hotfix: GSC Dataset warnings (cherry-pick from dev)

src/app/alternatives/[slug]/page.tsx

@@ -12,6 +12,7 @@ import { getBenchCreatedAt } from "@/lib/seo/bench-dates";
 import { capDescription } from "@/lib/seo-text";
 import { SectionLabel, SummaryStat } from "@/components/summary-stat";
 import { SITE } from "@/data/site";
+import { CREATOR_PUBLISHER, CITABLE_JSON_URL, DATASET_LICENSE, HF_DATASET_URL } from "@/lib/dataset-jsonld";
 import { loadAlternative } from "@/lib/alternatives";
 import { Breadcrumb } from "@/components/breadcrumb";
 import { buildBreadcrumbJsonLd, safeJsonLd } from "@/lib/jsonld";
@@ -120,14 +121,27 @@ export default async function AlternativePage({
           bench.metric,
           ...bench.results.map((r) => r.name),
         ].join(", "),
-        creator: { "@id": `${SITE.url}/#org` },
-        publisher: { "@id": `${SITE.url}/#org` },
+        creator: CREATOR_PUBLISHER,
+        publisher: CREATOR_PUBLISHER,
         isAccessibleForFree: true,
-        license: "https://creativecommons.org/licenses/by/4.0/",
+        license: DATASET_LICENSE,
         datePublished: getBenchCreatedAt(bench.slug).toISOString(),
         ...(citableAsOf(bench) ? { dateModified: bench.lastRunAt } : {}),
         variableMeasured: bench.metric,
         isBasedOn: benchUrl,
+        distribution: [
+          {
+            "@type": "DataDownload",
+            encodingFormat: "application/json",
+            contentUrl: `${SITE.url}/api/stat/${bench.slug}`,
+          },
+          {
+            "@type": "DataDownload",
+            encodingFormat: "application/json",
+            contentUrl: CITABLE_JSON_URL,
+          },
+        ],
+        sameAs: [HF_DATASET_URL],
       },
       {
         "@type": "Article",

src/app/benchmarks/[slug]/page.tsx

@@ -325,36 +325,34 @@ export default async function BenchmarkPage({
           }
         : null,
   });
-  const datasetNode = {
-    ...buildBenchDatasetJsonLd({
-      slug: benchmark.slug,
-      name: benchmark.seoTitle ?? benchmark.title,
-      alternateName: benchmark.title,
-      // Google Rich Results validator caps description at ~1000 chars even
-      // though schema.org Dataset allows up to 5000. Keep it under 990 to
-      // avoid the "Invalid string length" warning that strips rich snippets.
-      description: capDescription(benchmark.abstract, 990),
-      url: benchmarkUrl,
-      variableMeasured,
-      category: benchmark.category,
-      datePublished: getBenchCreatedAt(benchmark.slug).toISOString(),
-      // For drafts (no measurement history) skip dateModified so the
-      // structured-data channel does not spoof freshness. Same rule
-      // applied to /api/citable, /api/stat and the MCP surface.
-      ...(citableAsOf(benchmark)
-        ? { dateModified: benchmark.lastRunAt }
-        : {}),
-      measurementTechnique: benchmark.methodology.join(" "),
-    }),
-    // Re-bind creator + publisher to the global @id reference so the bench
-    // Dataset resolves to the same Organization node emitted by layout.tsx
-    // when crawlers stitch the site graph back together. The helper sets
-    // an inline Organization for standalone consumption; the page-level
-    // override is the more accurate shape on a site that already declares
-    // the Organization globally.
-    creator: { "@id": `${SITE.url}/#org` },
-    publisher: { "@id": `${SITE.url}/#org` },
-  };
+  // Keep the inline creator/publisher from buildBenchDatasetJsonLd. An
+  // earlier override rebound them to `{ "@id": ... }` references
+  // pointing at the Organization emitted by layout.tsx, but Google
+  // Search Console validates each Dataset in isolation - it does not
+  // stitch cross-document @id refs - and flagged the reference-only
+  // shape as "missing field creator" on /benchmarks/*. The inline Org
+  // still shares the `@id` with the layout node so JSON-LD stitchers
+  // (Perplexity, StatisticalReport isBasedOn) resolve them as one.
+  const datasetNode = buildBenchDatasetJsonLd({
+    slug: benchmark.slug,
+    name: benchmark.seoTitle ?? benchmark.title,
+    alternateName: benchmark.title,
+    // Google Rich Results validator caps description at ~1000 chars even
+    // though schema.org Dataset allows up to 5000. Keep it under 990 to
+    // avoid the "Invalid string length" warning that strips rich snippets.
+    description: capDescription(benchmark.abstract, 990),
+    url: benchmarkUrl,
+    variableMeasured,
+    category: benchmark.category,
+    datePublished: getBenchCreatedAt(benchmark.slug).toISOString(),
+    // For drafts (no measurement history) skip dateModified so the
+    // structured-data channel does not spoof freshness. Same rule
+    // applied to /api/citable, /api/stat and the MCP surface.
+    ...(citableAsOf(benchmark)
+      ? { dateModified: benchmark.lastRunAt }
+      : {}),
+    measurementTechnique: benchmark.methodology.join(" "),
+  });
   // StatisticalReport companion. Only emitted when we have a defensible
   // leader so the shape never publishes an Observation with a fabricated
   // measured value. temporalCoverage uses the ISO 8601 interval form:

src/app/compare/[slug]/page.tsx

@@ -17,6 +17,7 @@ import { capDescription } from "@/lib/seo-text";
 import { Breadcrumb } from "@/components/breadcrumb";
 import { buildBreadcrumbJsonLd, safeJsonLd } from "@/lib/jsonld";
 import { SITE } from "@/data/site";
+import { CREATOR_PUBLISHER, DATASET_LICENSE } from "@/lib/dataset-jsonld";
 import type { Benchmark } from "@/types/benchmark";
 import {
   computeInputsHash,
@@ -711,8 +712,10 @@ export default async function ComparePage({
     name: `${a.name} vs ${b.name} OpenChainBench measurements`,
     description: `Side by side live measurements for ${a.name} and ${b.name} on ${shared.length} shared OpenChainBench benchmarks.`,
     url,
-    creator: { "@id": `${SITE.url}/#org` },
-    license: "https://creativecommons.org/licenses/by/4.0/",
+    creator: CREATOR_PUBLISHER,
+    publisher: CREATOR_PUBLISHER,
+    isAccessibleForFree: true,
+    license: DATASET_LICENSE,
     measurementTechnique: `${SITE.url}/methodology`,
     variableMeasured: shared.map((s) => ({
       "@type": "PropertyValue",

src/app/products/[slug]/page.tsx

@@ -15,6 +15,7 @@ import {
 } from "@/data/provider-registry";
 import { Breadcrumb } from "@/components/breadcrumb";
 import { buildBreadcrumbJsonLd, safeJsonLd } from "@/lib/jsonld";
+import { CREATOR_PUBLISHER, CITABLE_JSON_URL, DATASET_LICENSE } from "@/lib/dataset-jsonld";
 import {
   fetchHlBuilderStats,
   isHlBuilderSlug,
@@ -445,8 +446,22 @@ export default async function ProviderPage({
           name: a.benchmark.title,
           description: capDescription(a.benchmark.subtitle, 990),
           url: `${SITE.url}/benchmarks/${a.benchmark.slug}`,
-          creator: { "@id": `${SITE.url}/#org` },
-          license: "https://creativecommons.org/licenses/by/4.0/",
+          creator: CREATOR_PUBLISHER,
+          publisher: CREATOR_PUBLISHER,
+          isAccessibleForFree: true,
+          license: DATASET_LICENSE,
+          distribution: [
+            {
+              "@type": "DataDownload",
+              encodingFormat: "application/json",
+              contentUrl: `${SITE.url}/api/stat/${a.benchmark.slug}`,
+            },
+            {
+              "@type": "DataDownload",
+              encodingFormat: "application/json",
+              contentUrl: CITABLE_JSON_URL,
+            },
+          ],
         })),
       },
       // NB: previously emitted a SoftwareApplication node here, but Google's

src/lib/dataset-jsonld.ts

@@ -60,13 +60,25 @@ const KEYWORDS = [
   "infrastructure",
 ];
 
-const CREATOR_PUBLISHER = {
+/** Embedded Org node used for Dataset `creator` / `publisher`. Exported
+ *  so pages emitting a Dataset outside the home graph can inline the
+ *  full node (Google Search Console flags an `@id`-only reference as
+ *  "missing field creator" when the referenced Org isn't declared on
+ *  the same page). */
+export const CREATOR_PUBLISHER = {
   "@type": "Organization",
   "@id": `${SITE.url}/#org`,
   name: SITE.name,
   url: SITE.url,
 } as const;
 
+/** Full JSON download of the current benchmark corpus. Used as the
+ *  default `distribution.contentUrl` for Dataset nodes emitted outside
+ *  `/benchmarks/[slug]` (products, alternatives, compare) so every
+ *  Dataset carries a machine-readable download link, satisfying the
+ *  Google Dataset "encodingFormat" / "contentUrl" requirements. */
+export const CITABLE_JSON_URL = `${SITE.url}/api/citable`;
+
 /**
  * Site-wide Dataset entry. Emitted on the home page so Google Dataset
  * Search and Perplexity have a single canonical record pointing at both