seo batch B: answers OG + compare visible FAQ + bench grounding-first

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

@@ -75,14 +75,25 @@ export async function generateMetadata({
       type: "article",
       url,
       siteName: SITE.name,
-      images: [{ url: `${SITE.url}/opengraph-image`, width: 1200, height: 630 }],
+      // Use the referenced bench's OG card instead of the site-wide
+      // default. Perplexity / ChatGPT / Claude Deep Research scrape the
+      // og:image alongside the answer text; showing the bench-specific
+      // leader card gives the AI a real datapoint next to the answer
+      // rather than the generic homepage tile.
+      images: [
+        {
+          url: `${SITE.url}/api/og/${ans.bench.slug}`,
+          width: 1200,
+          height: 630,
+        },
+      ],
     },
     twitter: {
       card: "summary_large_image",
       site: SITE.twitter,
       title,
       description,
-      images: [`${SITE.url}/twitter-image`],
+      images: [`${SITE.url}/api/og/${ans.bench.slug}`],
     },
   };
 }

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

@@ -507,9 +507,24 @@ export default async function BenchmarkPage({
       <h1 className="mt-5 display text-3xl sm:text-4xl md:text-5xl tracking-tight text-ink break-words">
         {benchmark.title}
       </h1>
-      <p className="mt-4 max-w-3xl text-lg sm:text-xl text-ink-muted leading-snug break-words">
-        {benchmark.subtitle}
-      </p>
+      {/* Grounding-line first-under-H1: Perplexity / ChatGPT search /
+          Claude Deep Research scrape the first prose sentence after the
+          H1 as the "citable claim." Previously the methodology-flavoured
+          subtitle sat here; the current-leader sentence carries a
+          concrete provider + number + unit that AI answer engines quote
+          verbatim ("Codex leads head lag at 0.8 s (p50, 24h)..."). Fall
+          back to the subtitle when no defensible leader exists (draft,
+          insufficient, cold start). */}
+      {sentence ? (
+        <p className="mt-4 max-w-3xl text-lg sm:text-xl text-ink leading-snug break-words">
+          <span className="font-medium">{sentence}</span>{" "}
+          <span className="text-ink-muted">{benchmark.subtitle}</span>
+        </p>
+      ) : (
+        <p className="mt-4 max-w-3xl text-lg sm:text-xl text-ink-muted leading-snug break-words">
+          {benchmark.subtitle}
+        </p>
+      )}
 
       {/* Compact "At a glance" card. Rendered as a native <details>
           element so it collapses to a one line summary on click while

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

@@ -912,6 +912,32 @@ export default async function ComparePage({
         </div>
       </section>
 
+      {/* Visible FAQ parity with the FAQPage JSON-LD above. Google
+          requires visible Q&A text matching the schema entries or it
+          silently drops the rich result — the compare pages had FAQ
+          jsonld without any visible questions, so the FAQPage never
+          rendered in the SERP snippet. */}
+      <section className="mt-12 max-w-3xl">
+        <h2 className="text-[11px] font-medium uppercase tracking-[0.18em] text-ink-muted">
+          Frequently asked questions
+        </h2>
+        <div className="mt-4 space-y-4">
+          {faqEntries.map((e) => (
+            <details
+              key={e.q}
+              className="border border-rule rounded-lg px-4 py-3"
+            >
+              <summary className="cursor-pointer text-sm font-medium text-ink leading-tight">
+                {e.q}
+              </summary>
+              <p className="mt-3 text-sm leading-relaxed text-ink-soft">
+                {e.a}
+              </p>
+            </details>
+          ))}
+        </div>
+      </section>
+
       <section className="mt-12 max-w-3xl">
         <h2 className="text-[11px] font-medium uppercase tracking-[0.18em] text-ink-muted">
           How this pair was selected