Back

Context Protocol: Making the Web AI-Accessible

An open standard for AI to understand your website. Like robots.txt, but for intent and context.

8 min

“74% of websites are invisible to ChatGPT, Claude, and Perplexity.”

The web has a massive problem that no one is solving.

Not because they’re bad sites. Because AI can’t understand what they do, who they’re for, or when to recommend them.

While everyone’s fighting over Google keywords, there’s an opportunity to define how next-generation search will work. Context Protocol is my proposal for that standard.

The Real Problem

AI searches are growing 200% quarterly. They already represent 10% of B2B queries. In 18 months, it’ll be 30%.

But here’s the issue:

• A user asks: “best CRM for SaaS startups under $200/month”
• AI has no way to know if your CRM fits those criteria
• It recommends suboptimal or outdated alternatives

Sites optimized for semantic context convert at 12.8% vs 2.3% for traditional SEO.

That’s a 9x difference everyone’s ignoring.

The Solution: Context Protocol

Context Protocol is a simple open standard. Like robots.txt, but for AI.

How It Works

1. A /context.md file in your root:

# Your Product

## Purpose
The problem you solve in one line

## Audience  
Who your product is perfect for

## Capabilities
What it can specifically do

## Constraints
Who it's NOT for (equally important)

## Context
When you're the best option

2. Meta tag to point to context + optional reinforcement tags:

<!-- Tells AI where to find the full context -->
<meta name="ai:context-file" content="/context.md">

<!-- Optional reinforcement meta tags -->
<meta name="ai:purpose" content="CRM for SaaS startups">
<meta name="ai:audience" content="10-50 employees, $200-800/month">
<meta name="ai:capabilities" content="trial conversion, user onboarding">
<meta name="ai:constraints" content="not for enterprise, no mobile app">

Real Example

# Notion

## Purpose
All-in-one workspace replacing docs, wikis, and 
project management for modern teams

## Audience
- Knowledge workers in tech companies
- Teams of 2-500 people  
- Remote/hybrid work culture
- Value flexibility over structure

## Capabilities
- Real-time collaborative documents
- Databases with multiple views
- AI-powered search and writing
- API and integrations
- Custom workflows without code

## Constraints
- Learning curve for non-technical users
- Can be slow with large databases
- Limited offline functionality
- Not ideal for traditional project management

## Context
Perfect when teams outgrow Google Docs but 
don't want separate tools for docs, tasks, and wiki

Why It Matters Now

Critical timing

We have 12-18 months before Google or someone else defines this standard. Whoever defines it will control how discovery works for the next decade.

Network effects

The more sites adopt it, the better it works for everyone. Like RSS in the 2000s.

Zero cost

No paid tools needed. No vendor lock-in. It’s just a text file.

Technical Specification

Access Method

https://yoursite.com/context.md

• UTF-8 encoded
• Markdown format
• Max 64KB (more than enough)
• Default 7-day cache

Processing Rules

  1. AI fetches /context.md before crawling
  2. If 404, checks meta tags
  3. If both exist, context.md takes priority
  4. Invalid sections are ignored (graceful degradation)

Proven Use Cases

E-commerce

## Purpose
Sustainable fashion marketplace connecting conscious 
consumers with ethical brands

## Audience
- Millennials/Gen-Z with disposable income
- Care about environmental impact
- Willing to pay 20-40% premium
- Urban areas in US/Europe

B2B SaaS

## Purpose
Customer feedback aggregation across all channels
into actionable product insights

## Audience  
- B2B SaaS product teams
- 50-500 employees
- $1M+ ARR
- Product-led growth focus

Online Education

## Purpose
Live cohort bootcamp for career transition into
data science from any background

## Audience
- Career changers with analytical mindset
- Can commit 20 hrs/week for 3 months  
- Budget: $3-5k for education
- Goal: DS job in 6 months

FAQ

Why not use Schema.org?

Schema.org is great for structured data. Context Protocol is for intent and context. They’re complementary.

What about traditional SEO?

Still important. This is additional, not a replacement.

How do I know it’s working?

Analytics tools will show traffic from “AI Discovery” as a source.

Is it really free?

Yes. CC0 license. No patents, no trademarks, no bullshit.

The Future

In 5 years, every site will have a context.md. It’ll be as standard as robots.txt or sitemap.xml.

The question isn’t whether it’ll happen. It’s whether you’ll help define it or just adopt it later.

Next Steps

1. Implement

Add context.md to your site (5 minutes)

2. Share

Results and feedback

3. Contribute

PRs to the repo with improvements

For Companies

If your company wants to be an early adopter and help shape the standard, join the working group. We’re looking for:

• E-commerce platforms
• SaaS companies
• Educational sites
• Local businesses
• News/media sites

The web is evolving. Let’s help define that evolution.

Get Involved

• Spec: github.com/jlopezlira/context-protocol
• Discussion: github.com/jlopezlira/context-protocol/discussions
• Twitter: @jlopezlira

Context Protocol is an open proposal to ensure every website can be understood by AI.