GEO-SEO Claude

What it is

GEO-SEO Claude is Zubair Trabzada’s toolkit for Claude Code. You type /geo audit and a site address. Claude then checks whether ChatGPT, Perplexity, Gemini and Google’s AI answers can find, read and quote that site. A full audit sends five subagents out side by side. It ends in one score from 0 to 100 and a ranked list of fixes. The ordinary SEO basics are checked too.

1. You run the installer once
2. You type /geo audit with your site address
3. Claude fetches pages and runs five subagents
4. Claude scores the site and writes GEO-AUDIT-REPORT.md
5. You apply the fixes and upload any files yourself

Who it suits

Site owners and marketers who want an AI search checkup, and freelancers or small agencies who want a report to hand a client. Extra commands track prospects, write proposals and compare month with month. Its audits read public pages, write files into the folder where you run Claude Code and publish nothing. The agency commands keep their data in ~/.geo-prospects.

Where to use it

Claude Code is the only home, on macOS, Linux and Windows. Windows needs Git Bash.

• Claude Code on macOS or Linux: install.sh
• Claude Code on Windows: install-win.sh, run from Git Bash and not from PowerShell
• Any public website. The skills tell Claude to check robots.txt first
• Not for Cursor, Codex or a chat window. The README documents none of them

How to set it up

Get it saves a zip of Mindket’s setup notes in your downloads folder, with a page for each system. The toolkit itself installs from the maker’s repository.

1. Check your tools first. You need Python 3.10 or newer, Git and Claude Code. On Windows, also Git Bash.
2. Clone it and run the installer. It copies the skills into ~/.claude and installs the Python packages.
3. Restart Claude Code and test it. The maker’s docs say Claude Code reads skills at startup, so a session that was already open will not see them.

git clone https://github.com/zubair-trabzada/geo-seo-claude.git
cd geo-seo-claude
./install.sh

On Windows, run these lines in Git Bash and use the other installer:

git clone https://github.com/zubair-trabzada/geo-seo-claude.git
cd geo-seo-claude
./install-win.sh

For macOS and Linux, the README also gives a one-command route. It runs the script from GitHub without you reading it first, so the clone route above is the safer first choice.

curl -fsSL https://raw.githubusercontent.com/zubair-trabzada/geo-seo-claude/main/install.sh | bash

The installer asks questions only in a terminal. There it asks whether to go on without Claude Code and whether to add Playwright. Run through curl, it goes on without asking and skips Playwright.

How to use it

Open Claude Code in any folder and type /geo with a subcommand and a site address. Claude fetches the pages, runs the checks and, for most commands, writes a file into that folder. You read it and decide what to change.

The skills cap an audit at 50 pages, wait one second between requests and give each page 30 seconds. Claude does the fetching and scoring. You apply the fixes and upload any file it writes.

Example

You run a small software site and want a first checkup before you change anything.

/geo audit https://yoursite.com

After several minutes Claude writes GEO-AUDIT-REPORT.md into your folder, with a GEO score from 0 to 100, the issues by severity, quick wins and a 30-day action plan.

The score is a weighted average of six areas. AI citability counts for 25% and brand authority for 20%.

For newcomers and experts

If you are new. GEO means optimizing for the answers AI tools write, not for a list of links. Start with /geo quick, then /geo crawlers. The crawler check reads your robots.txt, then the robots tags and headers of a few pages. The score is the toolkit’s own weighted estimate. It is not a count of how often ChatGPT names you.

For experts. The six areas and their weights, from geo/SKILL.md:

The scripts print JSON, and you can run them yourself from a clone once its packages are in. These lines are CLAUDE.md’s, the first one setting up the packages:

python3 -m venv .venv && . .venv/bin/activate && pip install -r requirements.txt
python3 scripts/fetch_page.py <url>
python3 scripts/citability_scorer.py <url>
python3 scripts/brand_scanner.py "<brand name>"
python3 scripts/llmstxt_generator.py <url>

Good to know

The toolkit is free and MIT licensed. Its files do not always agree with each other.

• Windows: install-win.sh installs the packages into your own Python with pip --user and makes no virtual environment. It also leaves the python3 commands in the skills as written, so a step that names python3 may fail where Git Bash has no python3
• The PDF report needs pandoc and Chrome at a macOS path. The README’s ReportLab script is not in the repository
• Counts differ: the README says 13 sub-skills, geo/SKILL.md says 14, and the repository holds 15
• Market figures in the README are the maker’s own editorial text. The toolkit does not measure them
• Updates: /geo update copies the latest files over the installed ones, so edits you made to an installed skill are lost. It does not redo the installer’s Python setup, so running the installer again from a fresh clone is the surer update
• Removing: ./uninstall.sh or the README’s rm line. The folder ~/.geo-prospects, with your prospect data, stays until you delete it
• Data: the brand check calls the Wikipedia and Wikidata APIs and has Claude look the brand up on other sites. Page text and your prompts go to Claude’s provider as in any Claude work
• Maker: Zubair Trabzada. The setup notes in the zip are Mindket’s