- Claude SEO tools come in three species: data-connected suites, self-contained skill libraries, and single-purpose specialists. The right pick depends on whether you need live data, a process library, or one job done well.
- There is no overall number one on this page. Eight jobs get eight verdicts, and competitors win most of them.
- Half the field has had zero commits since launch day. In a niche where AI search behaviour shifts monthly, the last-push date is a review criterion, and we list it for every tool.
Every tool below was reviewed from its actual repository: README, code tree, committed examples, licence file and last push date, on 11 July 2026. Where a rival is better than our own library at a job, this page says so, because that is the only way a best-of written by a participant deserves to rank.
I build one of the tools on this page: the Claude SEO Skills library. It is reviewed in the same format as everything else, its limits included, and it wins exactly one verdict below. Star counts and commit dates are verifiable on GitHub.
The verdicts, by job
Nobody should crown one tool across eight different jobs. Pick your row:
The three species of Claude SEO tool
Data-connected suites pair skills with live data through MCP servers: real rankings, real backlinks, real AI share-of-voice. They are better at every data job, and they cost money because the data does. Self-contained libraries encode process: install free, bring your own exports. Specialists do one job properly. Comparing across species without saying which is which is how dishonest best-of pages get written.
Every tool, reviewed
Format is identical for all ten, ours included: what it is, what is genuinely good, honest limits verifiable in the repo, and who it is best for.
AgriciDaniel claude-seo
Data-connected suite ·
A Claude Code plugin with 25 sub-skills and 18 specialist agents that run in parallel, up to 15 agents on one site audit. 10,980 stars, CI with 326 passing tests, and a falsifiability methodology: every recommendation states how you would know it failed.
Genuinely good: The biggest and most rigorous ecosystem in the space, actively maintained (pushed this month). SE Ranking's official repo forks its Google scripts, which tells you where the centre of gravity sits. Optional DataForSEO and Firecrawl extensions add live data.
Honest limits: Two-tier model: the public MIT repo plus a private early-access mirror behind a paid community. The full agent fleet is a heavyweight run for small questions.
Best for: Most people searching "claude seo tool". Start here if you want the ecosystem default. (MIT; Plugin install; extensions need API keys.)
SE Ranking official skills
Data-connected suite ·
SE Ranking's official catalogue: 26 skills calling their remote MCP over OAuth. Measures real things: brand citation share across ChatGPT, Perplexity, Gemini and AI Overview, actual backlink gaps, actual keyword and traffic figures per URL.
Genuinely good: The most professionally built repo we reviewed. 25 committed example runs against real public sites, locked output formats written to dated folders, graceful degradation, and a README that honestly documents rate limits and credit burn.
Honest limits: Requires a paid SE Ranking account with API access, and credits on big domains are real money. Data quality is SE Ranking's index quality. A data-free library physically cannot do these jobs, and this repo is better at them, full stop.
Best for: Agencies and in-house teams already on SE Ranking, or anyone whose deliverables need live numbers. (MIT; Two plugin commands + OAuth; paid account.)
GTM Engineer Skills
Self-contained library ·
Twelve chained skills forming a research-to-published-page pipeline for AI answer engines: brand research, GEO prompt research, Reddit mining, content planning, AI-citable charts, an AEO audit, and code-level fixes.
Genuinely good: Engineering rigour: a bundled zero-dependency crawler runs 16 deterministic pass/fail checks, the audit honestly splits script-checked facts from model-judged scores, and an eval suite guards regressions. Guidance footnotes published research including the KDD 2024 GEO paper. 1,254 stars earned.
Honest limits: Manual symlink install, no update path. Classic SEO is out of scope: no rank tracking, no backlink index, no GSC, and keyword volumes are estimates.
Best for: Startup and SaaS growth engineers who live in a terminal and want the whole AEO pipeline. (MIT; git clone + symlinks; no accounts.)
Claude SEO Skills (this site)
Self-contained library ·
Our library: 43 free skills encoding the processes a 120-person agency runs across 500+ client campaigns. Audits, GSC analysis, content strategy, AI search, outreach and calculators, each with intake gates, checks and a locked output format.
Genuinely good: Real sample output published for every flagship skill, including runs on our own properties that scored imperfectly. Install is one command, MIT, no accounts, no email gate. The agent-skills discovery index ships sha256 digests so installs are verifiable.
Honest limits: No live data connection by design: rank, volume and backlink jobs need your GSC and Bing exports, or a vendor MCP alongside. Most skills do not have a dedicated page yet, and this site is weeks old (our own Website Score run gave it 16/25 and we published that).
Best for: SEOs and founders who want agency process, free, and are happy to bring their own exports. (MIT; One command, no accounts.)
GEO Analyzer
Specialist ·
Not a skill: a local MCP server (npm) with two tools that score content for AI citability. Claim density per 100 words, answer frontloading, semantic triples, sentence length against Google's observed extraction chunks, as repeatable 0 to 10 scores.
Genuinely good: The only tool here giving repeatable numeric GEO measurements rather than a checklist, with methodology cited to the KDD 2024 GEO paper and Dejan's 7,060-query grounding study. Analysis stays on your machine.
Honest limits: Single-purpose: content-level GEO only, one URL at a time, no technical SEO or rank data. Needs your own Anthropic API key, roughly 14 US cents per analysis.
Best for: Content teams who want a before and after number when rewriting for AI citability. (MIT; One claude mcp add + API key.)
Pre-Launch Website Audit
Specialist ·
One orchestrator skill with five sub-audits (technical SEO, AI accessibility, security hygiene, performance, on-page) that fingerprints your stack first, then tailors every check, with profiles for 15 stacks including vibe-coded Lovable and Bolt apps.
Genuinely good: Runs on bash alone with fallbacks for every check, deepens if Screaming Frog MCP or DataForSEO are present, and cross-references the five audits instead of running silos. Catching Cloudflare Bot Fight Mode as an invisible AI-crawler killer is a genuinely sharp check.
Honest limits: Deliberately single-purpose (its own frontmatter says when not to trigger it). Deepest tier needs a paid Screaming Frog licence. Zero commits since release day.
Best for: Go-live checks on staging sites, especially AI-generated builds. (MIT; One plugin command, no accounts.)
SEO/GEO/AEO Skill (Alex Labat)
Specialist ·
The single-skill archetype and the lowest barrier in the list: one skill file uploaded as a ZIP in the Claude desktop app. Audits a URL across SEO, GEO and AEO and produces a scored report with downloadable DOCX and PDF.
Genuinely good: Genuinely disciplined for a single file: forces a Quick versus Full audit choice, instructs the model never to flag something missing without confirming it, crawls beyond the homepage, and locks the deliverable format. The landing page explains it to non-technical users better than any README here.
Honest limits: It sees the web through WebFetch, which strips head metadata, canonicals and JSON-LD, the exact artefacts an SEO audit inspects, so schema and meta findings are inferences from a lossy view. No live data. Untouched since March, and no LICENSE file despite the open-source claim.
Best for: Business owners on claude.ai who want a readable first-pass audit with zero setup. (None committed; Download ZIP, upload in Claude. Done.)
lionkiii claude-seo-skills
Data-connected suite ·
52 commands, the broadest catalogue here: server log analysis, migrations, hreflang, llms.txt generation, a robots.txt audit covering 17 named AI crawlers, plus 10 Ahrefs and 9 GSC commands for live data.
Genuinely good: Real engineering care: every command self-checks its MCP dependencies and degrades gracefully, a 142-check smoke test validates the install, and the README documents genuine Claude Code gotchas with verification scripts.
Honest limits: The Ahrefs MCP sits behind a paid Ahrefs plan, so a fifth of the catalogue is dead weight without it, and the GSC path means self-hosting the author's separate MCP server. 18 stars means thin community vetting.
Best for: Technical SEOs already paying for Ahrefs who want its data inside Claude Code. (MIT; Two plugin commands; paid Ahrefs for full use.)
SEO AEO Pro (mangollc)
Self-contained library ·
An agency plugin: 14 slash commands backed by 6 skills, with a per-client workspace model. /new-client scaffolds a folder, everything writes into it, and /report assembles a monthly client report.
Genuinely good: The client-workspace model is the differentiator, and /enhance-aeo reverse-engineering AI queries from PAA chains, Reddit and Quora then testing which of a client's keywords earn AI Overview citations is genuinely clever. Honest about its own estimates.
Honest limits: Effectively a one-day release untouched for months: no commits since February, no LICENSE file committed, no example outputs anywhere in the tree, and everything routes through a Firecrawl dependency.
Best for: Small agencies wanting client folders and GBP content, who accept the staleness risk. (Claimed MIT, no file; Plugin + Firecrawl CLI login.)
25-in-1 (01clauding / OpenClaw)
Self-contained library ·
A 25-skill markdown suite: keyword research, an 80-rule audit rubric, GEO, reporting. Documentation in four languages. It is a mirror of the OpenClaw project's suite, sharing identical push timestamps with the original.
Genuinely good: The reference layer is decent teaching material: rule IDs across 7 audit categories, correct citation of the KDD 2024 GEO paper, CWV thresholds. As a free pile of SEO prompting patterns to read and cannibalise, it delivers.
Honest limits: The data-shaped skills have no data source: rank tracker, backlinks and monitoring ship with no backing API. Its claim to be the only Claude suite with GEO is contradicted by four other tools on this page. Hardcoded statistics are date-stamped February 2026 and aging silently. Abandoned within two days of launch.
Best for: Reading material, honestly. Install one of the maintained libraries instead. (MIT; git clone + copy.)
How to choose in 60 seconds
Still stuck: start free. Install a process library today, run your first audit this hour, and add a data connection the day a deliverable needs live numbers. The audit walkthrough shows the whole first hour with real output.
FAQ
What is a Claude SEO tool?
Three species share the name: skill libraries you install into Claude (markdown files encoding workflows), plugins and MCP servers that connect Claude to live data sources like Ahrefs or Search Console, and single-purpose specialists that do one job such as scoring content for AI citability. This page reviews all three and says which is which.
Which Claude SEO tools are actually free?
The self-contained libraries: our 43-skill library, GTM Engineer Skills, and the stale-but-readable 25-in-1 suite, all MIT with no accounts. The specialists are free to install but GEO Analyzer bills your own API key per analysis. Anything with live rank, volume or backlink data always needs a paid data source behind it.
Do these replace Ahrefs or Semrush?
No, and the honest tools say so. A skill has no crawl index or clickstream data, so rank tracking, search volumes and backlink profiles still come from a data provider, either through their MCP servers inside Claude or through exports you paste. What skills replace is the process layer: the audit checklist, the analysis, the report.
What is the difference between a skill and an MCP server?
A skill teaches Claude a process: steps, checks, an output format. An MCP server gives Claude a capability: live data or actions from an external system. They compose. SE Ranking's catalogue is skills on top of their MCP, and our library pairs with whatever data sources you already have.