diff --git a/packages/markitdown/src/markitdown/converters/_ipynb_converter.py b/packages/markitdown/src/markitdown/converters/_ipynb_converter.py
index 6bc1c1673f..62925f5ef3 100644
--- a/packages/markitdown/src/markitdown/converters/_ipynb_converter.py
+++ b/packages/markitdown/src/markitdown/converters/_ipynb_converter.py
@@ -24,26 +24,26 @@ def accepts(
         mimetype = (stream_info.mimetype or "").lower()
         extension = (stream_info.extension or "").lower()
 
-        if extension in ACCEPTED_FILE_EXTENSIONS:
-            return True
-
         for prefix in CANDIDATE_MIME_TYPE_PREFIXES:
             if mimetype.startswith(prefix):
-                # Read further to see if it's a notebook
+                # Check the JSON structure, not just mentions of notebook fields.
                 cur_pos = file_stream.tell()
                 try:
                     encoding = stream_info.charset or "utf-8"
                     notebook_content = file_stream.read().decode(encoding)
+                    notebook = json.loads(notebook_content.lstrip("\ufeff"))
                     return (
-                        "nbformat" in notebook_content
-                        and "nbformat_minor" in notebook_content
+                        isinstance(notebook, dict)
+                        and type(notebook.get("nbformat")) is int
+                        and type(notebook.get("nbformat_minor")) is int
+                        and isinstance(notebook.get("cells"), list)
                     )
-                except (ValueError, LookupError):
+                except (ValueError, LookupError, RecursionError):
                     return False
                 finally:
                     file_stream.seek(cur_pos)
 
-        return False
+        return extension in ACCEPTED_FILE_EXTENSIONS
 
     def convert(
         self,
diff --git a/packages/markitdown/tests/test_ipynb_detection.py b/packages/markitdown/tests/test_ipynb_detection.py
new file mode 100644
index 0000000000..036047910e
--- /dev/null
+++ b/packages/markitdown/tests/test_ipynb_detection.py
@@ -0,0 +1,93 @@
+import io
+import json
+import sys
+
+import pytest
+
+from markitdown import MarkItDown, StreamInfo
+from markitdown.converters._ipynb_converter import IpynbConverter
+
+
+@pytest.mark.parametrize(
+    "payload",
+    [
+        {"description": "nbformat_minor describes a notebook version"},
+        {"notebook": {"nbformat": 4, "nbformat_minor": 5, "cells": []}},
+        {"nbformat": 4, "nbformat_minor": 5, "description": "Format documentation"},
+        {
+            "nbformat": {"type": "integer"},
+            "nbformat_minor": {"type": "integer"},
+            "cells": [],
+        },
+    ],
+    ids=["text-value", "nested-notebook", "missing-cells", "schema-fields"],
+)
+@pytest.mark.parametrize("entry_point", ["file", "stream"])
+def test_json_notebook_references_are_preserved(payload, entry_point, tmp_path):
+    """Mentioning notebook fields must not discard an ordinary JSON document."""
+    text = json.dumps(payload)
+    converter = MarkItDown()
+    if entry_point == "file":
+        path = tmp_path / "documentation.json"
+        path.write_text(text, encoding="utf-8")
+        result = converter.convert(path)
+    else:
+        result = converter.convert_stream(
+            io.BytesIO(text.encode("utf-8")),
+            stream_info=StreamInfo(mimetype="application/json"),
+        )
+
+    assert result.markdown == text
+
+
+@pytest.mark.parametrize(
+    "data, charset",
+    [
+        (b'["nbformat", "nbformat_minor"]', "utf-8"),
+        (b'"nbformat_minor"', "utf-8"),
+        (b'{"nbformat": 4, "nbformat_minor": 5, "cells": [}', "utf-8"),
+        (b"[]", "unknown-charset"),
+        (
+            b"[" * sys.getrecursionlimit()
+            + b'"nbformat_minor"'
+            + b"]" * sys.getrecursionlimit(),
+            "utf-8",
+        ),
+    ],
+    ids=["array", "string", "invalid-json", "unknown-charset", "deep-json"],
+)
+def test_ipynb_json_probe_rejects_non_notebooks_without_consuming_stream(data, charset):
+    stream = io.BytesIO(b"prefix" + data)
+    stream.seek(len(b"prefix"))
+    position = stream.tell()
+
+    assert not IpynbConverter().accepts(
+        stream, StreamInfo(mimetype="application/json", charset=charset)
+    )
+    assert stream.tell() == position
+    assert stream.read() == data
+
+
+@pytest.mark.parametrize(
+    "encoding, charset",
+    [
+        ("utf-8", None),
+        ("utf-8-sig", "utf-8"),
+        ("utf-8-sig", "utf-8-sig"),
+        ("utf-16", "utf-16"),
+    ],
+)
+def test_ipynb_json_probe_preserves_notebook_conversion(encoding, charset):
+    notebook = {
+        "nbformat": 4,
+        "nbformat_minor": 0,
+        "cells": [{"cell_type": "markdown", "source": ["# Notebook\n"]}],
+    }
+    stream = io.BytesIO(json.dumps(notebook).encode(encoding))
+    stream_info = StreamInfo(mimetype="application/json", charset=charset)
+
+    assert IpynbConverter().accepts(stream, stream_info)
+    assert stream.tell() == 0
+    result = MarkItDown().convert_stream(stream, stream_info=stream_info)
+    assert result.markdown == "# Notebook\n"
+    assert result.title == "Notebook"