DEV Community

TeHsuWang
TeHsuWang

Posted on

Is Your Website Invisible to AI? Stop Optimizing for Google, Start Optimizing for ChatGPT

#ai #seo #aeo #opensource

67% of users now get their first answer from an AI assistant rather than clicking a search result. Your site might rank #1 on Google and still be invisible to ChatGPT, Perplexity, and Google AI Overview.

The problem: SEO optimizes for ranking algorithms. AI engines don't rank -- they cite. If your content can't be extracted and quoted, it doesn't exist.

This is AEO (Answer Engine Optimization), and we built an open-source tool for it.

One Command 

npx aeoptimize scan your-site.com
Enter fullscreen mode Exit fullscreen mode 
AEO Readability Report
Score: 66/100  AI Readability: Good

  Structure        ██████████████░░░░░░ 18/25
  Citability       █████████████░░░░░░░ 16/25
  Schema           ███████░░░░░░░░░░░░░  7/20
  AI Metadata      █████████████████░░░ 13/15
  Content Density  ████████████████░░░░ 12/15

Warnings:
  ! Found 2 H1 headings. Use exactly one H1 per page.
  ! JSON-LD missing fields: @type, name, description

Top Suggestions:
  -> Add FAQ section with question-format headings
  -> Add AI-relevant schema types (FAQPage, Article, HowTo)
Enter fullscreen mode Exit fullscreen mode

No API keys. No signup. Runs offline.

What aeoptimize Measures

17 rules across 5 dimensions, scored 0-100:

Dimension Max What AI engines care about
Structure 25 Clear headings, short paragraphs (<150 words), FAQ sections
Citability 25 Self-contained statements an AI can quote without context
Schema 20 JSON-LD that LLMs use as source of truth (FAQPage, Article)
AI Metadata 15 llms.txt file, robots.txt AI crawler rules
Content Density 15 Signal-to-noise ratio, vocabulary diversity

The scoring is deterministic -- same input, same score. No LLM involved in the base scan, so you can run it in CI/CD without worrying about cost or rate limits.

Benchmarking Real Websites

Website Score Why
nextjs.org/docs 75 Good structure, has llms.txt, missing some schema
shiheintelligent.com 66 Has schema but incomplete, 2 H1s, no FAQ
stripe.com/docs 59 Great content but zero JSON-LD (-20 points)
anthropic.com 52 Landing page with little extractable content

Even well-built sites lose 20+ points from missing JSON-LD alone. That's the difference between being cited and being ignored.

Case Study: Fixing a 66-Score Site

shiheintelligent.com scored 66/100. The three biggest issues:

  1. Two H1 headings -- AI engines treat H1 as the page topic. Two H1s create ambiguity.
  2. Incomplete JSON-LD -- Schema existed but missing @type, name, description. AI crawlers can't categorize the content.
  3. No FAQ section -- The site answers common questions in paragraph form, but without FAQ schema, AI engines can't extract them as Q&A pairs.

Fixing these three issues alone could push the score above 80.

More Than a CLI

aeoptimize works in multiple modes:

Framework plugins -- auto-generate llms.txt and JSON-LD on build:

// vite.config.ts
import { aeoPlugin } from 'aeoptimize/vite';
export default defineConfig({ plugins: [aeoPlugin()] });

// next.config.mjs
import { withAeo } from 'aeoptimize/next';
export default withAeo({});
Enter fullscreen mode Exit fullscreen mode

Pre-commit hook -- block commits that drop AEO score:

npx aeoptimize hook install              # min score: 60
npx aeoptimize hook install --min-score 80
Enter fullscreen mode Exit fullscreen mode

GitHub Action -- check every PR:

- uses: dexuwang627-cloud/aeoptimize/action@main
  with:
    path: dist
    min-score: 60
Enter fullscreen mode Exit fullscreen mode

Claude Code skills -- interactive, AI-powered optimization:

claude plugin marketplace add dexuwang627-cloud/aeoptimize
# Then: /aeo-scan, /aeo-generate, /aeo-transform
Enter fullscreen mode Exit fullscreen mode

The /aeo-transform skill uses your Claude subscription to restructure content: split paragraphs, extract FAQ schemas, remove keyword stuffing, fix dangling references. Zero extra cost.

Multi-AI Scoring

For higher confidence, aeoptimize can dispatch scoring to multiple AI engines simultaneously:

npx aeoptimize scan your-site.com --multi-ai
Enter fullscreen mode Exit fullscreen mode

It detects gemini and copilot CLIs on your system, sends each the page content for independent evaluation, then merges with the rule engine using weighted consensus. The result shows per-AI insights:

Score: 72/100 (Rule Engine: 61 | AI Consensus: 83)

AI Insights:
  Claude:  "FAQ section lacks schema markup"
  Gemini:  "Missing llms.txt reduces discoverability"
Enter fullscreen mode Exit fullscreen mode

Built by Three AIs

The project itself was built as a multi-AI collaboration:

  • Claude -- architecture, core engine (17 rules, scanner, generator), security audit
  • Gemini -- Vite plugin
  • Copilot -- Next.js plugin

50 tests, 4 rounds of automated code review (code quality, security, architecture, final). The biggest competitor in this space has 4,800+ stars but zero tests.

The SEO vs AEO Shift

SEO AEO
Goal Rank higher Get cited
Audience Search crawler Language model
Key metric Click position Citation accuracy
Content style Keyword-rich Self-contained, structured
Structured data Nice to have Essential

This shift is already happening. Google AI Overview appears in 43% of queries. Perplexity processes 200M+ queries/month. Content that AI can't parse is content that doesn't exist.

Try It 

npx aeoptimize scan your-site.com
Enter fullscreen mode Exit fullscreen mode

GitHub: https://github.com/dexuwang627-cloud/aeoptimize
npm: https://www.npmjs.com/package/aeoptimize

Star the repo if it helped. Open issues if it didn't.

Top comments (1)

Collapse
 
ali_muwwakkil_a776a21aa9c profile image
Ali Muwwakkil •

Most developers overlook that optimizing for AI assistants involves training language models with domain-specific data, not just tweaking SEO. In our experience, integrating RAG architectures can make data retrieval more dynamic and context-aware, giving AI models a better grasp of your content. This approach ensures AI provides accurate info and keeps users engaged longer. - Ali Muwwakkil (ali-muwwakkil on LinkedIn)