diff --git a/aborist/qa/keys.py b/aborist/qa/keys.py
index 5da7adf..a25cf0c 100644
--- a/aborist/qa/keys.py
+++ b/aborist/qa/keys.py
@@ -35,6 +35,22 @@ def _canonical_json(obj) -> str:
     return json.dumps(obj, sort_keys=True, separators=(",", ":"), ensure_ascii=False)
 
 
+def canonical_question(question: str) -> str:
+    """Canonical equivalence-class form of `question` text.
+
+    Same four-step normalization that `question_hash` hashes, exposed
+    so callers can dedup BEFORE hashing — e.g. inject the canonical
+    form into the user message used for `conversation_hash`, while
+    still sending the verbatim question to the LLM. Without this split
+    `"who is batman"`, `"who is batman?"`, and `"who is the batman?"`
+    all collapse on `question_hash` but each hits `conversation_hash`
+    differently, missing cache.
+    """
+    canon = canonicalize(question).lower().rstrip(_QUESTION_TRAILING_STRIP)
+    tokens = [t for t in canon.split() if t not in _QUESTION_ARTICLE_STRIP]
+    return " ".join(tokens)
+
+
 def question_hash(question: str) -> str:
     """SHA-256 of canonicalized + lowercased + punctuation-stripped +
     article-stripped question text.
@@ -74,10 +90,7 @@ def question_hash(question: str) -> str:
     Apostrophes aren't either — ``X's`` is a different question from
     ``X``.
     """
-    canon = canonicalize(question).lower().rstrip(_QUESTION_TRAILING_STRIP)
-    tokens = [t for t in canon.split() if t not in _QUESTION_ARTICLE_STRIP]
-    canon = " ".join(tokens)
-    return _sha256(canon)
+    return _sha256(canonical_question(question))
 
 
 # Trailing punctuation that carries no semantic difference at the end
diff --git a/aborist/qa/query.py b/aborist/qa/query.py
index 7670d96..7545b8d 100644
--- a/aborist/qa/query.py
+++ b/aborist/qa/query.py
@@ -54,6 +54,7 @@ from aborist.qa.concepts import (
 )
 from aborist.qa.keys import (
     cache_key,
+    canonical_question,
     conversation_hash,
     governance_policy_hash,
     model_profile_hash,
@@ -707,6 +708,13 @@ def query(
     # user-turn instructions more reliably than system-turn rules under long
     # context). The sources + question land last so they are the most-recent
     # tokens before generation begins.
+    #
+    # Decoupled hash form vs LLM form: `messages` carries the verbatim
+    # question (Hermes sees the user's natural phrasing); `canonical_messages`
+    # substitutes the canonical question (no articles, no trailing
+    # punctuation) so `conversation_hash` collapses on the equivalence class.
+    # `who is batman`, `who is batman?`, `who is the batman?` → same chash.
+    canonical_q = canonical_question(question)
     messages = [
         {"role": "system", "content": policy["system_prompt"]},
     ]
@@ -720,10 +728,16 @@ def query(
             "content": f"Sources:\n\n{context}\n\n---\n\nQuestion: {question}",
         }
     )
+    canonical_messages = list(messages[:-1]) + [
+        {
+            "role": "user",
+            "content": f"Sources:\n\n{context}\n\n---\n\nQuestion: {canonical_q}",
+        }
+    ]
     context_root = _context_root([h.document_root for h in chosen])
     qhash = question_hash(question)
     mhash = model_profile_hash(model_id, revision, quantization)
-    chash = conversation_hash(messages)
+    chash = conversation_hash(canonical_messages)
     ghash = governance_policy_hash(policy)
     ckey = cache_key(
         context_root,
diff --git a/aborist/qa/runner.py b/aborist/qa/runner.py
index c5b9fdd..e20e167 100644
--- a/aborist/qa/runner.py
+++ b/aborist/qa/runner.py
@@ -24,6 +24,7 @@ from aborist.merkle import MerkleTree, proof_to_dict
 from aborist.qa.client import ChatClient
 from aborist.qa.keys import (
     cache_key,
+    canonical_question,
     conversation_hash,
     governance_policy_hash,
     model_profile_hash,
@@ -125,6 +126,13 @@ def ask(
     # System sets the policy; a user-turn reminder restates the verbatim-quote
     # rule one message before the document arrives. Document + question land
     # last as the most-recent tokens before generation.
+    #
+    # Decoupled hash form vs LLM form: `messages` carries the verbatim
+    # question; `canonical_messages` substitutes the canonical question
+    # (no articles, no trailing punctuation) so `conversation_hash`
+    # collapses on the question equivalence class. See aborist/qa/query.py
+    # for the rationale.
+    canonical_q = canonical_question(question)
     messages = [
         {"role": "system", "content": policy["system_prompt"]},
     ]
@@ -141,10 +149,19 @@ def ask(
             ),
         }
     )
+    canonical_messages = list(messages[:-1]) + [
+        {
+            "role": "user",
+            "content": (
+                f"Document:\n\n{document_text}\n\n---\n\n"
+                f"Question: {canonical_q}"
+            ),
+        }
+    ]
 
     qhash = question_hash(question)
     mhash = model_profile_hash(model_id, revision, quantization)
-    chash = conversation_hash(messages)
+    chash = conversation_hash(canonical_messages)
     ghash = governance_policy_hash(policy)
     ckey = cache_key(
         document_root,
diff --git a/tests/test_query.py b/tests/test_query.py
index e9fdd62..1c6e2d7 100644
--- a/tests/test_query.py
+++ b/tests/test_query.py
@@ -133,6 +133,65 @@ def test_query_different_question_different_cache(tmp_path):
     assert len(client.calls) == 2  # both missed cache, both called client
 
 
+def test_query_question_equivalence_class_dedups_cache(tmp_path):
+    """Fox 2026-04-29 catch: 'who is batman?', 'who is batman', and
+    'who is the batman?' should all hit the same cache. Question_hash
+    canonicalizes correctly, but conversation_hash used to hash the
+    LITERAL question text in the user message — so each variant got
+    its own chash and missed cache. Fix: canonical_question form is
+    substituted into the messages list used for conversation_hash,
+    while the LLM still receives the verbatim question."""
+    main_db = tmp_path / "corpus.db"
+    qa_db = tmp_path / "qa.db"
+    conn = connect(main_db)
+    try:
+        ingest_source(conn, FakeSource(DOCS))
+    finally:
+        conn.close()
+
+    captured: list = []
+
+    def _capture(messages, **kw):
+        captured.append(messages)
+        return "stub answer"
+
+    # First variant — populates cache.
+    r1 = query(
+        question="What is anarchism?",
+        qa_db=qa_db,
+        chat_client=StubClient(answer=_capture),
+        model_id="m",
+        single_db=main_db,
+    )
+    # Second & third variants — must hit the same cache_key.
+    r2 = query(
+        question="What is anarchism",  # no trailing ?
+        qa_db=qa_db,
+        chat_client=StubClient(answer=_capture),
+        model_id="m",
+        single_db=main_db,
+    )
+    r3 = query(
+        question="what is the anarchism?",  # leading article
+        qa_db=qa_db,
+        chat_client=StubClient(answer=_capture),
+        model_id="m",
+        single_db=main_db,
+    )
+
+    assert r1["cache_key"] == r2["cache_key"] == r3["cache_key"]
+    assert r1["status"] == "cache_miss_then_written"
+    assert r2["status"] == "cache_hit"
+    assert r3["status"] == "cache_hit"
+    # Only the first call reached the LLM.
+    assert len(captured) == 1
+    # And it received the verbatim question, not the canonical form.
+    user_text = "\n".join(
+        m["content"] for m in captured[0] if m["role"] == "user"
+    )
+    assert "What is anarchism?" in user_text
+
+
 def test_query_no_sources_when_empty_corpus(tmp_path):
     main_db = tmp_path / "empty.db"
     qa_db = tmp_path / "qa.db"