diff --git a/myst_nb/core/read.py b/myst_nb/core/read.py
index 2b3dae4b..1d8e819c 100644
--- a/myst_nb/core/read.py
+++ b/myst_nb/core/read.py
@@ -31,6 +31,8 @@ class NbReader:
     """The configuration for parsing markdown cells."""
     read_fmt: dict | None = dc.field(default=None)
     """The type of the reader, if known."""
+    support_cell_ids: bool = False
+    """Whether the format supports stable cell IDs"""
 
 
 def standard_nb_read(text: str) -> nbf.NotebookNode:
@@ -75,7 +77,11 @@ def create_nb_reader(
             if commonmark_only:
                 # Markdown cells should be read as Markdown only
                 md_config = dc.replace(md_config, commonmark_only=True)
-            return NbReader(partial(reader, **(reader_kwargs or {})), md_config)  # type: ignore
+            return NbReader(
+                partial(reader, **(reader_kwargs or {})),
+                md_config,
+                support_cell_ids=suffix == ".ipynb",
+            )  # type: ignore
 
     # a Markdown file is a special case, since we only treat it as a notebook,
     # if it starts with certain "top-matter"
@@ -89,6 +95,7 @@ def create_nb_reader(
             ),
             md_config,
             {"type": "plugin", "name": "myst_nb_md"},
+            support_cell_ids=False,
         )
 
     # if we get here, we did not find a reader
diff --git a/myst_nb/core/render.py b/myst_nb/core/render.py
index 0110d3dc..084f6a34 100644
--- a/myst_nb/core/render.py
+++ b/myst_nb/core/render.py
@@ -182,7 +182,7 @@ def render_nb_cell_code(self: SelfType, token: SyntaxTreeNode) -> None:
         )
 
         cell_id = token.meta["id"]
-        if cell_id:
+        if cell_id and not cell_id.startswith("RANDOM_CELL_ID"):
             self.document.note_implicit_target(cell_container, cell_container)
             slug = "cell-id=" + cell_id
             cell_container["slug"] = slug
diff --git a/myst_nb/sphinx_.py b/myst_nb/sphinx_.py
index dba36531..344a7779 100644
--- a/myst_nb/sphinx_.py
+++ b/myst_nb/sphinx_.py
@@ -87,6 +87,12 @@ def parse(self, inputstring: str, document: nodes.document) -> None:
             return super().parse(inputstring, document)
         notebook = nb_reader.read(inputstring)
 
+        if not nb_reader.support_cell_ids:
+            # mark randomly generated cell IDs
+            for cell in notebook["cells"]:
+                if "id" in cell:
+                    cell["id"] = "RANDOM_CELL_ID_" + cell["id"]
+
         # update the global markdown config with the file-level config
         warning = lambda wtype, msg: create_warning(  # noqa: E731
             document, msg, line=1, append_to=document, subtype=wtype