Skip to content
FAHD MOHAMED
All work

Professional work · Generative Engine Optimization · Crawler

GEO Audit Platform

Audits how ready a website is to be cited by AI answer engines, then writes the fixes.

GEO categories scored
10
GEO categories scored
source: src/lib/types.ts
page types classified
16
page types classified
source: src/lib/pipeline
export formats
4
export formats
source: reports
Audit result: GEO score, category scores, top improvements

The problem

Search is shifting to AI answers. Businesses need to know whether ChatGPT-style engines can understand, trust and cite their site.

My role

  • Engineered the platform: AI engineering, GEO analysis pipeline and platform development

Engineered during my role as AI & Software Engineer at Elenor Marketing Agency, Cairo (Jan 2026 – present). Screenshots come from sanitized demo environments: client and business data are replaced with demo data.

What it is

Paste a URL. A Playwright and Cheerio crawler maps the site and classifies pages into 16 types, then extracts business facts and checks name, address and phone consistency. It scores 10 GEO categories from 0 to 100 and generates prioritized recommendations and content, exported as PDF, DOCX, Markdown or JSON. The LLM layer treats crawled text as untrusted input, and falls back to templates without a key.

Key features

  • Staged job pipeline with retry and live progress
  • 10 scored GEO categories plus issue list
  • PDF and DOCX report generation
  • SSRF protection: internal hosts and private ranges are refused

Architecture

Input

  • URL + config
  • SSRF validator

Pipeline

  • Crawl
  • Classify
  • Extract facts
  • Score

Generate

  • Recommendations
  • Content
  • LLM (optional)

Output

  • PDF
  • DOCX
  • MD / JSON

AIPrompted with anti-hallucination rules; crawled content treated as untrusted

AIDeterministic template fallback

Stack

App

Next.js 15React 19TailwindRadix

Pipeline

PlaywrightCheerioPrismaPostgreSQL

Output

OpenRouter / OpenAI SDKpdfkitdocx

Engineering challenges

01

A crawler that can’t be turned on you

User-supplied URLs are validated against localhost and private IP ranges before any request. I confirmed this while capturing: a localhost audit is rejected.

Run locally. The audit shown is a real run against example.com (the IANA documentation domain), without an LLM key.

Start an audit
Full audit report

Next project

AI SEO Agent System