What it is
Uristocrat.com is a media site that covers sneakers, sports, culture, and technology. We post several times a day as part of the Uristocrat Knowledge Harness. Our main challenge was getting our pages indexed by Google.
Here is some background: uristocrat.com has been around since 2008, mostly as a WordPress site with lots of pages and posts. Now it runs on Ghost. Not every page migrated successfully, which led to Google indexing issues and a significant drop in search traffic. After setting up this agent, Uristocrat’s Google Search traffic increased 14 times.
This is where the Ghost SEO Skill comes in.
The SEO Fixes for Uristocrat agent uses the Ghost SEO Skill to handle these daily tasks:
- It logs into Google Search Console and pulls a list of non-indexed URLs.
- Diagnoses each one: missing metadata, broken canonical, thin content, wrong reason.
- It applies fixes directly through the Ghost MCP, editing only existing posts.
- It submits the fixed URLs for re-crawling, making sure to stay within Google’s daily limits.
- It writes a report and updates a memory file, so the next day’s run picks up right where the last one ended.
This agent only handles SEO and indexing. It never publishes or deletes posts, and it does not change the title or slug of any post that is already ranking. The edits are small and careful, focusing on rebuilding Google’s trust in the site rather than making unnecessary changes.
The three-month arc
The results cover the period from late April, when the campaign began, to July 8. Here is what the agent’s Search Console shows.
| Date | Indexed pages (GSC) |
|---|---|
| May 27 | 867 |
| June 22 | 1,174 |
| July 4–8 | 1,520 |
There were 653 more indexed pages, about a 75% increase during this period. These numbers come straight from the agent’s Search Console on the dates shown.
Here is what happened behind those numbers:
Phase one was the metadata backfill, from late April to mid-June. The site had years of posts missing meta titles, descriptions, or excerpts. The agent went through the archive in batches and filled in about 950 posts by June 12. By late June, every published post had a meta description, no matter the year. That campaign is finished, and the agent has stopped spending time on it.
Phase two was about making structural changes. Ghost automatically creates a thin archive page for every tag on the site. There were about 2,349 of these, each just a list of post titles with no content. Google was using its crawl budget on these pages instead of the real articles. The agent flagged this issue in every run for weeks. On June 18, the noindex fix was added to the tag template. In the agent’s words, it was “the biggest structural SEO win since the metadata campaign started.”
Phase three was automation and cleanup, from late June to July. On June 23, the agent started submitting re-crawl requests directly in the browser, so I no longer had to do it manually. This meant it could diagnose and resubmit in the same run, without my help. By early July, live URL checks showed that new editorial posts were being indexed the same day or the next. On July 8, the agent noindexed 70 thin-content posts, 23 sports-roundup aggregators, and 47 closed job listings to keep Google from wasting crawl budget on them.
How was this implemented?
The Ghost SEO Skill gives explicit instructions for accessing the Ghost site, writing metadata, and submitting to Google Console. That was the hard part. We installed a Ghost MCP on Uristocrat.com and set up a scheduled routine in Claude to run the agent. You can set up the whole system for a new Ghost site in just one afternoon.
Why the result is about how it runs, not what it does
Any of these fixes could be done by hand in an afternoon. The real improvement came from how the system runs, not from each single edit.
It runs daily, so the fixes compound. A quarterly SEO audit identifies many problems and fixes them all at once. This runs every morning, so a post published at midnight has complete metadata and a re-crawl request submitted before I wake up. New problems never get a chance to age.
It is designed to be careful. It only edits existing posts and never publishes, deletes, or changes the slug of a post that is already indexed. That is why I can let it run on a live site without watching it. If something goes wrong, the worst that happens is a small metadata change on one post, not a broken URL that hurts rankings.
It solves problems instead of just repeating tasks. The best thing it did was realize that the non-indexed backlog was not a metadata issue. The real problem was structural: tag archives and thin roundup posts that Google did not index because of quality. A less smart system would have kept rewriting metadata endlessly. This one knew when the metadata work was done, reported it, and pushed for the theme change that actually mattered.
It keeps track of what is done. Each run updates a memory file, so the next day it knows the tag fix went live on June 18, the metadata is complete, and which posts are already indexed. It does not waste time on problems that are already solved.
It is built to fail safely. If Chrome was not logged in or running, the agent would stop and report the problem instead of guessing at fixes without real Search Console data. I would rather it do nothing than have it make up work.
It leaves the big decisions to me. Editorial choices, such as whether to noindex the daily notes template or unpublish two empty posts from 2010, are flagged in the report but not applied automatically. The agent handles the routine work while I focus on the judgment calls.
The honest edges
The total number of non-indexed pages did not go down during this period. It actually went up, from about 3,900 to 4,350. That might sound bad, but it is because the site publishes every day and Ghost keeps making new tag pages. The agent sees this as a structural issue that is mostly intentional now: thin archive and roundup pages that Google is right to skip. The important number is that real editorial pages being indexed went up by 75%.
The last big fix, noindexing the tag and author archive templates at the theme level, is still partly up to me, since it is a theme change and an editorial choice, not just a metadata update. The agent brings this up in almost every run. That is how the system works as designed: it does what it should without me and tells me the two or three things only I should decide.
I did not hire an SEO person or spend my mornings in Search Console. I built the agent once using the Ghost SEO Skill, and it has run the job every day since. Indexed pages rose by 75% in a quarter on a live site, while I read a report instead of doing the work.
That is exactly what I do for you. Find the workflow that quietly eats your team’s time, build the agent that runs it end-to-end, and hand it back working. I ran it on my own business first.