diff --git a/scripts/generate_site.py b/scripts/generate_site.py
index ab49e477..d4313b05 100644
--- a/scripts/generate_site.py
+++ b/scripts/generate_site.py
@@ -1373,10 +1373,26 @@ def _plain_markdown(text: str) -> str:
     text = re.sub(r"<[^>]+>", " ", text)
     text = re.sub(r"!\[([^]]*)\]\([^)]*\)", r"\1", text)
     text = re.sub(r"\[([^]]+)\]\([^)]*\)", r"\1", text)
-    text = re.sub(r"[`*_~]", "", text)
+    # Code spans keep their text whole: validate_schemas.py read
+    # validateschemas.py once every underscore was taken for emphasis.
+    pieces = re.split(r"(`[^`]*`)", text)
+    text = "".join(
+        piece.strip("`") if piece.startswith("`")
+        else re.sub(r"(?<!\w)_+|_+(?!\w)", "", re.sub(r"[*~]", "", piece))
+        for piece in pieces
+    )
     return re.sub(r"\s+", " ", text).strip()
 
 
+def _shorten(plain: str, limit: int = 157) -> str:
+    """A description cut at a word, never inside one: six snippets ended
+    in "op..." or "Fo..."."""
+    if len(plain) <= limit:
+        return plain
+    cut = plain[: limit + 1].rsplit(" ", 1)[0] if " " in plain[:limit] else plain[:limit]
+    return cut.rstrip(" ,;:-(.") + "..."
+
+
 def _page_description(markdown: str, relative: Path, title: str) -> str:
     key = relative.as_posix()
     specific = {
@@ -1398,6 +1414,12 @@ def _page_description(markdown: str, relative: Path, title: str) -> str:
         about = PAGE_DATA.get(key, {}).get("about", {})
         pinned = any(a.get("version") for a in (about if isinstance(about, list) else [about]))
         kind = "source-pinned emulator firmware profile" if pinned else "emulator firmware profile"
+        if not PAGE_DATA.get(key, {}).get("hasPart"):
+            # 152 cores that load no file announced paths and hashes.
+            kind = kind.replace(" firmware", "")
+            trees = (" The data directories it reads are listed."
+                     if "\n## Data directories\n" in markdown else "")
+            return f"{title}: {kind}. No BIOS or firmware file is required.{trees}"
         return f"{title}: {kind} with paths, hashes, requirements and validation behavior."
     if key.startswith("systems/") and relative.stem != "index":
         return f"{title}: indexed firmware files, hashes, variants, provenance and platform or emulator usage."
@@ -1412,7 +1434,7 @@ def _page_description(markdown: str, relative: Path, title: str) -> str:
             continue
         plain = _plain_markdown(stripped)
         if len(plain) >= 35:
-            return plain[:157].rstrip(" ,;:-") + ("..." if len(plain) > 157 else "")
+            return _shorten(plain)
     return f"{title}: RetroBIOS source-traced retrogaming firmware reference."
 
 
@@ -1733,7 +1755,9 @@ def decorate_markdown_pages(docs: Path) -> None:
             page_path = "" if len(relative.parts) == 1 else "/".join(relative.parts[:-1]) + "/"
         else:
             page_path = relative.with_suffix("").as_posix() + "/"
-        canonical = urllib.parse.urljoin(SITE_URL, page_path)
+        # Encoded as mkdocs encodes the canonical link and the sitemap:
+        # idtech4a++ named .../idtech4a++/ here and .../idtech4a%2B%2B/ there.
+        canonical = urllib.parse.urljoin(SITE_URL, urllib.parse.quote(page_path))
         schema_type = "WebSite" if relative.as_posix() == "index.md" else "TechArticle"
         if relative.as_posix() == "data.md":
             schema_type = "Dataset"
diff --git a/tests/test_site_emulator_page.py b/tests/test_site_emulator_page.py
index 8f0e60af..fd364aad 100644
--- a/tests/test_site_emulator_page.py
+++ b/tests/test_site_emulator_page.py
@@ -69,5 +69,49 @@ class WhoShipsTheBuildAndWhy(unittest.TestCase):
                       "found only as rebuilt copies) |", page)
 
 
+class WhatASearchSnippetSays(unittest.TestCase):
+    def test_a_core_loading_no_file_promises_no_hash(self):
+        """152 pages of cores that load no file announced paths and hashes."""
+        rel = Path("emulators/lutro.md")
+        saved = dict(gs.PAGE_DATA)
+        self.addCleanup(lambda: (gs.PAGE_DATA.clear(), gs.PAGE_DATA.update(saved)))
+        gs.PAGE_DATA[rel.as_posix()] = {}
+        text = gs._page_description("# Lutro\n", rel, "Lutro")
+        self.assertNotIn("hashes", text)
+        self.assertIn("No BIOS or firmware file is required.", text)
+        gs.PAGE_DATA[rel.as_posix()] = {"hasPart": [{"name": "a.bin"}]}
+        self.assertIn("paths, hashes", gs._page_description("# Lutro\n", rel, "Lutro"))
+
+    def test_a_long_paragraph_is_cut_at_a_word(self):
+        plain = ("files/?sha1= is the address that does not move: it opens the card "
+                 "wherever the file is now, " * 4)
+        text = gs._shorten(plain)
+        self.assertTrue(text.endswith("..."))
+        self.assertLessEqual(len(text), 160)
+        self.assertTrue(plain.startswith(text[:-3]))
+        self.assertIn(plain[len(text) - 3], " ,:")
+
+    def test_code_keeps_its_underscores(self):
+        self.assertEqual(gs._plain_markdown("only `validate_schemas.py` and _this_"),
+                         "only validate_schemas.py and this")
+
+    def test_structured_url_is_encoded_like_the_canonical(self):
+        """mkdocs encodes idtech4a++ as idtech4a%2B%2B in the canonical link
+        and the sitemap, and the JSON-LD named the unencoded URL."""
+        import json
+        import tempfile
+
+        with tempfile.TemporaryDirectory(dir=REPO_ROOT / "tmp") as tmp:
+            page = Path(tmp) / "emulators" / "idtech4a++.md"
+            page.parent.mkdir()
+            page.write_text("# idTech4A++\n\nSome text long enough to be a description.\n")
+            gs.decorate_markdown_pages(Path(tmp))
+            text = page.read_text()
+        data = json.loads(text.split('<script type="application/ld+json">')[1]
+                          .split("</script>")[0].replace("<\\/", "</"))
+        self.assertEqual(data["url"],
+                         "https://abdess.github.io/retrobios/emulators/idtech4a%2B%2B/")
+
+
 if __name__ == "__main__":
     unittest.main()