---
title: "AI Agent Readiness Audit — Can AI Agents Read Your Website? | Code4X"
url: "https://code4x.dev/"
canonical: "https://code4x.dev/"
description: "Is your website ready for AI agents? Code4X audits 55 dimensions of agent discovery, readability, and interaction, then generates deployable fixes."
format: "markdown-mirror"
updated: "2026-09-29"
agent_ready: true
---

# AI Agent Readiness Audit

> Is your website ready for AI agents? Code4X audits 55 dimensions of agent discovery, readability, and interaction, then generates deployable fixes.

- **URL**: https://code4x.dev/
- **Format**: Machine-readable Markdown Mirror (RFC 9727 / Accept: text/markdown)
- **Publisher**: Code4X Agent Readiness Platform

---

============================================================= 1. HERO 

**IS YOUR WEBSITE READY FOR AI AGENTS?**

# Your site has a third audience.

> Most sites turn them away at the front door. You never find out, because a page that fails an agent still looks perfectly fine to you.

Find out what yours does.

Your domain
>

              Scan my site→

Free AI agent readiness audit. No email. Eight checks in 15 seconds.

## What your site was built for

1. 1995 Humans Your site: built layout · accessibility · brand
2. 2005 Search engines Your site: maybe sitemaps · schema · meta tags
3. 2026 AI agents Your site: nothing discovery · readability · interaction

 ============================================ 2. READINESS READOUT (section) 
 The 31 checks is the rubric size: static copy, not a reading. Every other
     figure here comes from the scanner — each carries data-stat so the backend
     can swap it, and data-stub so the pre-launch sweep for placeholder numbers
     stays one grep. The bar widths carry them too, because the drawing has to
     move when the number does. 

**HOW READY IS THE WEB?**

## Most sites are not ready.

> We scan public URLs against the same deterministic rubric. Here is what we have found so far.

Sites scanned

3,567

Checks per scan

31

Average score

53/100

### Most failed checks

1. Agent discoverability 99%
2. Agent readability 96%
3. Agent interaction 94%

### By outcome

Failing: 872

Partial: 1,996

Scoring 80+: 699

**failing
**partial
**scoring 80+

[Score my site→](/scan)
[How we score it](/registry)

 ============================================== 3. DOES SEO COVER THIS? 
 Three published figures, each with its source. Static copy, not scanner
     output — no data-stat, no data-stub. Every source is an outbound link: the
     arrow is decorative and the sr-only text is what announces the new tab, so
     nothing has to read the word "source" aloud. 

**DOES SEO COVER THIS?**

## Google reads you. Agents might not.

> Search crawlers render JavaScript. The crawlers behind ChatGPT, Claude and Perplexity do not. Your rankings were earned by a machine that reads your site differently.

42%

of JavaScript-rendered content never gets indexed by AI systems

[Learn more↗ at Onely, February 2026, opens in a new tab](https://www.onely.com/blog/seo-tips-for-dynamic-content-and-dynamic-sites/)

#1

on Google and still invisible to ChatGPT, Claude and Perplexity

[Learn more↗ at Vercel and MERJ, opens in a new tab](https://vercel.com/blog/the-rise-of-the-ai-crawler)

57.5%

of all web traffic is now machines, not humans

[Learn more↗ at Cloudflare, June 2026, opens in a new tab](https://radar.cloudflare.com)

 ============================================== 4. WHAT DO WE CHECK? 
 The four layer names and their check counts. design.md clears both for
     public pages; the enumerated registry behind them is not cleared and is
     not here. The counts are product fact, not scanner output — no data-stat,
     no data-stub. 

**WHAT DO WE CHECK?**

## Agents ask four different questions.

> Finding you, reading you, and acting on your site are three separate problems. A fourth watches what the engines say about you afterwards.

15Checks

### Discovery

Can an agent find you and understand what you offer?

16Checks

### Readability

Can an agent read your content without guessing?

13Checks

### Interaction

Can an agent complete a task on your site?

6Checks

### Representation

Can an agent describe you accurately, month after month?

 ================================================== 5. WHAT DO WE FIX? 
 The patch is an illustration of the fix layer, not a reading of anyone's
     site — static copy, no data-stat, no data-stub. It reuses the global .diff
     component, so the added lines expand in off the observer that is already
     watching every .diff on the page. 

**WHAT DO WE FIX?**

## We hand you the fix, with the finding.

> Every failed check has a fix, and the crawl already gathered what the fix needs. So we fix it for you.

robots.txt
Generated patch

- User-agent: *
- Disallow: /
 
+ User-agent: GPTBot
+ Disallow: /            # training
 
+ User-agent: ChatGPT-User
+ Allow: /               # citation
 
+ Content-Signal: ai-train=no, search=yes

Deployable files

written from your crawl

Code patches

matched to your stack

Written instructions

where a human has to decide

 ================================================ 6. WHAT DOES IT COST? 
 Three consequences of failing the checks above. Argument, not measurement:
     static copy, no data-stat, no data-stub, no figures to keep in sync. 

**WHAT DOES IT COST?**

## You just stop being remembered.

> The agent reads what it can, answers with whatever it found, and moves on.

### The answer goes to someone else

An agent recommends whoever it could read. There is no second page of results to be on.

### Wrong answers, your name on them

Stale structured data and unreadable pages mean an agent quotes a price or a claim you retired years ago.

### No way to catch up later

The months an agent skipped you are gone. Unlike ads, there is nothing to switch back on.

 ============================================== 7. ISN'T THIS JUST SEO? 
 A real table, so it is marked up as one: tool names are row headers, the
     two readings are column headers. Static copy, no data-stat, no data-stub.
     Nothing here names a competitor's internals — only what each tool
     publishes about itself. 

**ISN'T THIS JUST SEO?**

## Others tell you what is wrong.

> Most tools watch whether AI mentions your brand. Nobody checks whether an agent can use your site once it arrives, and nobody hands you the fix.

| Tool | What it measures | What it leaves you with |
| --- | --- | --- |
| Lighthouse | Four agentic checks, pass or fail | A pass rate, no score, nothing to act on |
| Ahrefs | Whether AI mentions your brand | A visibility number, and no reason for it |
| Semrush | Mentions and sentiment, on simulated prompts | A dashboard your developers cannot use |
| Code4X | Whether an agent can find, read and use your site | A score, the evidence behind it, and the files that fix it |

 ================================================ 8. WHERE DO WE START? 
 Prices are order-of-magnitude placeholders, not quotes — see the marker on
     each one. Static copy: no data-stat, no data-stub. 

**WHERE DO WE START?**

## Three ways in.

> Every engagement starts the same way. A scan, then a conversation with an engineer.

One-off

### Full audit

Every dimension scored, evidence for each lost point, the generated fix files, and an MCP connection to query it all.

 PLACEHOLDER PRICE — replace with real figure 

from $4,500

Delivered in 5 working days

Most common

### Retainer

Monthly re-audit on the same rubric, before and after deltas, and drift monitoring as the standards move.

 PLACEHOLDER PRICE — replace with real figure 

from $1,800/mo

Rolling, three-month minimum

Premium

### Live agent testing

A real agent driven through your key flows, with a record of where it succeeded and where it stalled.

 PLACEHOLDER PRICE — replace with real figure 

from $6,000

Scoped per site

Start here

### Free scan

Fifteen seconds, no email. You will know whether any of the above is worth a conversation.

[Scan my site→](/scan)
[Talk to an engineer](/contact)

 ================================================ EXPLORE ALL RESOURCES 
 Comprehensive site map linking all guides, tools, and resources 

**EXPLORE MORE**

## Guides, tools & resources

> Everything you need to understand and implement AI visibility — ten technical guides, eight free tools, deep research blogs, and resources to keep you current as standards evolve.

 TECHNICAL GUIDES 

### Technical guides

Start with the first four guides below — they cover the failures that account for most audits.

[Tier 1 #### llms.txt What it is, what the file looks like, and the honest state of the evidence.](/llms-txt)
[Tier 1 #### AI crawlers & robots.txt Block training, allow citations. Template for 13 named agents.](/ai-crawlers-robots-txt)
[Tier 1 #### JavaScript & AI crawlers Most AI crawlers never run JavaScript. How to check and fix it.](/javascript-and-ai-crawlers)
[Tier 1 #### Content negotiation Serve markdown instead of HTML — cuts reading cost by 80-99%.](/content-negotiation)
[Tier 2 #### Schema drift When structured data contradicts your visible page.](/schema-drift)
[Tier 2 #### Lighthouse agentic browsing What Google's experimental audit scores and what it skips.](/lighthouse-agentic-browsing)
[Tier 2 #### WebMCP Let sites expose callable tools to agents. Adoption is near zero.](/webmcp)
[Category #### Google AI Overviews What actually influences inclusion, and why there is no schema.](/ai-overviews)

 KEY CONCEPTS 

### Concepts & comparisons

[#### AEO without the hype What Answer Engine Optimization is, what works, and what is oversold.](/aeo)
[#### AEO vs GEO vs SEO They describe the same discipline. Why the naming fight exists.](/aeo-vs-geo)
[#### Agency vs in-house Where an AEO agency ends and where engineering needs to begin.](/aeo-agency-vs-engineering)
[#### For engineering leaders Why this matters to your stack and who needs to be involved.](/for-engineering-leaders)

 TOOLS & RESOURCES 

### Tools & reference

[Free #### Readiness scan Eight checks in 15 seconds, no email. Every result is verifiable.](/scan)
[Free #### Eight free tools Generators, calculators, and utilities for AEO implementation.](/tools)
[Articles #### Blogs & research Technical investigations into AI search, self-audits, and crawler behaviour.](/blogs)
[Reference #### Glossary 27 terms defined plainly: agent readiness, AEO, GEO, WebMCP and more.](/glossary)
[Reference #### FAQ Twenty direct answers on audits, pricing, and what is genuinely contested.](/faq)
[Reference #### All 55 checks The complete rubric: every dimension, every check, every layer.](/registry)
[Reference #### Resources & index A living tracker of standards, benchmarks, and drift log.](/resources)

 CASE STUDIES & SERVICES 

### Services & proof

[#### Services & pricing Four ways to work with us, from free scans to monthly retainers.](/services)
[#### Our own score Our site audited hourly, including the two checks we fail.](/proof)
[#### Case studies We launched in 2026. Real case studies will go here as we grow.](/case-studies)
[#### MCP delivery Query your audit via the Model Context Protocol. Agent-native reporting.](/mcp)

---
*Machine mirror served for AI agents, citation crawlers, and LLM assistants. Canonical URL: [https://code4x.dev/](https://code4x.dev/).*