
+  test('non-composing inserts still advance by the inserted length', () => {
+    const c = Cursor.fromText('abcX', 80, 3) // between "abc" and "X"
+    const after = c.insert('de')
+    expect(after.text).toBe('abcdeX')
+    expect(after.offset).toBe(5)
+  })
+
+  test('astral-plane insert advances by its UTF-16 unit count', () => {
+    const c = Cursor.fromText('X', 80, 0)
+    const after = c.insert('😀') // 2 UTF-16 code units
+    expect(after.text).toBe('😀X')
+    expect(after.offset).toBe(2)
+  })
+})
diff --git a/src/utils/Cursor.ts b/src/utils/Cursor.ts
index 4622ea0acc..660bf1a359 100644
--- a/src/utils/Cursor.ts
+++ b/src/utils/Cursor.ts
@@ -875,11 +875,17 @@ export class Cursor {
       insertString +
       this.text.slice(endOffset)
 
-    return Cursor.fromText(
-      newText,
-      this.columns,
-      startOffset + insertString.normalize('NFC').length,
-    )
+    // Cursor.fromText NFC-normalizes the whole newText, so compute the new
+    // offset from the normalized prefix-plus-insert rather than normalizing
+    // insertString in isolation. Otherwise a combining mark that composes with
+    // the last character of the prefix (e.g. "e" + U+0301 -> "é") shortens the
+    // normalized text by one unit that this offset doesn't account for, landing
+    // the cursor one position too far (past following text).
+    const newOffset = (
+      this.text.slice(0, startOffset) + insertString
+    ).normalize('NFC').length
+
+    return Cursor.fromText(newText, this.columns, newOffset)
   }
 
   insert(insertString: string): Cursor {