Key takeaways
- Three-mode workflow (Audit → Plan → Execute) produces scored findings, prioritized roadmaps, and ready-to-deploy fixes — not just advice
- Broadest agent support in the category — works with Claude Code, Codex, Gemini CLI, Cursor, Copilot, Windsurf, Cline, Aider, Devin, and ChatGPT Custom GPTs via AGENTS.md standard
- 35 bundled Python scripts cover the full diagnostic surface: Core Web Vitals, schema validation, hreflang, broken links, readability, duplicate content, AI crawler access, and more — plus HTML, Excel, and PDF report export
- Eval suite with 15 scenarios and 63 assertions shows 100% pass rate with the skill vs 87% baseline — adds ~50s and ~19K tokens per task
FAQ
What is ultimate-seo-geo?
A cross-platform SEO and GEO skill for AI coding agents. It runs full site audits with scored findings, generates implementation plans, and produces executable fixes for both traditional search and AI search engines.
Which AI tools does ultimate-seo-geo support?
It supports 10+ tools via the AGENTS.md standard: Claude Code (plugin marketplace), OpenAI Codex, Gemini CLI, Cursor, GitHub Copilot, Windsurf, Cline, Aider, Devin, and ChatGPT Custom GPTs with a dedicated setup bundle.
How does it compare to geo-seo-claude?
geo-seo-claude (8,075 stars as of June 2026) has deeper GEO-specific scoring but is Claude Code only. ultimate-seo-geo has broader agent support and more diagnostic scripts (35 vs implicit), but is newer with far fewer stars (48).
What makes ultimate-seo-geo different?
The three-mode Audit → Plan → Execute workflow and 35 bundled diagnostic scripts that actually run against live URLs. Most competitors provide guidance; this one produces concrete, deployable artifacts.
Overview
ultimate-seo-geo is an SEO and GEO skill built around the AGENTS.md cross-platform standard, making it one of the few tools in this category that works beyond Claude Code. Its three-mode workflow — Audit → Plan → Execute — is designed to produce deployable artifacts rather than recommendations.
Key stats (as of June 11, 2026): 48 stars, 10 forks, MIT license, Python. Created March 23, 2026. Actively maintained — v1.9.0 shipped May 28, 2026 (last push).
| Attribute | Value |
|---|---|
| Stars | 48 |
| Forks | 10 |
| License | MIT |
| Language | Python |
| Created | March 23, 2026 |
| Latest version | v1.9.0 (May 28, 2026) |
| Diagnostic Scripts | 35 |
| Agent Support | 10+ (AGENTS.md standard) |
| Dependencies | Python 3.8+, requests, beautifulsoup4 |
| Category | SEO/GEO Agent Skills |
Architecture
Two-layer progressive disclosure:
- Layer 1 — AGENTS.md (~27KB): Auto-loaded by all AGENTS.md-compatible tools. Contains routing logic, condensed procedures, script references, and quality gates. Stays under Codex's 32KB limit.
- Layer 2 — SKILL.md (routing shell) + references/procedures/ + scripts/: Loaded on demand when deeper detail or execution is needed. Since the v1.8.4 restructure, the monolithic SKILL.md has been split into a small routing shell plus per-section procedure files that hosts pull only when needed.
This is a practical solution to the context window problem — front-load the essentials, lazy-load the rest.
Three-Mode Workflow
Mode 1 — Audit: Fetches the site, runs all diagnostic scripts, and produces an SEO Health Score (0–100) with prioritized findings. Each finding includes severity, evidence, and a fix suggestion.
Mode 2 — Plan: Converts audit findings into a phased implementation roadmap with effort/impact scoring and owner assignment. Useful for handing off to developers or clients.
Mode 3 — Execute: Produces the actual fixes — JSON-LD schema, meta tag rewrites, redirect maps, robots.txt configurations — and verifies them. This is where most competing tools stop at "you should fix this" while ultimate-seo-geo generates the code.
Most requests run all three modes in sequence.
Diagnostic Scripts (35)
The bundled Python scripts are the real differentiator — they run against live URLs and return structured data. The count has grown from 25 at launch to 35 as of v1.9.0:
| Script | Purpose |
|---|---|
| generate_report.py | Full-site HTML dashboard (orchestrates all scripts) |
| validate_schema.py | JSON-LD validation (pure stdlib) |
| robots_checker.py | robots.txt rules + AI crawler allow/block status |
| pagespeed.py | Core Web Vitals via PageSpeed Insights API |
| hreflang_checker.py | All 8 hreflang rules + bidirectional return tags |
| internal_links.py | Link graph, orphan pages, anchor text, crawl depth |
| broken_links.py | 4xx/5xx broken links + redirect counts |
| security_headers.py | HSTS, CSP, X-Frame-Options — weighted score |
| entity_checker.py | Wikidata, Wikipedia, sameAs entity signals |
| llms_txt_checker.py | llms.txt presence + format validation |
| readability.py | Flesch-Kincaid grade + sentence stats |
| duplicate_content.py | Near-duplicate detection |
| sitemap_checker.py | Sitemap discovery + sanity checks |
| local_signals_checker.py | LocalBusiness / tel / address signals |
Plus 21 more covering redirects, social meta, image alt text, article structure, link profiles, IndexNow, dependency preflight (requirements-check.py), and eval transcript scoring (score_eval_transcript.py). By default audits use the seed URL with modest crawl limits; a --crawl-deep flag (added in v1.8.5) widens link and canonical coverage, and reports now export to HTML, Excel, and PDF. Optional MCP extensions (Firecrawl, DataForSEO) are documented for JS-rendered sites and large audits.
Coverage
SEO
Technical SEO (Core Web Vitals, crawlability, JS rendering, security headers), on-page optimization, content and E-E-A-T assessment, schema markup (deprecation-aware — HowTo and SpecialAnnouncement removed), keywords and topic clusters, link building, local SEO, international SEO (hreflang), programmatic SEO, site migrations, and analytics setup.
GEO
Platform coverage for Google AI Overviews, AI Mode, ChatGPT Search, Perplexity, and Bing Copilot. Citability scoring with 134–167 word answer block optimization. Brand mention strategy across YouTube, Reddit, Wikipedia, and LinkedIn. AI crawler management for GPTBot, OAI-SearchBot, PerplexityBot, and ClaudeBot. llms.txt generation and RSL 1.0 machine-readable AI licensing.
Cross-Platform Support
| Platform | How It Loads | Script Execution |
|---|---|---|
| Claude Code | Plugin marketplace + SKILL.md | Full |
| OpenAI Codex | AGENTS.md auto-loaded (32KB limit) | Full |
| Gemini CLI | GEMINI.md → imports AGENTS.md | Full |
| Cursor IDE | AGENTS.md + .cursor/rules/ | Full |
| GitHub Copilot | AGENTS.md + copilot-instructions.md | Full (agent mode) |
| Windsurf / Cline / Aider / Devin / Goose / Amp | AGENTS.md auto-loaded | Full |
| ChatGPT Custom GPT | Uploaded instructions + knowledge files | Limited (no scripts) |
| Claude Desktop | Upload SKILL.md to Project Knowledge | No shell |
Eval Results
Benchmarked with 15 scenarios and 63 assertions (expanded from 14/60 at launch):
| Metric | With Skill | Without Skill | Delta |
|---|---|---|---|
| Pass rate | 100% | 87% | +13 pts |
| Avg time | 103s | 54s | +49s |
| Avg tokens | 83K | 64K | +19K |
Test scenarios include YMYL publisher audits, local business + schema, SaaS schema, migration plans, recipe content (no URL), news/paywall, international hreflang, traffic drop routing, GEO platform routing, an execute-mode risk gate (robots.txt), and fabrication checks.
Who Made This
Myk Pono — creator and maintainer. The repo was created March 23, 2026 and has shipped steadily since: v1.8.x releases through early April added PDF export, an MCP extensions doc, and --crawl-deep, and v1.9.0 (May 28, 2026 — the latest push) hardened audit fetching and added rendering-drift checks.
Changes Since the March Profile
- Stars grew from 12 to 48 and forks to 10 — still a small project, but no longer brand-new.
- Diagnostic scripts expanded from 25 to 35; evals from 14 scenarios/60 assertions to 15/63 (pass-rate deltas unchanged).
- New since March:
--crawl-deepmode, Excel/PDF report export, optional Firecrawl/DataForSEO MCP extensions, a plugin agents bundle, and Goose/Amp added to supported tools. - Pricing unchanged: free and open source (MIT); optional PageSpeed and extension API keys are the only external costs.
What Developers Say
No attributed third-party developer testimonials for ultimate-seo-geo were findable in searches of Hacker News, Reddit, X, or developer blogs as of June 11, 2026 — results circle back to the repo itself and this profile. Treat the 48 stars and 10 forks as the most honest adoption signal available.
Bottom Line
ultimate-seo-geo trades star count for substance — 35 diagnostic scripts, a structured eval suite, and the broadest agent support in the category via the AGENTS.md standard. Recommended for developers and marketers on non-Claude agents (Codex, Cursor, Gemini CLI, Copilot) who want scripted audits plus deployable fixes — nothing else in the category covers that surface. Not recommended if you need community-validated tooling: at 48 stars with no third-party testimonials, you are betting on one maintainer's pace. Outlook: positive — a 4x star climb, monthly releases, and a v1.9.0 push in late May show real momentum, but it needs visible adoption beyond the author to close the gap with geo-seo-claude.
See the full SEO/GEO Agent Skills comparison for how it stacks up against geo-seo-claude, claude-seo, and 6 other tools in the category.