How to build a Claude SEO skill for your workflow — and actually rank with it.

The Rank-Tracking Skill is the canonical starting point for anyone learning how to build a Claude SEO skill for their workflow. The inputs are explicit — domain, keyword list, target location, device type, result depth — and the success criteria are easy to verify by spot-checking a handful of positions manually against what Claude reports.

To build it: add the SerpAPI MCP to Claude Code by running claude mcp add serpapi --url https://mcp.serpapi.com/mcp in your terminal, then write a Skill.md that defines your keyword list, location parameters, schedule cadence, and the Markdown table format you want positions stored in. Save the resulting ZIP via Settings > Capabilities > Skills and trigger it with a single prompt. Once the skill runs cleanly for a week, document the exact instructions that produced the working result — those instructions become your reusable template for every new client domain.

Keep your first rank-tracking skill to 20–50 keywords connected to your most important pages. Verify that position numbers, tracked URLs, and device labels all appear correctly before expanding. A skill that runs reliably at small scale is worth far more than an ambitious one that silently returns wrong data at large scale.

Search Console holds more actionable intelligence than almost any paid SEO tool — and most teams look at it once a quarter. A Claude Search Console Analysis Skill changes that by turning a GSC Performance export into a structured weekly briefing: pages where clicks fell > 20% month-over-month, queries with impressions > 500 and CTR < 2%, pages ranking in positions 8–15 that are one optimisation away from page-one, and clusters of queries suggesting keyword cannibalization.

Build the skill by writing a Skill.md that defines exactly how Claude should ingest a GSC CSV export: which columns map to which analysis steps, what thresholds trigger an alert, and what the output format looks like. A well-specified skill will return a ranked opportunity table in under 60 seconds — work that would take a junior SEO two hours to replicate manually.

For deeper integration, wire the skill to the official Google Search Console MCP server, which allows Claude to pull live data without you exporting anything. This is the setup used by the most sophisticated Claude SEO workflows and pairs naturally with the MCP connection patterns covered in our Google and Meta Ads guide.

The Technical SEO Audit Skill is built around one core idea: Screaming Frog can crawl everything, but a 40,000-row export is useless without a system that tells you what to fix first. The skill ingests the export — or a subset of it — and returns a prioritised report that separates sitewide structural issues (broken internal links, redirect chains, missing canonical tags, crawl budget waste) from page-level issues (missing meta descriptions, duplicate title tags, slow Core Web Vitals).

To build it, write a Skill.md that defines: which Screaming Frog columns Claude should parse, how to classify issues by severity, and what the output sprint table should include — issue type, affected URL count, fix description, estimated effort, and estimated traffic impact. Activate it via Settings > Capabilities > Skills and run it on a fresh export each month.

The most powerful version of this skill connects to a site-crawl MCP so Claude can trigger the crawl itself rather than waiting for a manual export. That is the pattern described in the advanced MCP integration guide, and it is what makes the difference between a diagnostic tool and a true autonomous SEO agent.

The Content Optimisation Skill takes an existing article and a target keyword and returns a structured improvement plan: entities the page is missing relative to the top 10 ranking pages, heading structure tweaks, internal link opportunities from your own site, schema markup gaps, and AI-search citation readiness. One skill run replaces a manual content audit that typically takes an experienced SEO 90 minutes per article.

Build the skill by defining a clear input format in your Skill.md: the article text (pasted or fetched via URL), the primary keyword, up to five competitor URLs for entity benchmarking, and the desired output sections. Pair it with a SERP Analysis Skill (see approach #7 below) so Claude can pull live competitor data rather than relying on what you paste manually. That combination is what advanced practitioners mean when they talk about a full Claude SEO skill workflow.

For bulk optimisation across a large site, this skill pairs naturally with Ryze AI's autonomous content layer, which can prioritise which pages to optimise first based on impression data and push the changes live without a review queue.

The Topical Cluster Builder Skill turns a single seed topic into a full content ecosystem. Feed the skill your domain, a seed keyword, and your current sitemap, and it returns: a pillar page outline, a list of supporting cluster articles with target keywords and search intent labels, a proposed internal linking structure, and an analysis of which sub-topics your competitors cover that you do not.

This is the skill that separates reactive content creation — writing what feels relevant — from systematic topical authority building. The best implementations pair it with live SERP data from a SerpAPI MCP so the cluster map reflects what is actually ranking now, not what Claude believes is probably ranking based on training data. For a worked example of this skill in action, see our guide on wiring Claude to live data sources via MCP.

Before writing anything — a new blog post, a landing page, a product description — the SERP Analysis Skill tells you exactly what the top-ranking pages have in common. Feed it a keyword and a set of competitor URLs; it returns the entities they all mention, the heading structures they share, the average content depth, the questions they answer, and the schema types Google is surfacing in rich results.

Build the skill by connecting Claude to SerpAPI via MCP and writing a Skill.md that defines the analysis steps and the brief template format your writers use. A well-tuned SERP Analysis Skill eliminates the research phase of content creation almost entirely — your writers receive a brief that is already benchmarked against what ranks, not against what your editorial team thinks should rank.

The GEO / AI-Readiness Skill is the newest category in this guide and the fastest-growing use case for Claude SEO skills in 2026. Google's own GEO guidance confirms that AI-search optimisation is still core SEO — the same index, the same quality signals — but with an additional layer: your content needs to be structured so that AI systems can extract and cite it accurately.

Build the skill by defining a scoring rubric in your Skill.md: Does the page have a clear, directly answerable lead paragraph? Does it define its primary entity explicitly? Does it include FAQ schema? Does it cite primary sources? Does it cover the sub-topics an LLM would expect given the main keyword? Claude scores each dimension and returns a prioritised list of changes.

This skill pairs naturally with the Content Optimisation Skill — run SERP Analysis first, then Content Optimisation, then GEO Readiness as a final pass. That three-skill workflow covers everything a page needs to rank in both traditional and AI-mediated search. For teams that want this workflow running automatically across their entire site, Ryze AI's GEO layer handles it without any Skill.md authoring required.

The Draft-First Publishing Skill is the most technically involved Claude SEO skill in this guide, and the one with the largest time savings at scale. It connects Claude to your CMS — WordPress, Webflow, Contentful — via the Playwright MCP and authenticated API credentials, then creates full draft posts: title tag, meta description, heading structure, body copy, internal links, image alt text, and schema markup, all generated from a brief and queued for human review before publication.

The critical design decision is the review gate. Publishing automation creates risk when Claude receives broad write-access to your site; starting with draft-only mode means every page gets a final editorial check before going live. Store all credentials in environment variables — never in the Skill.md file itself — and use restricted API permissions that allow draft creation but not direct publishing.

Once the skill runs reliably in draft mode, you can incrementally expand its permissions — for example, auto-publishing pages that score above a defined quality threshold on an internal rubric. That is the pattern described by practitioners who have used this skill to publish dozens of high-quality SEO pages per week with a one-person content team. For a deep dive on the MCP connection architecture that powers this workflow, see our MCP integration guide.