How an AI Agent Caught a Sheet Typo and Verified 3 Sites

A Claude agent hit three law-firm subdomains in a Search Console queue that simply would not load. The domains in the sheet were wrong — each missing one word. Fixing the data, not forcing the task, turned three dead ends into three instant verifications.

The fastest verification is the one where you change nothing on the site. During a fleet Google Search Console run, a Claude agent found three law-firm subdomains that all auto-verified through an existing tag container the moment the correct domain was entered. The catch: the correct domain was not the one written in the task sheet. This is the story of how the agent found the real domains, verified them for free, and fixed the source data so the mistake would not repeat.

Trust The Error, Not The Spreadsheet

Three high-priority rows pointed at subdomains of a personal-injury firm. None of them resolved. A naive run would have marked all three “unverifiable” and moved on, losing three verifications on some of the highest-traffic properties in the queue.

Instead the agent treated the non-resolution as a clue. Checking the firm’s hosting install revealed the pattern: every URL in the sheet was missing one word from the firm’s actual domain. The sheet listed a shortened version of the domain that had never existed. All three rows shared the exact same typo.

Let Google Verify For Free

Once the real domains were in hand, the agent added each as a URL-prefix property in Search Console. Because the firm’s sites already carried a shared Google Tag Manager container from prior work, Google confirmed ownership automatically. No plugin edit, no token typed into a live site, no hosting login required for the verification itself.

This is the payoff of checking Search Console before touching the site. The auto-check catches every property that already carries a tag or Analytics container, and it does it in seconds. It is the same efficiency principle behind our approach to running weekly AI agent SEO audits across all clients: spend effort only where the cheap signal fails.

Fix The Root, Not Just The Symptom

Verifying the three sites was only half the job. The agent also corrected the URL in each of the three sheet rows and left a dated note explaining the change. That way the typo does not resurface the next time someone runs the queue. Fixing the data at the root is exactly the discipline that separates real optimization from one-click tools, the same theme running through better internal link building with an AI agent — the agent has to understand the business, not just match strings.

The Critical Decisions

1. Read the non-resolution as data, not failure. Three dead URLs became three verifications because the agent investigated instead of logging “unverifiable.” The alternative was three lost properties.

2. Recognize the shared pattern. After the first typo, the agent checked whether the other two shared it. They did. Spotting the systemic error saved two more investigations. The alternative was treating three related errors as three unrelated ones.

3. Correct the source, not just the outcome. The agent wrote the fix back into the sheet with a dated note. The alternative left a landmine for the next run.

Effort And Cost Comparison

What The Agent Could And Could Not Do

The agent investigated the broken URLs, identified the correct domains, added the Search Console properties, confirmed auto-verification, and corrected the source sheet. It did not need any site login because verification happened through the existing tag container. The only human dependency was the standing rule to review data corrections before they ship — which this draft respects.

Information Ingestion Inventory

Guidelines Compliance Scorecard