diff --git a/src/anndata/_repr/html.py b/src/anndata/_repr/html.py
index a9342f2c9..e60b72bd4 100644
--- a/src/anndata/_repr/html.py
+++ b/src/anndata/_repr/html.py
@@ -74,6 +74,7 @@
     get_setting,
     is_backed,
     is_view,
+    render_markdown,
 )
 
 if TYPE_CHECKING:
@@ -466,7 +467,7 @@ def _render_custom_section(
     )
 
 
-def _render_header(
+def _render_header(  # noqa: PLR0912, PLR0915
     adata: AnnData, *, show_search: bool = False, container_id: str = ""
 ) -> str:
     """Render the header with type, shape, badges, and optional search box."""
@@ -537,8 +538,16 @@ def _render_header(
             )
             readme_content += truncation_note
 
-        escaped_readme = escape_html(readme_content)
-        # Truncate for no-JS tooltip (first 500 chars)
+        # Try rendering markdown; fall back to plain text
+        rendered_html = render_markdown(readme_content)
+        if rendered_html is not None:
+            escaped_readme = escape_html(rendered_html)
+            readme_format = "html"
+        else:
+            escaped_readme = escape_html(readme_content)
+            readme_format = "text"
+
+        # Truncate for no-JS tooltip (first 500 chars of source text)
         tooltip_text = readme_content[:TOOLTIP_TRUNCATE_LENGTH]
         if len(readme_content) > TOOLTIP_TRUNCATE_LENGTH:
             tooltip_text += "..."
@@ -547,6 +556,7 @@ def _render_header(
         parts.append(
             f'<span class="anndata-readme__icon" '
             f'data-readme="{escaped_readme}" '
+            f'data-readme-format="{readme_format}" '
             f'title="{escaped_tooltip}" '
             f'role="button" tabindex="0" aria-label="View README">'
             f"ⓘ"
diff --git a/src/anndata/_repr/static/repr.css b/src/anndata/_repr/static/repr.css
index e8a4faec9..50ffbebac 100644
--- a/src/anndata/_repr/static/repr.css
+++ b/src/anndata/_repr/static/repr.css
@@ -281,13 +281,120 @@
         line-height: 1.6;
         color: var(--anndata-text-primary);
 
+        h1,
+        h2,
+        h3,
+        h4 {
+            margin-top: 16px;
+            margin-bottom: 8px;
+            font-weight: 600;
+            color: var(--anndata-text-primary);
+        }
+
+        h1 {
+            font-size: 1.5em;
+            padding-bottom: 8px;
+            border-bottom: 1px solid var(--anndata-border-color);
+        }
+
+        h2 {
+            font-size: 1.3em;
+        }
+
+        h3 {
+            font-size: 1.1em;
+        }
+
+        p {
+            margin: 0 0 12px 0;
+        }
+
+        code {
+            background: var(--anndata-code-bg);
+            padding: 2px 6px;
+            border-radius: 3px;
+            font-family: var(--anndata-font-mono);
+            font-size: 0.9em;
+        }
+
         pre {
-            margin: 0;
-            white-space: pre-wrap;
-            word-wrap: break-word;
+            background: var(--anndata-code-bg);
+            padding: 12px;
+            border-radius: var(--anndata-radius);
+            overflow-x: auto;
+            margin: 12px 0;
             font-family: var(--anndata-font-mono);
             font-size: 0.9em;
             line-height: 1.5;
+            white-space: pre-wrap;
+            word-wrap: break-word;
+
+            code {
+                background: none;
+                padding: 0;
+            }
+        }
+
+        ul,
+        ol {
+            margin: 0 0 12px 24px;
+            padding-left: 0;
+            list-style-position: outside;
+        }
+
+        li {
+            margin: 4px 0;
+            padding-left: 4px;
+        }
+
+        a {
+            color: var(--anndata-link-color);
+            text-decoration: none;
+
+            &:hover {
+                text-decoration: underline;
+            }
+        }
+
+        blockquote {
+            margin: 12px 0;
+            padding: 8px 16px;
+            border-left: 3px solid var(--anndata-accent-color);
+            background: var(--anndata-bg-secondary);
+            color: var(--anndata-text-secondary);
+        }
+
+        table {
+            border-collapse: collapse;
+            margin: 12px 0;
+            font-size: 0.9em;
+            width: auto;
+        }
+
+        th,
+        td {
+            border: 1px solid var(--anndata-border-color);
+            padding: 6px 12px;
+            text-align: left;
+        }
+
+        th {
+            background: var(--anndata-bg-secondary);
+            font-weight: 600;
+        }
+
+        tbody tr:hover {
+            background: var(--anndata-bg-secondary);
+        }
+
+        hr {
+            border: none;
+            border-top: 1px solid var(--anndata-border-color);
+            margin: 12px 0;
+        }
+
+        img {
+            max-width: 100%;
         }
     }
 
diff --git a/src/anndata/_repr/static/repr.js b/src/anndata/_repr/static/repr.js
index 9d9747d8b..3e131b267 100644
--- a/src/anndata/_repr/static/repr.js
+++ b/src/anndata/_repr/static/repr.js
@@ -462,12 +462,17 @@ if (readmeIcon) {
         closeBtn.setAttribute("aria-label", "Close");
         header.appendChild(closeBtn);
 
-        // Content — plain text (no markdown parsing, XSS-safe via textContent)
+        // Content — rendered markdown (innerHTML) or plain text (textContent)
         const content = document.createElement("div");
         content.className = "anndata-readme__content";
-        const pre = document.createElement("pre");
-        pre.textContent = readmeContent;
-        content.appendChild(pre);
+        const readmeFormat = readmeIcon.dataset.readmeFormat;
+        if (readmeFormat === "html") {
+            content.innerHTML = readmeContent;
+        } else {
+            const pre = document.createElement("pre");
+            pre.textContent = readmeContent;
+            content.appendChild(pre);
+        }
 
         modal.appendChild(header);
         modal.appendChild(content);
diff --git a/src/anndata/_repr/utils.py b/src/anndata/_repr/utils.py
index fee7b33f0..c0d52a5cd 100644
--- a/src/anndata/_repr/utils.py
+++ b/src/anndata/_repr/utils.py
@@ -511,6 +511,40 @@ def escape_html(text: str) -> str:
     return html.escape(str(text))
 
 
+def render_markdown(text: str) -> str | None:
+    """Render markdown to HTML using any available renderer.
+
+    Tries common markdown libraries in order of likelihood to be installed
+    in a Jupyter environment. Returns ``None`` if no renderer is available,
+    signaling that the caller should fall back to plain text.
+
+    Raw HTML in the markdown source is escaped by each renderer to prevent
+    XSS (e.g. ``<script>`` tags become ``&lt;script&gt;`` in the output).
+    """
+    # markdown-it-py: dependency of rich (pip, hatch, many CLI tools)
+    try:
+        from markdown_it import MarkdownIt
+
+        # html=False escapes raw HTML tags in the source;
+        # enable table and strikethrough for GitHub-flavored markdown
+        md = MarkdownIt("commonmark", {"html": False})
+        md.enable("table")
+        md.enable("strikethrough")
+        return md.render(text)
+    except ImportError:
+        pass
+
+    # mistune: dependency of nbconvert (Jupyter)
+    try:
+        import mistune
+
+        return mistune.create_markdown(escape=True)(text)
+    except ImportError:
+        pass
+
+    return None
+
+
 def sanitize_for_id(text: str) -> str:
     """Sanitize a string for use as an HTML id attribute."""
     # Replace non-alphanumeric chars with underscore
diff --git a/tests/repr/test_repr_core.py b/tests/repr/test_repr_core.py
index 5a7e04a26..ad1f8ab60 100644
--- a/tests/repr/test_repr_core.py
+++ b/tests/repr/test_repr_core.py
@@ -774,13 +774,16 @@ def test_readme_tooltip_truncated(self, validate_html):
         v.assert_text_visible("...")
 
     def test_readme_data_attribute_contains_content(self, validate_html):
-        """Test data-readme attribute contains full content."""
+        """Test data-readme attribute contains content and format flag."""
         adata = AnnData(np.zeros((10, 5)))
         adata.uns["README"] = "Test content"
         html = adata._repr_html_()
         v = validate_html(html)
         v.assert_element_exists(".anndata-readme__icon")
-        v.assert_attribute_value(".anndata-readme__icon", "data-readme", "Test content")
+        # Content is either rendered markdown (html) or plain text
+        assert "Test content" in v.html
+        # Format flag must be present
+        assert 'data-readme-format="html"' in v.html or 'data-readme-format="text"' in v.html
 
     def test_readme_icon_accessibility(self, validate_html):
         """Test readme icon has accessibility attributes."""
diff --git a/tests/visual_inspect_repr_html.py b/tests/visual_inspect_repr_html.py
index ab7ed17b9..1d8c93a1e 100644
--- a/tests/visual_inspect_repr_html.py
+++ b/tests/visual_inspect_repr_html.py
@@ -2111,8 +2111,8 @@ def format(self, obj, context):
         "17. README Icon",
         adata_readme._repr_html_(),
         "When <code>uns['README']</code> contains a string, a small ⓘ icon appears in the header. "
-        "Click the icon to open a modal with the README content displayed as plain text "
-        "(raw markdown source, not rendered). Press Escape or click outside to close.",
+        "Click the icon to open a modal with rendered markdown (if markdown-it-py or mistune "
+        "is installed) or plain text as fallback. Press Escape or click outside to close.",
     ))
 
     # Test 18: README icon in No-JS mode
@@ -2915,7 +2915,7 @@ def __repr__(self):
         "</ul>"
         "<b>Evil README (click icon to open modal):</b><br>"
         "<ul>"
-        "<li>README is displayed as plain text via textContent, so no vectors can fire</li>"
+        "<li>README is rendered as markdown (if renderer available) with raw HTML escaped</li>"
         "<li>Contains: script tags, event handlers, style injection, closing tags</li>"
         "<li>Unicode: RTL override, null bytes, emoji</li>"
         "<li>Template injection attempts, 50KB size bomb</li>"