diff --git a/conditioning.py b/conditioning.py
index e6280754..91143214 100644
--- a/conditioning.py
+++ b/conditioning.py
@@ -10668,9 +10668,10 @@ class FunPackVideoRefinerV2(FunPackVideoRefiner):
                   "splits into scenes and each scene is enhanced instead.")
         if _enhance_base:
             _before = prompt_to_encode
+            _ref = self._v2_enhancer_reference(prompt_to_encode, _enhance_sources)
             prompt_to_encode, _enhance_status = self._v2_enhance_prompt(
                 _enhance_clip, prompt_to_encode, _enhance_system, cache=_enhance_cache,
-                reference=self._v2_enhancer_reference(prompt_to_encode, _enhance_sources),
+                reference=_ref,
                 seed=_enhance_seed, temperature=prompt_enhance_temperature,
                 top_p=prompt_enhance_top_p, max_length=prompt_enhance_max_length,
                 thinking=prompt_enhance_thinking, image=prompt_enhance_image,
@@ -10679,7 +10680,8 @@ class FunPackVideoRefinerV2(FunPackVideoRefiner):
             print(f"[FunPackVideoRefinerV2] {_enhance_status}")
             self._v2_enhanced_prompts.append({
                 "scene": None, "before": _before, "after": prompt_to_encode,
-                "status": _enhance_status, "image": prompt_enhance_image is not None})
+                "status": _enhance_status, "image": prompt_enhance_image is not None,
+                "reference": _ref.strip()})
 
         # Studio's `enhanced_prompt` output, for a node encoding on its own (Editor link
         # "Enhanced prompt, fallback to prompt + postfix"). The same text that link's
@@ -10688,9 +10690,10 @@ class FunPackVideoRefinerV2(FunPackVideoRefiner):
         if prompt_enhance_output:
             _whole = str((_link_texts or {}).get("full_prompt") or "").strip()
             if prompt_enhance and _whole:
+                _ref = self._v2_enhancer_reference(_whole, _enhance_sources)
                 _after, _st = self._v2_enhance_prompt(
                     _enhance_clip, _whole, _enhance_system, cache=_enhance_cache,
-                    reference=self._v2_enhancer_reference(_whole, _enhance_sources),
+                    reference=_ref,
                     seed=_enhance_seed, temperature=prompt_enhance_temperature,
                     top_p=prompt_enhance_top_p, max_length=prompt_enhance_max_length,
                     thinking=prompt_enhance_thinking, image=prompt_enhance_image,
@@ -10698,7 +10701,7 @@ class FunPackVideoRefinerV2(FunPackVideoRefiner):
                 print(f"[FunPackVideoRefinerV2] Enhanced prompt output: {_st}")
                 self._v2_enhanced_prompts.append({
                     "scene": None, "before": _whole, "after": _after, "status": _st,
-                    "image": prompt_enhance_image is not None})
+                    "image": prompt_enhance_image is not None, "reference": _ref.strip()})
                 _whole = _after
             self._v2_enhanced_output = _whole
 
@@ -10938,9 +10941,10 @@ class FunPackVideoRefinerV2(FunPackVideoRefiner):
                     if _enhance_own and split_scene_texts:
                         _enhanced_texts = []
                         for _i, t in enumerate(split_scene_texts):
+                            _ref = self._v2_enhancer_reference(t, _enhance_sources)
                             _after, _st = self._v2_enhance_prompt(
                                 _enhance_clip, t, _enhance_system, cache=_enhance_cache,
-                                reference=self._v2_enhancer_reference(t, _enhance_sources),
+                                reference=_ref,
                                 seed=_enhance_seed, temperature=prompt_enhance_temperature,
                                 top_p=prompt_enhance_top_p,
                                 max_length=prompt_enhance_max_length,
@@ -10949,7 +10953,8 @@ class FunPackVideoRefinerV2(FunPackVideoRefiner):
                             print(f"[FunPackVideoRefinerV2] Scene {_i + 1}: {_st}")
                             self._v2_enhanced_prompts.append({
                                 "scene": _i, "before": t, "after": _after, "status": _st,
-                                "image": prompt_enhance_image is not None})
+                                "image": prompt_enhance_image is not None,
+                                "reference": _ref.strip()})
                             _enhanced_texts.append(_after)
                         split_scene_texts = _enhanced_texts
                     scene_refinement_keys = [set(s.get("keys") or set()) for s in canon_scenes]
diff --git a/movie_editor/frontend/composer.js b/movie_editor/frontend/composer.js
index 5b4b6d81..5073b6d1 100644
--- a/movie_editor/frontend/composer.js
+++ b/movie_editor/frontend/composer.js
@@ -1060,6 +1060,15 @@
       det.append(el("summary", null, "What went in"), el("div", "enh-before-text", before));
       card.append(det);
     }
+    // What was appended after the prompt (Enhance ▸ Reference) — the model read it too.
+    const ref = String(it.reference || "").trim();
+    if (ref) {
+      const n = (ref.match(/^\[(Shortcut|Lore)\] /gm) || []).length;
+      const det = el("details", "enh-before");
+      det.append(el("summary", null, `Reference sent (${n} ${n === 1 ? "entry" : "entries"})`),
+        el("div", "enh-before-text enh-ref-sent", ref));
+      card.append(det);
+    }
     if (it.status) card.append(el("div", "enh-status", String(it.status)));
     return card;
   }
diff --git a/movie_editor/frontend/styles.css b/movie_editor/frontend/styles.css
index 2189fa85..a4ec0d67 100644
--- a/movie_editor/frontend/styles.css
+++ b/movie_editor/frontend/styles.css
@@ -2133,6 +2133,7 @@ body.tour-mode { overflow: hidden; }
 .enh-before summary { font-size: 11px; color: var(--muted); cursor: pointer; }
 .enh-before-text { margin-top: 4px; padding-left: 8px; border-left: 2px solid var(--line); font-size: 12px; line-height: 1.5;
   color: var(--muted); white-space: pre-wrap; overflow-wrap: anywhere; user-select: text; }
+.enh-ref-sent { max-height: 260px; overflow-y: auto; }
 .enh-status { margin-top: 8px; font-size: 10.5px; line-height: 1.4; color: var(--faint); }
 .eng-enhance-host { margin-top: 16px; }
 
diff --git a/tests/test_v2_category_alignment.py b/tests/test_v2_category_alignment.py
index e09028f8..b5973548 100644
--- a/tests/test_v2_category_alignment.py
+++ b/tests/test_v2_category_alignment.py
@@ -2329,3 +2329,16 @@ def test_studio_skips_its_own_enhancement_when_told_nothing_reads_it(tmp_path):
     out, clip = _enhanced_output(tmp_path, prompt_enhance=True, prompt_enhance_scenes=False,
                                  prompt_enhance_output=True)
     assert out == "a vivid cat on a sunlit sill"
+
+
+def test_last_run_readout_shows_the_reference_that_was_sent(tmp_path, monkeypatch):
+    import templates
+    monkeypatch.setattr(templates, "load_shortcut_db", lambda: {"shortcuts": {
+        "k1": {"name": "Mira", "replacements": ["red coat"], "triggers": ["/mira"]}}})
+    refiner = FunPackVideoRefinerV2()
+    refiner._v2_state_path = lambda refinement_key: str(tmp_path / "state.json")
+    refiner.refine_v2("a cat", GeneratingClip("a vivid cat"), "Unrated", "ref-readout",
+                      link_texts={"full_prompt": "a cat"}, prompt_enhance=True,
+                      prompt_enhance_output=True, prompt_enhance_shortcuts=["k1"])
+    item = refiner._v2_enhanced_prompts[-1]
+    assert item["before"] == "a cat" and item["reference"].count("[Shortcut] Mira\nred coat") == 1