diff --git a/src/lib/ai/types.ts b/src/lib/ai/types.ts
index bf4b47fc2..4dcbdc1bf 100644
--- a/src/lib/ai/types.ts
+++ b/src/lib/ai/types.ts
@@ -75,7 +75,16 @@ export const insightRecommendationSchema = z.union([
   }),
 ]);
 
-export const insightResultSchema = z.object({
+/**
+ * The fields every insight reply must carry. Loose on purpose: the briefing
+ * prompt also asks for `dailyBriefing`, `trendAnnotations` and
+ * `storyboardAnnotations`, each read downstream (the briefing card, the
+ * dashboard snapshot, the today read), and a strict object dropped exactly
+ * those from a reply that validated, while a reply that did not validate
+ * kept them. Validation checks these fields and leaves the rest of the reply
+ * as it came.
+ */
+export const insightResultSchema = z.looseObject({
   insightType: z.string().optional(),
   summary: z.string(),
   classification: z.enum([
diff --git a/src/lib/insights/briefing-request/__tests__/parse-briefing-reply.test.ts b/src/lib/insights/briefing-request/__tests__/parse-briefing-reply.test.ts
new file mode 100644
index 000000000..68de24fa8
--- /dev/null
+++ b/src/lib/insights/briefing-request/__tests__/parse-briefing-reply.test.ts
@@ -0,0 +1,37 @@
+/**
+ * A reply that fits the insight schema keeps the blocks the prompt asks for
+ * beside it. The strict object used to drop `dailyBriefing` from exactly the
+ * replies that validated.
+ */
+import { describe, expect, it } from "vitest";
+
+import { parseBriefingReply } from "../grounding";
+
+const VALID = {
+  summary: "Steady week.",
+  classification: "gut",
+  findings: [],
+  correlations: [],
+  recommendations: [],
+  dataQuality: { coverage: "good", gaps: [], confidence: "hoch" },
+  disclaimer: "Consult your doctor.",
+};
+
+const BLOCKS = {
+  dailyBriefing: { paragraph: "Readings held steady.", signalsOfDay: [] },
+  trendAnnotations: [{ metric: "weight", note: "flat" }],
+  storyboardAnnotations: [],
+};
+
+describe("parseBriefingReply", () => {
+  it("keeps the briefing blocks of a reply that validates", () => {
+    const parsed = parseBriefingReply(JSON.stringify({ ...VALID, ...BLOCKS }));
+    expect(parsed).toMatchObject(BLOCKS);
+    expect(parsed).toMatchObject({ summary: "Steady week." });
+  });
+
+  it("keeps a reply that does not validate as it was parsed", () => {
+    const loose = { summary: "No classification.", ...BLOCKS };
+    expect(parseBriefingReply(JSON.stringify(loose))).toEqual(loose);
+  });
+});
diff --git a/src/lib/insights/briefing-request/completion.ts b/src/lib/insights/briefing-request/completion.ts
index a63f86804..923c29720 100644
--- a/src/lib/insights/briefing-request/completion.ts
+++ b/src/lib/insights/briefing-request/completion.ts
@@ -45,8 +45,8 @@ export function runBriefingCall(
     // `getInsightsSystemPrompt` returned the old `{changed, stable, drivers,
     // …}` shape and never asked the model for dailyBriefing, so the hero
     // strip's "Re-run analysis" button kept producing a payload with no
-    // briefing block. The reply tolerates the strict shape via
-    // `insightResultSchema.safeParse`'s soft fallback to the parsed object.
+    // briefing block. `insightResultSchema` is loose, so a reply that
+    // validates keeps these blocks, and one that does not is kept as parsed.
     systemPrompt: buildSystemPromptWithReferences(
       ctx.locale,
       prompt.referenceMetrics,
diff --git a/src/lib/insights/comprehensive-generate.ts b/src/lib/insights/comprehensive-generate.ts
index 137f488aa..8ae3feb59 100644
--- a/src/lib/insights/comprehensive-generate.ts
+++ b/src/lib/insights/comprehensive-generate.ts
@@ -342,8 +342,8 @@ async function rerollBriefingParagraph(args: {
 
 /**
  * Parse + (best-effort) validate a comprehensive provider reply. Returns the
- * parsed payload (validated when it matches `insightResultSchema`, else the
- * raw object so the richer optional blocks survive), or null when the reply
+ * parsed payload (validated when it matches `insightResultSchema`, which
+ * keeps the richer optional blocks, else the raw object), or null when the reply
  * is not valid JSON even after fence-stripping. Null signals the caller to
  * run its one corrective retry.
  */
diff --git a/tests/integration/routes/parity/__snapshots__/insights-generate.P11-strict-reply.json.snap b/tests/integration/routes/parity/__snapshots__/insights-generate.P11-strict-reply.json.snap
index 646d8ce72..5a2a4783c 100644
--- a/tests/integration/routes/parity/__snapshots__/insights-generate.P11-strict-reply.json.snap
+++ b/tests/integration/routes/parity/__snapshots__/insights-generate.P11-strict-reply.json.snap
@@ -14,6 +14,19 @@
           "confidence": "hoch"
         },
         "disclaimer": "Consult your doctor.",
+        "dailyBriefing": {
+          "paragraph": "Your readings held steady this week, with no surprises.",
+          "signalsOfDay": [
+            {
+              "sourceMetric": "weight",
+              "tone": "info",
+              "headline": "Weight is steady",
+              "nudge": "Keep weighing in at the same time each morning.",
+              "delta": null
+            }
+          ],
+          "keyFindings": []
+        },
         "briefingGeneratedAt": "anchor+0d+12:00:00"
       },
       "cached": false,
@@ -222,6 +235,19 @@
         "confidence": "hoch"
       },
       "disclaimer": "Consult your doctor.",
+      "dailyBriefing": {
+        "paragraph": "Your readings held steady this week, with no surprises.",
+        "signalsOfDay": [
+          {
+            "sourceMetric": "weight",
+            "tone": "info",
+            "headline": "Weight is steady",
+            "nudge": "Keep weighing in at the same time each morning.",
+            "delta": null
+          }
+        ],
+        "keyFindings": []
+      },
       "briefingGeneratedAt": "anchor+0d+12:00:00"
     },
     "snapshotHash": "4ac15bc271309c63b3b5d1e43a3367f7abdf952d3f4351bb1b22aa364bcb630a",