Cloudflare's Agent Readiness Score: Lighthouse for the Agentic Web
Cloudflare has launched a new Agent Readiness score and public adoption dataset to measure how well websites support AI agents. Here's what the data says and why web teams should care.
The web learned how to serve browsers. Then it learned how to serve search engines. Now it is starting to learn how to serve AI agents.
On April 17, Cloudflare launched Agent Readiness, a new scoring tool at isitagentready.com that checks whether a website exposes the signals and interfaces modern AI agents need to navigate, authenticate, and consume content efficiently. At the same time, Cloudflare added a new AI agent standards adoption dataset to Cloudflare Radar, giving the industry an early measurement of how prepared the public web actually is.1
The short version: not very.
According to Cloudflare's scan of 200,000 popular domains, 78% of sites have a robots.txt file, but most are still written for traditional crawlers rather than AI agents. Only 4% declare AI usage preferences using Content Signals, and just 3.9% support Markdown content negotiation by returning text/markdown when agents request it. More advanced discovery standards, including MCP Server Cards and API Catalogs defined in RFC 9727, appeared on fewer than 15 sites in the dataset.1
That gap is exactly what makes this announcement interesting. Cloudflare is not just shipping another AI feature. It is trying to turn a loose collection of emerging conventions into something web teams can measure, benchmark, and improve.
What the score actually checks
Cloudflare groups its checks into four categories:1
- Discoverability:
robots.txt,sitemap.xml, and HTTPLinkheaders defined in RFC 8288 - Content: whether pages can be served as Markdown for agents
- Bot access control: AI bot rules, Content Signals, and Web Bot Auth
- Capabilities: machine-discoverable APIs and agent-facing interfaces such as API Catalogs, OAuth discovery, MCP Server Cards, and WebMCP
In practice, that means Agent Readiness is less about SEO-style scoring and more about whether a site can participate in the emerging agentic web. Can an agent discover your docs? Can it fetch a cheaper, cleaner representation than raw HTML? Can it tell what it is allowed to do? Can it find structured capabilities instead of scraping your UI and guessing?
Those are real architectural questions now, not future ones.
Why this matters beyond Cloudflare
This is also a notable shift in how web standards spread. Historically, standards move slowly until somebody makes them visible and operational. Performance best practices existed before Lighthouse, but Lighthouse made them legible to teams and managers. Cloudflare is clearly aiming for the same effect here.
The company even built the service to be agent-consumable itself. The site publishes a stateless MCP endpoint and an Agent Skills index so agents can scan sites programmatically and learn how to fix failures.1 That detail matters because it turns the score from a static audit into something coding agents can act on.
It also builds on Cloudflare's earlier push for Markdown for Agents, announced in February, which converts HTML pages into Markdown when agents send the right Accept header. Cloudflare said that can reduce token usage by up to 80% on some pages, which is meaningful for any workflow where models repeatedly fetch documentation or knowledge-base content.2
The bigger takeaway
What stands out is not just the product, but the data. If Cloudflare's Radar numbers are directionally right, the public web is still in the very early innings of agent compatibility. Most sites can be visited by agents, but very few are designed for them.
That creates an unusual opportunity. For the first time in a while, small teams can adopt a new web pattern before large incumbents do. Supporting Markdown negotiation, publishing structured API metadata, and making access rules explicit are all practical improvements. They make sites easier for agents to use, but they also make interfaces cleaner and more machine-readable in general.
Whether Agent Readiness becomes the Lighthouse of the agent era is still an open question. But the underlying point is solid: if AI agents are going to become a real traffic source, websites need better contracts than scraped HTML and hope.
And right now, most of the web is still missing those contracts.
Sources
- Cloudflare: Introducing the Agent Readiness score
- Cloudflare Radar: AI Insights
- Cloudflare: Markdown for Agents
- RFC 9727: The API Catalog
- RFC 8288: Web Linking