Comparison · Practitioner
Claude SEO Audit Skill: Tested
Claude SEO is a free, open-source plugin for Claude Code that runs SEO audits from the terminal. Its README lists 26 sub-skills and 19 specialist agents covering technical SEO, E-E-A-T, schema, AI search (GEO), local, ecommerce and international SEO. It's a useful second opinion. Its GEO scores estimate citation likelihood; they don't measure real citations, so verify them.
Disclosure: MentionX publishes this page and sells AI visibility tracking, which overlaps with part of this skill. Facts below come from the project's GitHub repository, read on 30 October 2026 (CET). Claude SEO is a third-party project, not an Anthropic product.
What does Claude SEO check?
From the README (GitHub):
| Area | What it covers | Command |
|---|---|---|
| Full audit | Parallel agents, prioritised action plan | /seo audit <url> |
| Single page | Deep analysis of one URL | /seo page <url> |
| Technical | Crawlability, indexing, sitemaps, robots.txt, redirects, Core Web Vitals | /seo technical <url> |
| Content | E-E-A-T, thin pages, topic coverage | /seo content <url> |
| Schema | Detect, validate, generate Schema.org markup | /seo schema <url> |
| AI search | AI Overviews and generative engine optimisation | /seo geo <url> |
| Agent readiness | Including llms.txt and agentic browsing checks | /seo agentic <url> |
| Change monitoring | Baselines and comparisons stored in SQLite | /seo drift |
| Local and ecommerce | Business profile, map pack, product pages | /seo local, /seo ecommerce |
The README says a full audit takes 10 to 15 minutes. That's the project's own claim; we haven't timed it.
How do you install it?
The project documents a clone-and-install approach (GitHub):
git clone --depth 1 https://github.com/AgriciDaniel/claude-seo.git
bash claude-seo/install.shThen, inside Claude Code, run /seo setup to create its runtime and /seo doctor to check readiness, then /seo audit https://your-site.com.
What should you know about security?
An audit skill fetches URLs, runs scripts and may hold API keys. The project's v2.2.0 release notes mention fixes for installer credential injection, SSRF handling and API-key exposure (release notes). That shows active maintenance, and it's also a reminder to:
- [ ] Read
install.shbefore running it - [ ] Install the latest release, not an old copy
- [ ] Use API keys with the smallest scope available
- [ ] Run it on a machine without production secrets
- [ ] Review commands Claude Code asks permission to run
How reliable are its GEO findings?
The GEO module estimates how likely AI engines are to cite a page, based on page features. That's useful for finding obvious gaps: missing direct answers, weak structure, blocked AI crawlers. But no on-page audit can tell you whether ChatGPT or Perplexity actually cites you today. That depends on retrieval, the prompt and competing sources.
| Finding type | How much to trust it | How to verify |
|---|---|---|
| robots.txt blocks an AI crawler | High; it's checkable | Read robots.txt yourself |
| Schema errors | High | Google Rich Results Test |
| Core Web Vitals | High if from field data | PageSpeed Insights / CrUX |
| "Citation likelihood" score | Low to medium; it's a heuristic | Run real prompts and log citations |
| Content quality / E-E-A-T | Medium; it's an opinion | Human editorial review |
How does it compare with other audit approaches?
| Approach | Strength | Limitation |
|---|---|---|
| Claude SEO skill | Broad coverage, free, explains each recommendation | Needs a terminal and Claude Code; findings need verification |
| Desktop crawler | Exhaustive crawl data, repeatable | Less interpretation; AI search checks are limited |
| SEO suite site audit | Scheduled, dashboards, historical trends | Paid; generic recommendations |
| AI visibility tracker | Measures real citations across engines | Doesn't audit on-page technical issues |
| Manual expert review | Judgement and context | Slow and costly |
The skill fits best between a crawler and a human review: it turns raw checks into a prioritised plan you then validate.
What changed in recent releases?
According to the v2.2.0 release notes (release notes), the release:
- Closed an install-time credential-injection vector in extension installers
- Rejected SSRF authority-confusion bypasses
- Moved Google and Bing API keys out of request URLs and error output
- Fixed GSC totals falsely reporting "0 clicks" for low-volume queries
- Models AI Mode as a distinct citation engine from AI Overviews
Check the releases page for anything newer before installing.
Our test results
This section will show MentionX's hands-on test: three sites, audit findings compared with manual checks, and GEO scores compared with real prompt results. Until then, this page stays unpublished.
| Site | Real issues found | False positives | GEO score vs real citations |
|---|---|---|---|
| Pending test | — | — | — |
When is it a good fit?
| You are... | Fit |
|---|---|
| A freelancer who wants a quick first-call audit | Good, then check the findings yourself |
| An in-house team that wants a second opinion | Good |
| A team that needs ongoing AI citation tracking | Not a replacement; pair it with prompt tracking |
| Someone not comfortable in a terminal | Poor fit |
How should you use the output?
- Run
/seo auditand export the action plan. - Verify each high-priority item with the source tool (robots.txt, the Rich Results Test, PageSpeed Insights).
- Treat GEO scores as hypotheses. Pick 10–20 buyer prompts and check real answers.
- Fix verified issues first, then re-run
/seo drift compareto confirm changes.
Checklist
- [ ] Latest release installed, installer read
- [ ] Ran on a safe machine with minimal keys
- [ ] High-priority findings verified
- [ ] GEO claims checked against real prompts
- [ ] Re-audit scheduled
Related
Sources
- AgriciDaniel, claude-seo repository and README, accessed 30 Oct 2026
- AgriciDaniel, claude-seo v2.2.0 release notes, accessed 30 Oct 2026
- Anthropic, Claude Code, accessed 30 Oct 2026
Questions
What is the Claude SEO skill?
An open-source, MIT-licensed plugin for Claude Code that runs SEO audits with sub-skills and specialist agents, covering technical SEO, content, schema, AI search, local and ecommerce.
Is Claude SEO free?
The project describes itself as free and open source under the MIT licence. You still need access to Claude Code, and some optional extensions use paid data APIs.
Can you trust its AI search (GEO) scores?
Treat them as heuristics. The scores estimate citation likelihood; they don't measure actual citations. Verify with real prompt checks in each AI engine.