Run a technical SEO audit with Claude

The prompt works from the outside in: if the domain does not resolve or the certificate fails, nothing further down matters.

1. DNS. check_dns reads the public records for the bare domain and www: A and AAAA (where the site points), CNAME, and TXT (where the Search Console verification record usually lives). A www host that resolves but is not served correctly is a common, silent leak. Free.
2. TLS. check_tls checks each host's certificate: issuer, expiry date, hostnames covered and whether the chain is trusted. An expired or mismatched certificate stops Googlebot and visitors before any redirect runs. Free.
3. Redirects. check_redirects follows http and https, www and bare domain, hop by hop. All four should land on one canonical home page in a single permanent redirect. Free.
4. Robots. check_robots fetches robots.txt and tests real URLs against Googlebot's rules, showing the exact matching line. Free.
5. Sitemaps. list_sitemaps shows each submitted sitemap with errors, warnings and the date Google last downloaded it. One data pull.
6. Index status. inspect_url asks Google about specific pages: indexed or not, last crawl, robots and fetch result, and which canonical Google picked. One data pull per URL, so the prompt keeps it to five.
7. Speed. page_speed runs PageSpeed Insights: real-user Core Web Vitals when Google has enough traffic data for the page, plus the Lighthouse lab score and the top opportunities. One data pull per URL.

Eight data pulls fits inside the free plan's 15 a day. For a new site, run it once at launch and again after big releases. If you audit client sites every week, the numbers add up fast; Search Console MCP for agencies covers the plan that suits that work.

A real run on mcpsearchconsole.com on 2 October 2026. It inspected the home page and four blog pages and measured the home page speed: 7 data pulls in total.

The fix list the agent produced, in order:

1. Fix www (hosting). Issue a certificate that covers www.mcpsearchconsole.com and redirect www to the bare domain in one 301 or 308. Today any link to a www address ends in an error.
2. Cut home page JavaScript (developer). A lab LCP of 8.5 seconds on mobile is far above Google's 2.5 second "good" threshold. Start with the 349 KiB of unused JavaScript.
3. Resubmit the sitemap (SEO). Google last read it three months ago. Resubmit so new pages are found through it, not only through links.
4. Pick one trailing slash form (developer). Redirect /blog/ to /blog. Low risk today, since Google already picked the right one.
5. Confirm certificate auto-renewal (hosting). 45 days left is normal for Let's Encrypt, which renews automatically when configured; check that it is.

Sort every finding into one of four tiers, and fix the tiers in order. Within a tier, fix what affects the most traffic first.

• Tier 1: Google cannot reach or index the page. Certificate errors, DNS failures, robots.txt blocking important URLs, noindex on pages you want in search, 5xx errors. Fix these the same day; nothing else on the list matters until they are gone.
• Tier 2: links and signals are leaking. Redirect chains, temporary redirects that should be permanent, hosts that do not reach the canonical domain, Google-selected canonicals that differ from yours. These cost link equity and confuse which URL ranks. Fix within the week.
• Tier 3: the page loads slowly. Core Web Vitals outside Google's "good" thresholds (LCP over 2.5 s, INP over 200 ms, CLS over 0.1). Prefer field data; if there is none, use the lab numbers as a direction, not a verdict. Plan the work with a developer.
• Tier 4: housekeeping. Sitemaps not downloaded recently, trailing slash duplicates Google already resolved, certificates within 30 days of expiry on auto-renewal. Batch these into the next release.

For a client audit, hand over the table and the fix list together, with an owner per line (developer, hosting, content), and re-run the same prompt after the fixes ship. The second run is the proof the work was done.

• It is not a full crawl. The agent checks the pages you name and the ones in your sitemap. It will not find every broken internal link, orphan page or duplicate title across thousands of URLs; use a crawler for that and give the agent its export.
• Index status reflects Google's last crawl. A fix deployed today shows in the free live checks at once, but in inspect_url only after Google recrawls.
• Field speed data needs traffic. New and small sites often have no real-user Core Web Vitals, as in our own run. Lab scores vary between runs and from real visitors' experience.
• No server logs. The audit cannot see how often Googlebot actually hits each URL or which errors it met in between. Search Console's Crawl stats report and your logs cover that.
• Structured data and content quality are out of scope. inspect_url reports rich result status for a URL, but the audit does not judge whether your markup or content is good.