AI Context Files: Platform-Ready Setup

AI Context Files — Platform Guide

GEO Optimizer provides platform-specific context files in the ai-context/ folder. Each file is optimized for its platform’s format and character limits.

TL;DR: See SKILL.md for the quick-reference table. This doc has step-by-step setup for each platform.

Claude Projects

File: ai-context/claude-project.md
Limit: No practical limit — full context loaded
What you get: Full workflow, 9 Princeton methods with impact %, all scripts with flags, framework examples (Astro, Next.js, WordPress), GEO checklist, robots.txt block, llms.txt templates, JSON-LD schema examples

Setup

Go to claude.ai → click Projects in the sidebar
Click + New Project (e.g. “GEO Optimizer”)
Click Add content → Upload file → select ai-context/claude-project.md
Start a new conversation inside the Project

Claude will have persistent GEO context across every conversation in that Project.

What to expect

Immediate, specific advice without needing to re-explain what GEO is
Ready-to-paste robots.txt blocks, llms.txt, and schema JSON
The AI will start every site analysis with the audit command
Framework-specific schema code (Astro, Next.js, WordPress, PHP)

ChatGPT Custom GPT (GPT Builder)

File: ai-context/chatgpt-custom-gpt.md
Limit: 8,000 characters for system prompt
File size: ~4,500 characters — well within limit
Requires: ChatGPT Plus, Team, or Enterprise (paid plan)

Setup

Go to chat.openai.com → click your avatar → My GPTs
Click Create a GPT → go to the Configure tab
In the Instructions field, paste the full contents of ai-context/chatgpt-custom-gpt.md
Give the GPT a name (e.g. “GEO Optimizer”) and save

What to expect

Compressed but complete GEO knowledge: workflow, all 9 methods, schema templates
Starts every site query with the audit command
Generates ready-to-paste code blocks
Slightly less verbose than Claude Projects version (due to space constraints)

Note: Custom GPTs require a paid ChatGPT plan. If you’re on the free plan, use Custom Instructions instead.

ChatGPT Custom Instructions

File: ai-context/chatgpt-instructions.md
Limit: 1,500 characters per field (2 fields = 3,000 max)
File size: ~800 characters — fits in a single field
Requires: Free or paid ChatGPT account

⚠️ Honest limitation: Custom Instructions have a 1,500 character limit per field. The file covers only the essentials: role, workflow steps, script commands, and top 3 methods. For full context (all 9 methods, schema templates, framework examples), use the Custom GPT file (requires paid plan) or Claude Projects (free with account).

Setup

Go to chat.openai.com → click your avatar → Customize ChatGPT
Under “What would you like ChatGPT to know about you?” paste the contents of ai-context/chatgpt-instructions.md
Save. Active for all new conversations.

What to expect

ChatGPT will understand GEO context without re-explaining every time
Knows the 3 script commands and top 3 Princeton methods
Does NOT have the full 9-method table, schema templates, or framework examples
For complex tasks, paste relevant sections from ai-context/claude-project.md directly into the chat

Cursor

File: ai-context/cursor.mdc
Format: .mdc with YAML frontmatter
Limit: No character limit
Activation: Rules load when files match the globs pattern

Setup

mkdir -p .cursor/rules
cp ~/geo-optimizer-skill/ai-context/cursor.mdc .cursor/rules/geo-optimizer.mdc

Or copy manually:

In your project root, create .cursor/rules/ directory
Copy ai-context/cursor.mdc → .cursor/rules/geo-optimizer.mdc
Cursor loads the rule automatically when you open matching files

The globs pattern activates the rule for: *.html, *.astro, *.tsx, *.jsx, *.php, robots.txt, llms.txt, *.json

What to expect

Cursor AI (Cmd+K, Cmd+L, inline chat) will follow GEO rules when editing matching files
Imperative “Always/Never” format integrates well with Cursor’s rule system
Suggests correct schema types for each page type automatically
Reminds you to run the audit command before making recommendations

Frontmatter reference

description: "GEO Optimizer — make websites cited by AI search engines..."
globs: "**/*.html,**/*.astro,**/*.tsx,..."
alwaysApply: false  # Only applies when matching files are open

Change alwaysApply: true to load the rule in every Cursor session regardless of file type.

Windsurf

File: ai-context/windsurf.md
Format: Plain Markdown — NO YAML frontmatter
Limit: ~12,000 chars (UI activation)
Activation: Configured via Windsurf UI → Customizations → Rules (4 modes: Always On, Glob, Model Decision, Manual)

⚠️ Windsurf does NOT read YAML frontmatter from rule files — activation mode is configured in the Windsurf UI, not in this file. If you add YAML frontmatter here, Cascade reads it as literal text.

Setup

mkdir -p .windsurf/rules
cp ~/geo-optimizer-skill/ai-context/windsurf.md .windsurf/rules/geo-optimizer.md

Or copy manually:

In your project root, create .windsurf/rules/ directory
Copy ai-context/windsurf.md → .windsurf/rules/geo-optimizer.md
Open Windsurf → Customizations (top-right slider icon) → Rules
Find geo-optimizer.md in the list and set the activation mode

Activation modes (Windsurf UI)

Mode	When to use
Always On	Rule is injected into every Cascade AI context — most reliable
Glob	Rule activates only when files matching a pattern are open (e.g. `*/.html`)
Model Decision	Cascade decides when the rule is relevant
Manual	You manually enable/disable the rule per session

⚠️ Known Windsurf bug (2025): complex glob patterns sometimes fail to activate rules. If glob doesn’t work, use “Always On” as fallback.

What to expect

Same GEO “Always/Never” rules as the Cursor file — same content, UI-managed activation
Windsurf Cascade will follow GEO rules when editing HTML/Astro/robots.txt files
Activation is controlled via the UI panel, not via frontmatter in the file

Kiro

File: ai-context/kiro-steering.md
Format: Markdown with YAML frontmatter
Limit: No character limit
Activation: inclusion: fileMatch — loads only when matching files are open

Setup

mkdir -p .kiro/steering
cp ~/geo-optimizer-skill/ai-context/kiro-steering.md .kiro/steering/geo-optimizer.md

Or copy manually:

In your project root, create .kiro/steering/ directory
Copy ai-context/kiro-steering.md → .kiro/steering/geo-optimizer.md
Kiro loads the steering file automatically when files matching the pattern are open

What to expect

Same rules as Cursor/Windsurf — GEO “Always/Never” rules apply when editing HTML/Astro/robots.txt files
fileMatch inclusion means the steering doc is only loaded when relevant files are open (efficient)
Kiro will suggest correct schema types, remind you to run the audit command, and follow GEO best practices

Frontmatter reference

---
inclusion: fileMatch
fileMatchPattern:
  - "**/*.html"
  - "**/*.astro"
  - "**/*.tsx"
  - "**/*.jsx"
  - "**/*.php"
  - "**/robots.txt"
  - "**/llms.txt"
  - "**/*.json"
---

inclusion: fileMatch with a fileMatchPattern array is the Kiro equivalent of Cursor’s globs — the steering file is only injected into context when one of the matching files is active.

Platform Comparison

Platform	Context quality	Setup effort	Cost	Persistence
Claude Projects	⭐⭐⭐⭐⭐ Full	Upload file	Free account	Project-scoped
ChatGPT Custom GPT	⭐⭐⭐⭐ Compressed	Paste in builder	Paid plan	GPT-scoped
ChatGPT Custom Instructions	⭐⭐ Essentials only	Paste in settings	Free account	Account-wide
Cursor	⭐⭐⭐⭐ Rules format	Copy file	Free	Project-scoped
Windsurf	⭐⭐⭐⭐ Rules format	Copy file	Free	Project-scoped (UI activation)
Kiro	⭐⭐⭐⭐ Rules format	Copy file	Free	Project-scoped (fileMatch activation)

Updating context files

When the GEO Optimizer toolkit is updated, re-copy the relevant files:

# Update all IDE rule files
cp ~/geo-optimizer-skill/ai-context/cursor.mdc .cursor/rules/geo-optimizer.mdc
cp ~/geo-optimizer-skill/ai-context/windsurf.md .windsurf/rules/geo-optimizer.md
cp ~/geo-optimizer-skill/ai-context/kiro-steering.md .kiro/steering/geo-optimizer.md

# For Claude/ChatGPT: re-upload/re-paste the updated file manually

GEO Optimizer by Juan Camilo Auriti — https://github.com/auriti-labs/geo-optimizer-skill