diff --git a/src/server/lib/audit/discovery.test.ts b/src/server/lib/audit/discovery.test.ts
new file mode 100644
index 000000000..761cb1685
--- /dev/null
+++ b/src/server/lib/audit/discovery.test.ts
@@ -0,0 +1,40 @@
+import { describe, expect, it } from "vitest";
+import { parseRobotsTxt } from "@/server/lib/audit/discovery";
+
+const ORIGIN = "https://example.com";
+
+describe("parseRobotsTxt", () => {
+  it("honors a group that names the crawler", () => {
+    const robots = parseRobotsTxt(
+      ORIGIN,
+      ["User-agent: OpenSEO-Audit", "Disallow: /private/"].join("\n"),
+    );
+
+    expect(robots.isAllowed(`${ORIGIN}/private/report`)).toBe(false);
+  });
+
+  it("honors an allowance the crawler is named in", () => {
+    const robots = parseRobotsTxt(
+      ORIGIN,
+      [
+        "User-agent: *",
+        "Disallow: /",
+        "",
+        "User-agent: OpenSEO-Audit",
+        "Allow: /",
+      ].join("\n"),
+    );
+
+    expect(robots.isAllowed(`${ORIGIN}/pricing`)).toBe(true);
+  });
+
+  it("falls back to the wildcard group when the crawler is not named", () => {
+    const robots = parseRobotsTxt(
+      ORIGIN,
+      ["User-agent: *", "Disallow: /admin/"].join("\n"),
+    );
+
+    expect(robots.isAllowed(`${ORIGIN}/admin/settings`)).toBe(false);
+    expect(robots.isAllowed(`${ORIGIN}/pricing`)).toBe(true);
+  });
+});
diff --git a/src/server/lib/audit/discovery.ts b/src/server/lib/audit/discovery.ts
index 1113f1af2..575bebe1f 100644
--- a/src/server/lib/audit/discovery.ts
+++ b/src/server/lib/audit/discovery.ts
@@ -5,6 +5,8 @@ import robotsParser from "robots-parser";
 import { XMLParser } from "fast-xml-parser";
 import { isSameOrigin, normalizeUrl } from "./url-utils";
 
+/** The token the crawler identifies itself with; must match crawlPage's. */
+const CRAWL_USER_AGENT = "OpenSEO-Audit/1.0";
 const SITEMAP_FETCH_TIMEOUT_MS = 15_000;
 // robots.txt is checkpointed as durable Workflow step state (~1MiB cap, shared
 // with the rest of the step's return). RFC 9309 requires parsers to handle at
@@ -40,7 +42,7 @@ export interface RobotsResult {
 async function fetchRobotsTxtText(origin: string): Promise<string | null> {
   try {
     const response = await fetch(`${origin}/robots.txt`, {
-      headers: { "User-Agent": "OpenSEO-Audit/1.0" },
+      headers: { "User-Agent": CRAWL_USER_AGENT },
       signal: AbortSignal.timeout(10_000),
     });
 
@@ -63,7 +65,11 @@ export function parseRobotsTxt(
 
   const robots = robotsParser(`${origin}/robots.txt`, text);
   return {
-    isAllowed: (url: string) => robots.isAllowed(url) ?? true,
+    // Rules are matched against the token the crawler actually sends. Without
+    // it robots-parser only reads the `User-agent: *` group, so a group naming
+    // OpenSEO-Audit — to block it, or to admit it where `*` is disallowed —
+    // has no effect.
+    isAllowed: (url: string) => robots.isAllowed(url, CRAWL_USER_AGENT) ?? true,
     sitemapUrls: robots.getSitemaps(),
   };
 }
@@ -173,7 +179,7 @@ async function fetchSitemapDocumentWithRetry(sitemapUrl: string): Promise<{
   for (let attempt = 0; attempt <= SITEMAP_RETRIES; attempt++) {
     try {
       const response = await fetch(normalizedSitemapUrl, {
-        headers: { "User-Agent": "OpenSEO-Audit/1.0" },
+        headers: { "User-Agent": CRAWL_USER_AGENT },
         signal: AbortSignal.timeout(SITEMAP_FETCH_TIMEOUT_MS),
       });