SEO Refresher is a free Claude Skill that refreshes one decaying blog post around the queries it already half-ranks for. It reads your Google Search Console data, finds the striking-distance cluster at positions 4 to 20, and rewrites the post to serve it: a dedicated H2, a fixed title and meta, an extended FAQ, internal links both ways.
What does the SEO Refresher skill do?
It refreshes the one post you point it at, then stops. The skill is the standalone version of the refresh loop Relato runs on relato.com, packaged as one SKILL.md instruction file for Claude Code. Five steps:
- Step 1, pick the post: You name one post by slug, URL, or file path. It never chooses for you, with one exception: a prioritized queue file with checkbox rows in your repo, where it takes the topmost unchecked row.
- Step 2, pull the cluster: It pulls 90 days of Search Console queries for that page and filters to average positions between 4 and 20, with real impressions and near-zero clicks. The highest-impression query the post does not already answer in an H2 becomes the lift target.
- Step 3, refresh: A new H2 answers the lift target, with the query in the heading where it reads naturally. Depth expands around the cluster’s adjacent questions, and the title tag, meta description, FAQ, and internal links get updated to match.
- Step 4, quality gates: Every new passage runs through the Humanizer skill, em dashes it wrote are stripped, the frontmatter has to parse, and the lift target has to appear in the new H2, the title tag, the meta description, and at least one paragraph. A draft that fails its own checks gets fixed before you see it.
- Step 5, output: One summary block: the lift target, the cluster size, the section added, title and meta before and after, and every internal link it placed. If you work in git, it commits to a refresh branch and leaves pushing to you.
What is a striking-distance refresh?
A striking-distance refresh targets queries where your page already appears in search results, at positions almost nobody clicks. The page is indexed and Google already associates it with the query. What is missing is a section that answers it. That makes this the cheapest move in SEO: no new page, no backlinks, no waiting for a fresh URL to earn trust.
The skill’s method is specific. Add the query as an H2 with a real answer under it, and put it in the first half of the title tag. Rewrite the meta description around it, at about 155 characters. Extend the FAQ. Tighten the internal links both ways. The skill’s own method note treats 4 to 8 weeks as the window in which a move like position 12 to the top half of page one shows up, so the refresh is a bet you go back and check in Search Console.
A refresh is also narrower than a rewrite. The skill expands what the post already credibly claims and leaves paragraphs outside the cluster’s scope alone, because scope creep is the most common way a refreshed post ends up worse than the original. That discipline is what makes an AI content refresh safe on a post that already ranks.
What does a before and after look like?
Here is a worked sample on a made-up post, every edit traced to a rule in the skill file. Say a post at /blog/content-audit/ has sat at position 12 for “content audit checklist” for a quarter, with impressions, almost no clicks, and no heading on the phrase.
The skill would add an H2 titled “The content audit checklist” with the actual checklist under it. Its title tag moves from “How to audit your content” to “Content audit checklist: how to audit your content”, putting the lift target in the first half. Two cluster questions join the FAQ, and the meta description gets rewritten around the phrase. Links go both ways: two to four outbound on descriptive anchors, one to three inbound from posts that mention audits without linking here.
The run ends with the skill’s literal output format, filled in:
Refreshed: /blog/content-audit/
Lift target: content audit checklist
Cluster size: 6 striking-distance queries
Section added: The content audit checklist
FAQ items added: 2
Title before: How to audit your content
Title after: Content audit checklist: how to audit your content
Meta before: Learn how to audit your website content.
Meta after: A content audit checklist for deciding what to update,
merge, or retire, with the queries each page should answer.
Outbound links added: (content inventory template, /blog/content-inventory/)
Inbound links added: (/blog/content-strategy/, content audit)
Humanizer summary: 3 em dashes, 1 banned phrase, 0 hedges
Production URL: pending deploy
The numbers are invented for the illustration. The block format is the skill’s own; on a real run every value comes from your Search Console data and your repo.
What will it refuse to do?
Plenty, on purpose. The refusals keep an automated refresh from quietly damaging a blog:
- Batching. One post per invocation. Ten refreshes means ten runs, each reviewable on its own.
- Inventing Search Console numbers. If it cannot pull the data, it asks for a 90-day CSV export or refuses the run.
- Inventing facts. No manufactured statistics, citations, quotes, or named people. The refresh deepens what the post already establishes.
- Refreshing thin posts. Under 600 words, the post comes back unchanged with a suggestion to rewrite instead.
- Chasing the wrong query. If the cluster names a query the post cannot honestly serve, wrong intent or wrong audience, it proposes a net-new piece and stops.
- Downstream work. No social threads, no emails. Those sit after editorial review, and it refuses if asked.
If its Humanizer self-check fails, it returns the draft marked work-in-progress with the failures named, and stops.
What can it not know without Search Console data?
Without Search Console, there is no cluster, and the skill refuses the run rather than guessing. The method depends on knowing which queries your page already shows for and at what position, and only your own GSC property holds that: a live connection, or a CSV export covering 90 days, grouped by query.
Three honest limits sit behind that one. Your impressions describe your property’s visibility, a floor on demand rather than the market’s search volume. The skill sees search data and your repo, so conversions stay outside its view; whether the lifted traffic is worth having is your call, made with your analytics. And position movement arrives over weeks, so the output is a reviewed draft plus a hypothesis you verify later.
How do you install the SEO Refresher skill?
You save one file and invoke the skill by name. For SEO Refresher:
- Save the SKILL.md file to
~/.claude/skills/seo-refresher/SKILL.mdon your machine. - Have your Search Console data ready: a connection, or a 90-day CSV export for the property.
- In Claude Code, invoke
/seo-refresherand point it at one post. The refreshed draft lands in your repo, ready for review.
There is no server and no account, because the skill is its instruction file. The gated Google Doc in the bundle repeats these steps with install notes.
New to Skills in general? The install guide on the skills hub covers how installation works for the whole series, including claude.ai.
Who is this Skill for?
Teams with a blog old enough to have decaying posts.
- Heads of content who want this part of the workflow to stop being negotiated piece by piece.
- Production leads who need the step to be predictable.
- Solo content marketers using Claude Code as a second pair of hands.
- Any team trying to make the LLM part of the pipeline reliable instead of magical.
What’s in the bundle?
Four things, all free:
- SKILL.md (raw): the full Claude Code Skill file, ready to drop into
~/.claude/skills/seo-refresher/. - One-pager PDF: a printable summary covering the cover, the 4-step process, what the Skill returns, and install instructions.
- Google Doc (gated): a Drive-resident copy plus install notes. Submit your email to access; we also tag you for the Claude Skills for content operations email sequence so you hear when the next Skill lands.
- Card image: the same hero you can repost when you share it.
Why we built it
Refreshing a decaying post was the piece of our own pipeline that most depended on who happened to do it. One person would fix the title, another would rewrite half the post and break what ranked. Moving the rules into a file made the step repeatable: same filters, same gates, same output block, whoever runs it. This page distributes that file.
This Skill is part of Relato’s Claude Skills for content operations series, ten Skills that together cover the content lifecycle from brief through refresh. SEO Refresher is Skill 2 of 10.
It stops at one post per run. Multiple refreshes happen by re-invoking the Skill once per post.
Frequently asked questions
What is a Claude Skill?
A Claude Skill is a reusable instruction set Claude Code loads on demand. It lives as a SKILL.md file with frontmatter and a body. You install once. Claude Code finds it next time you invoke it by name.
Do I need a paid Claude plan to use this?
You need Claude Code. The Skill itself is free.
Does it work without Google Search Console?
No. The striking-distance cluster comes from your own Search Console data, as a live connection or a 90-day CSV export grouped by query. Without that, the skill refuses the run instead of inventing numbers.
Will it invent statistics or sources?
No. A hard rule in the skill bans invented statistics, citations, and named people. The refresh expands depth on what the post already credibly claims and on what the cluster names.
Can I edit the Skill to match my team’s voice or workflow?
Yes, and we recommend it. The skill reads a writing-style.md if your project has one, and every threshold, from the position band to the FAQ range, is plain markdown you can change. Your edits are enforced on the next run.
Will more Skills follow?
Yes. The full series is ten Skills covering brief drafting, SEO refresh, FAQ generation, humanizing, internal links, brief-to-outline, editor checklist, repurposing, calendar generation, and refresh queue ranking. Submit your email above to be told when each one lands.
Ready to streamline your content operations?
Start your free trial and see the difference.