Your website just got a new performance metric, and it's not about page speed. On April 17, Cloudflare shipped isitagentready.com, a public scanner that scores any URL on how prepared it is for AI agents. Paste a domain, get a number out of 100, see which checks passed and which failed. For the first time, "agent readiness" moved from gut feeling to measurable infrastructure.
The score matters for marketing executives because AI agents are becoming a primary channel for content discovery. When Claude Code, Cursor, or ChatGPT retrieves your documentation, product pages, or thought leadership content, the format and discoverability of that content directly affects whether your brand gets cited in AI-generated answers. Cloudflare's analysis of 200,000 top domains found that only 3.9% of sites support markdown content negotiation, and only 4% have declared their AI usage preferences. The window to differentiate is open.
But here's the problem: the composite score is structurally misleading if you stop reading after the number.
What the Scanner Actually Checks
The tool runs sixteen checks across five categories. Understanding what each category measures is the difference between chasing a vanity metric and making decisions that affect pipeline.
Discoverability covers the basics: robots.txt, sitemap, and Link headers. These tell agents where your content lives and how to navigate it. 78% of sites have a robots.txt, but most are written for traditional search crawlers, not AI agents.
Content Accessibility tests whether your site serves markdown when an agent requests it via the Accept: text/markdown header. Cloudflare's own benchmarks show an 80% token reduction when serving markdown instead of HTML. A blog post that consumes 16,180 tokens as HTML drops to 3,150 as markdown. For AI systems processing millions of pages, that's a significant cost and latency saving.
Bot Access Control checks whether your robots.txt includes AI-specific rules and Content Signals, a draft standard for declaring how you want your content used: training, search indexing, or live agent answers.
Protocol Discovery is where the score gets interesting for B2B companies. This category checks for MCP Server Cards, API Catalogs (RFC 9727), OAuth discovery, and Agent Skills indexes. Cloudflare found that MCP Server Cards and API Catalogs together appear on fewer than 15 sites in their entire dataset. It's early.
Commerce checks for emerging payment protocols like x402 that let agents transact directly with websites.
The Preset Problem
One practitioner ran the same website through the scanner and got 33 or 67 depending on which preset they selected. The tool offers three modes: All Checks, Content Site, and API/Application. A content-only blog that scores poorly on API Catalog and OAuth discovery isn't failing at agent readiness; it's being measured against checks that don't apply to its use case.
This is the critical insight for marketing executives: the composite number is meaningless without context. A marketing site doesn't need an MCP Server Card. A developer documentation portal probably does. Before you brief your team on "improving our agent readiness score," understand which checks actually matter for your site type.
What Actually Moves the Needle
For most B2B marketing sites, three interventions deliver the highest return:
Markdown content negotiation. When an agent sends Accept: text/markdown, serve clean markdown instead of HTML. Vercel measured a 99.6% reduction in page content size when serving markdown to agents. If you're on Cloudflare, their Markdown for Agents feature handles this at the CDN layer with no origin changes. If you're not, your engineering team can implement content negotiation at the application layer.
Content Signals in robots.txt. Add a single line declaring your AI usage preferences: Content-Signal: search=yes, ai-input=yes, ai-train=yes. This tells agents whether you consent to training, search indexing, and live answer generation. The implementation is a one-liner, and it signals intent to every agent that reads your robots.txt.
Link headers for discovery. HTTP responses can include a Link header that points agents to your sitemap, llms.txt, and other discovery files without requiring them to parse HTML. A single rule in your CDN configuration can wire up all the discovery endpoints an agent needs.
The CFO Question
Your CFO will ask: what's the ROI of agent readiness? The honest answer is that we don't have clean attribution data yet. One study of 381 pages across 6 websites found no statistically significant increase in bot traffic from serving markdown. But that study measured traffic, not citations. The more relevant question is whether your content gets cited in AI-generated answers, and that's harder to measure.
What we do know: agents are already a meaningful traffic source. One agency reports 3x more agent visits than human visits per day. If your content is invisible to agents because it's buried in HTML soup, you're missing a channel that's growing faster than organic search.
The investment is modest. Most of the high-impact changes are configuration, not development. Markdown content negotiation on Cloudflare is a toggle. Content Signals is a line in robots.txt. Link headers are a CDN rule. The engineering lift is measured in hours, not sprints.
A Two-Week Pilot
Run the scanner on your primary marketing domain. Select the "Content Site" preset unless you're running an API or application. Document which checks failed and why. Then prioritize:
- Week 1: Implement Content Signals in robots.txt and Link headers for discovery. Both are configuration changes with no code required.
- Week 2: Enable markdown content negotiation. If you're on Cloudflare, this is a feature toggle. If not, scope the engineering work and decide whether the token reduction justifies the investment.
Measure agent traffic before and after. Track whether your content appears in AI-generated answers for your target queries. The data will tell you whether to invest further.
The score is a diagnostic, not a destination. Use it to find the gaps that matter for your site type, close them, and move on. The agents are already reading your content. The question is whether they're reading it efficiently.
