SEO/GEO Agent Skills

Overview

A new category of AI agent skills has emerged around SEO and GEO (Generative Engine Optimization) — tools that give coding agents the ability to audit and optimize websites for both traditional search engines and AI-powered search. This comparison covers 11 open-source tools, from focused GEO audit skills and deterministic rule engines to comprehensive marketing libraries.

The market has validated fast. Between March and June 2026, stars exploded across the board — the category leader sits at 32,921, four tools cleared 7,000, and the adjacent commercial AI-visibility tier produced a unicorn: Profound raised a $96M Series C at a $1B valuation. AI-referred traffic grew 527% year-over-year and Gartner projects a 50% drop in traditional search traffic by 2028. GEO is no longer an early-mover bet; it is becoming a budget line.

---

Market Map

---

Category Segments

GEO-First Tools

geo-seo-claude (8,075 stars) is the original GEO-first entry — it explicitly prioritizes GEO over traditional SEO. Its citability scoring engine analyzes content blocks against research showing optimal AI-cited passages are 134-167 words. Five parallel subagents cover AI visibility, platform analysis, technical SEO, content quality, and schema markup, synthesizing into a composite GEO Score (0-100). The command surface grew from 12 to 15 between March and June — and the three additions (/geo prospect, /geo proposal, /geo compare) all tilt toward agency workflows rather than audit depth. Monetizes through a paid Skool community teaching GEO agency sales ($2K-$12K/month engagements).

GEO Optimizer Skill (463 stars, new to this comparison) is the engineer's GEO entry from Auriti Labs. Despite "skill" in the name, it is pip-installable Python software — a CLI, library, MCP server, and Astro integration backed by 1,720 tests — that scores sites against 47 research-backed methods anchored to the Princeton GEO paper (KDD 2024) and AutoGEO (ICLR 2026). Its differentiator: geo citations queries live answer engines to report whether ChatGPT, Perplexity, and Gemini actually cite your domain — closing the outcome loop every other audit tool leaves open. Open core: the engine is MIT, the GeoReady SaaS adds monitoring from $19/month.

Comprehensive SEO Suites (with GEO)

claude-seo (8,686 stars) is the fastest riser in the category — up from 2,656 in March. The v2.0.0 "Security + AI-Search Major Release" (May 25, 2026) nearly doubled coverage to 25 sub-skills, repositioned the project as AI-search first with passage citability scoring, and added Google APIs (Search Console, CrUX, GA4), PDF/Excel reporting, and an evidence-based posture on llms.txt. Eighteen specialist subagents run up to 15 in parallel, backed by 271 tests. The extensions system (DataForSEO, Firecrawl, Banana) remains the differentiator — real search data instead of heuristics.

Agentic-SEO-Skill (643 stars) is the most architecturally ambitious — 16 sub-skills, 10 specialist agents, and 88 Python scripts as evidence collectors, up from 33 at launch. Unique "reasoning-first" approach with confidence labels (Confirmed, Likely, Hypothesis) on each finding. Platform support tripled from 3 to 9 agents — Claude Code, Codex, Antigravity IDE, Claude Cowork, Cursor, Windsurf, Continue.dev, GitHub Copilot, and Cline. GitHub-specific SEO optimization remains a standout niche feature.

SEO/GEO Optimizer (33 stars) from 199 Biotechnologies covers the broadest optimization surface — SEO, GEO, AEO (Answer Engine Optimization), voice assistants, and social media. Python stdlib only, zero external dependencies. The oldest project in the category (November 2025) has answered the maintenance question: an active solo maintainer shipped IndexNow submission and position-bias checks through late May 2026, plus an unusually honest documentation pass that stripped unreproduced 2024-era GEO statistics.

Deterministic Rule Engines

SEO Audit Skill (280 stars, new to this comparison) from SEOmator takes the opposite bet from the prompt-driven majority: 251 deterministic rules across 20 categories, with Core Web Vitals measured through a real browser and raw-vs-rendered DOM comparison for SPA sites. It ships as a TypeScript CLI (~2,800 npm downloads/month), an Electron desktop app, and a Claude Code skill scoped to a single Bash(seomator:*) permission — the agent interprets the report, it does not perform the audit. Vendor-maintained as a funnel to SEOmator's $49-$279/month SaaS; development has slowed to maintenance since the v3.0 sprint, but a May 2026 patch hardened the LLM reporter against prompt injection.

Content-First Tools

SEO Machine (7,121 stars) remains the only tool focused on creating long-form, SEO-optimized content rather than auditing existing sites. It has grown into a full research-to-publish pipeline with GA4, Search Console, DataForSEO, and WordPress integrations. The caution: no commits since April 10, 2026 — adoption more than doubled while development paused, and the GEO gap remains unaddressed.

Broad Marketing Libraries

Marketing Skills (32,921 stars) is the category leader by roughly 6x, more than doubling since March, covering CRO, copywriting, SEO, analytics, and growth engineering across 50+ skills in 8 categories. No longer Claude Code only — v2.x added OpenAI Codex, Cursor, and Windsurf support. SEO/GEO is one capability among many; it remains the broadest but shallowest option on GEO specifically, with a commercial layer (the Magister "AI CMO") forming around the free core.

Multi-Agent Skill Libraries

seo-geo-claude-skills (2,092 stars, up more than 4x from March) still has the broadest agent support — 35+ agents via the skills.sh registry, including Cursor, Codex, Gemini CLI, and Windsurf. Twenty skills now ship with 5 consolidated commands (down from 9), and v9.9.10 (June 5, 2026) added MCP integration — closing its clearest earlier feature gap — plus a HOT/WARM/COLD memory system. Custom frameworks: CORE-EEAT Content Benchmark and CITE Domain Rating.

Full-Stack Audit + Execute

ultimate-seo-geo (48 stars) from Myk Pono is built around the AGENTS.md cross-platform standard and keeps shipping fast — v1.9.0 landed May 28, 2026. Three-mode workflow: Audit produces a scored SEO Health Score (0–100), Plan converts findings into a phased roadmap, and Execute generates deployable artifacts (JSON-LD, meta rewrites, redirect maps, robots.txt). Diagnostic scripts grew from 25 to 35, the eval suite expanded to 15 scenarios and 63 assertions (100% pass rate vs 87% baseline), and reports now export to HTML, Excel, and PDF. Supports Claude Code, Codex, Gemini CLI, Cursor, Copilot, Windsurf, Cline, Aider, Devin, Goose, Amp, and ChatGPT Custom GPTs. Still tiny on stars, still technically dense.

Consultant-Style Plugins

seo-geo-consultant (3 stars) is now dormant — the March 10, 2026 initial commit remains the only activity, with no releases, issues, or pull requests since. The "senior SEO & GEO consultant" persona with four workflow modes and Next.js-specific guidance survives as an interesting UX design reference, not a working tool.

---

Feature Comparison

---

Trends

Stars exploded — and the money followed. Every actively maintained tool multiplied its stars between March and June 2026: Marketing Skills 14,937 → 32,921, claude-seo 2,656 → 8,686, geo-seo-claude 3,320 → 8,075, SEO Machine 2,816 → 7,121, seo-geo-claude-skills 452 → 2,092. The clearest signal that this is a real market came from the adjacent commercial tier: Profound, the AI-visibility monitoring platform, raised a $96M Series C at a $1B valuation. The open-source skills in this comparison are the bottom-up wedge into the same budget that unicorn is chasing top-down.

SaaS vendors are skill-ifying. Established SEO SaaS companies now ship official agent skills — SE Ranking and SurgeGraph both released theirs — and two vendor-backed entries join this comparison: SEOmator's seo-audit-skill and Auriti Labs' geo-optimizer-skill with its GeoReady platform. The pattern is consistent: MIT-license the engine, monetize the monitoring and reporting layer. Expect every SEO SaaS to have an official skill within a year.

llms.txt reality check. Most tools in this category generate or validate llms.txt, but no major AI crawler consumes it in production. claude-seo's v2 moved to an explicitly evidence-based posture on it, and 199 Biotechnologies stripped unreproduced GEO statistics from its docs — early signs the category is self-correcting from hype toward verifiable claims. GEO Optimizer's live citation checking points at the better question: not "is the site theoretically optimized" but "do answer engines actually cite it."

The determinism split. A new architectural axis emerged: prompt-driven analysis (claude-seo, geo-seo-claude, Agentic-SEO-Skill) versus deterministic engines the agent merely invokes (seo-audit-skill's 251 rules, geo-optimizer-skill's tested CLI). Deterministic tools are reproducible and CI-friendly; prompt tools are broader and more strategic. The hybrid — scripts for evidence, LLM for reasoning — is where Agentic-SEO-Skill and ultimate-seo-geo sit.

Consolidation has started. Of 11 tools, one is dead (seo-geo-consultant — dormant since its single March 2026 commit) and one is paused (SEO Machine — no commits since April 10 despite doubling stars). The active leaders are pulling away: claude-seo and geo-seo-claude are in a genuine dead heat at ~8K stars each, Marketing Skills owns the generalist lane, and seo-geo-claude-skills owns multi-agent distribution. Watch ultimate-seo-geo and geo-optimizer-skill as engineering-dense dark horses with single-maintainer risk.

---

Bottom Line

If you want comprehensive SEO that is now AI-search first, claude-seo (8,686 stars) has the broadest feature set — 25 sub-skills, citability scoring, and real data via DataForSEO and Google APIs. If you want GEO-first focus, geo-seo-claude (8,075) remains the deepest on citability and AI crawler analysis, now with agency tooling built in. If you need skills beyond Claude Code, seo-geo-claude-skills (35+ agents) is still the distribution leader, with ultimate-seo-geo covering 10+ tools via AGENTS.md. And if you want deterministic, CI-friendly audits instead of prompt-driven analysis, the new vendor-backed entries — seo-audit-skill (251 rules) and geo-optimizer-skill (47 methods plus live citation checks) — are the engineering-grade picks. Skip seo-geo-consultant (dormant) and pair the paused SEO Machine with a GEO-aware tool if content production is your bottleneck.