---
license: cc-by-4.0
language:
- en
pretty_name: Agentic Readiness Index
tags:
- seo
- geo
- generative-engine-optimization
- ai-agents
- agentic-web
- webmcp
- llms-txt
- lighthouse
- web-standards
size_categories:
- n<1K
configs:
- config_name: default
  data_files: "agentic_readiness_index.csv"
---

# The Agentic Readiness Index

**How ready is the web for AI agents?** This dataset measures the *agentic readiness* of leading websites: how well an AI agent (ChatGPT, Perplexity, an agentic browser) can read, understand, and act on a page. It is the open companion to the study *"Built for Eyes, Not Agents"* (DOI: 10.13140/RG.2.2.10508.48000).

Maintained by **Rahul Kumar** (AI Researcher & Founder, Incrementors). Open data, CC BY 4.0. Contributions and re-runs welcome.

## Why this exists
Search visibility is measured exhaustively. Page speed is measured continuously. **Agentic readiness — whether the software agents now acting for users can actually use your site — is barely measured at all.** As AI agents mediate a growing share of discovery and purchase, this becomes a first-class metric. The Index is an open, reproducible baseline and a living leaderboard.

## How it's measured (fully reproducible)
Google Lighthouse 13.3.0, experimental `agentic-browsing` category, one command per site:

```
npx lighthouse@13.3.0 <url> --only-categories=agentic-browsing --chrome-flags="--headless --no-sandbox"
```

The overall **agentic_score** is a pass-ratio across checks that map to what an agent must do:

| Capability | Check column | Meaning |
|---|---|---|
| Read | `accessibility_tree` | Is the machine-readable structure well-formed |
| Understand | `llms_txt` | Valid `llms.txt` telling AI what the site is |
| Act | `webmcp` | Browser-side WebMCP tools registered |
| Trust | `cumulative_layout_shift` | Layout stable, not a moving target |

`PASS` / `FAIL` / `NA` (not applicable, e.g. capability absent). Scores are single-run on the home page and the category is experimental, so treat individual numbers as indicative; the value is the pattern across the set.

## Columns
`rank`, `tier` (ai-native / big-brand / pub-saas / smb), `site`, `url`, `agentic_score`, `accessibility_tree`, `cumulative_layout_shift`, `llms_txt`, `webmcp`.

## Key findings (v1, 22 sites, June 2026)
- Median score **0.50**; mean 0.55. Even leading sites are half-ready at best.
- **0 of 22** registered browser-side WebMCP — the "act" layer is essentially unbuilt.
- **68%** failed the accessibility-tree (read) check.
- Only **23%** served a valid `llms.txt`.
- **82%** passed layout stability — the human-era metric the industry already optimizes.
- **Budget does not predict readiness:** the two lowest scores are martech giants Salesforce (0.10) and HubSpot (0.19); a small business scored above Apple, Nike, and Walmart.

## Leaderboard (v1, June 2026)

| Rank | Site | Score | Read (a11y) | Trust (CLS) | Understand (llms.txt) | Act (WebMCP) |
|---|---|---|---|---|---|---|
| 1 | OpenAI | 1.0 | PASS | PASS | NA | NA |
| 2 | Stripe | 1.0 | PASS | PASS | PASS | NA |
| 3 | Target | 1.0 | PASS | PASS | PASS | NA |
| 4 | Vercel | 0.67 | FAIL | PASS | PASS | NA |
| 5 | Notion | 0.67 | PASS | PASS | FAIL | NA |
| 6 | Shopify | 0.67 | FAIL | PASS | PASS | NA |
| 7 | Kaire & Heffernan | 0.67 | PASS | PASS | FAIL | NA |
| 8 | isaWOW | 0.67 | FAIL | PASS | PASS | NA |
| 9 | Incrementors | 0.67 | PASS | PASS | FAIL | NA |
| 10 | Nike | 0.52 | PASS | FAIL | NA | NA |
| 11 | Perplexity | 0.50 | FAIL | PASS | NA | NA |
| 12 | Apple | 0.50 | FAIL | PASS | NA | NA |
| 13 | Coca-Cola | 0.50 | FAIL | PASS | NA | NA |
| 14 | New York Times | 0.50 | FAIL | PASS | NA | NA |
| 15 | Glo Medspa | 0.50 | FAIL | PASS | NA | NA |
| 16 | Emerald Pools | 0.50 | FAIL | FAIL | NA | NA |
| 17 | Anthropic | 0.33 | FAIL | PASS | FAIL | NA |
| 18 | Cursor | 0.33 | FAIL | PASS | FAIL | NA |
| 19 | Walmart | 0.33 | FAIL | PASS | FAIL | NA |
| 20 | Arch Dental | 0.33 | FAIL | PASS | FAIL | NA |
| 21 | HubSpot | 0.19 | FAIL | FAIL | FAIL | NA |
| 22 | Salesforce | 0.10 | FAIL | FAIL | NA | NA |

## Citation
> Kumar, R. (2026). *Built for Eyes, Not Agents: Measuring How Ready 22 of the World's Leading Websites Are for the Agentic Web.* DOI: 10.13140/RG.2.2.10508.48000

## Roadmap
v1 = 22 sites. Planned: expand to 100+ sites, add per-industry indices, re-run quarterly to track the web's agentic readiness over time. To suggest a site or contribute a re-run, open a discussion.

## Author
Rahul Kumar (@4AMKarmYogi) — AI Researcher & Founder, Incrementors. AI SEO · GEO · applied LLMs · AI automation. incrementors.com · linkedin.com/in/rahulaimarketing · researchgate.net/profile/Rahul-Kumar-543