Agentic SEO CLI

Scores how readable a web page is to the AI crawlers behind ChatGPT, Claude, Perplexity and Gemini. Same HTML in, same score out, offline, no API key.

python3 cli.py audit https://example.com --html out/report.html

The problem it solves

The crawlers that feed AI answer engines do not run JavaScript. They read the raw HTML your server sends, in source order, and stop after roughly the first 5,000 to 6,000 characters.

Your browser shows you the page after JavaScript has run. That is a different page. So a site can look complete to you, and to Google, while being thin or blank to ChatGPT.

There is no simple way to see the version the crawler sees. This tool fetches it, scores it out of 100, and shows you the difference between the two.

Install

Python 3.10 or newer, one dependency.

git clone https://github.com/hasandenizuk/agentic-seo-cli.git
cd agentic-seo-cli
python3 -m venv .venv
./.venv/bin/pip install -r requirements.txt

Rendered capture is optional and uses Playwright. Without it the audit runs on raw HTML only and says so in the report.

What you get

A score out of 100 across nine areas: whether real content is in the raw HTML, whether the page answers its question early, schema, semantic markup, entity clarity, navigation signals, form controls, freshness dates and citations. Plus the JavaScript gap, the content a person sees that a crawler cannot, reported first because it is usually the biggest finding.

Scores are page-type aware. A documentation page is not marked down for lacking a product schema.

Exit code 1 when a page falls below a threshold you set, so it can gate a deployment.

Documentation

• Usage covers every command and flag: auditing a URL or a folder of saved HTML, bulk auditing a Screaming Frog crawl, and scaffolding the agent-discovery files.
• Scoring explains the nine areas, how a score is built, what counts as a blocker, and the four bands from Agent-Ready down to Rebuild.
• Use cases has worked examples: a pre-deploy gate, a CI job, a multilingual site, and auditing a full crawl.
• Research is the evidence the tool is built on, including the measurement of 500 million crawler requests that found no JavaScript execution.
• Comparison sets it against AEO checkers, LLM-based SEO tools and classic crawlers, and says plainly where it is the wrong choice.
• Extending shows how to add a check or a page type.

Use it from an AI coding assistant

The repo ships a skill, so an assistant can run the audit and then finish the judgement-only checks without touching the deterministic score.

python3 skills/agentic-seo-audit/scripts/install.py claude   # or codex, or cursor

Then ask it to audit agent readiness on a URL. See the skill.

Author

Built by Hasan Deniz. Part of the micro-tools collection - small, focused utilities for web development and SEO.

• Website: hasandeniz.com
• GitHub: github.com/hasandenizuk
• LinkedIn: linkedin.com/in/hasandeniz

License

Apache 2.0 License. See LICENSE.