Cloudflare launched an agent readiness score. It checks four things. Faro checks thirty-plus. Before you read this as a vendor comparison, understand what that gap actually means — because it matters for how you decide what to fix on your site.
Cloudflare is one of the most important pieces of infrastructure on the internet. More than 20% of all websites run behind it. When Cloudflare ships something in the AI agent space, it gets attention — and it should. But the specific problem it's solving is not the same problem most businesses have when they discover that ChatGPT, Claude, and Perplexity cannot find or recommend them.
What Cloudflare's Agent Readiness Score Checks
Cloudflare's agent readiness score focuses on infrastructure access signals: whether your site blocks AI crawlers, whether you have an agents.json file, whether you accept payment protocols like x402, and whether your robots.txt configuration permits agent traffic. These are real, necessary checks. If you're blocking GPTBot or ClaudeBot in robots.txt, no amount of structured data or llms.txt optimization will matter — the agent cannot read your site at all.
The Cloudflare check answers a binary question: can an agent get to your site? That's the infrastructure layer, and it's the first thing to fix. But it's not the only thing. In Faro's scan data, 41% of sites that pass infrastructure access checks still score below 50/100 overall. They let agents in; they just give them nothing useful to work with.
What Happens After the Agent Gets In
An AI agent that can access your site then has to understand it. That's a different problem entirely. It reads your pricing page and needs to find structured, machine-readable pricing — not a design-heavy page with slider widgets and modal comparisons. It checks your llms.txt file for context about what you do, who you serve, and what your API looks like. It looks for JSON-LD structured data to confirm basic facts about your business. It checks whether you have an OpenAPI spec, an MCP server, or OAuth documentation if it's trying to take an action rather than just read about you.
None of these are infrastructure questions. They're content and signal quality questions. Cloudflare doesn't audit them because Cloudflare is an infrastructure platform. That's not a criticism — it's an accurate description of scope.
The 7 Categories Faro Checks
Faro's AI readiness scan runs 30+ checks across six categories. Here's how each one maps to what an AI agent actually needs:
| Category | What it checks | Cloudflare covers? |
|---|---|---|
| Agent Permissions | robots.txt AI stanzas, GPTBot, ClaudeBot, Gemini, Content-Signal header | Partial |
| Content Discoverability | Sitemap, llms.txt, link headers, canonical tags | No |
| Structured Data | JSON-LD, sameAs, OKF bundle, Open Graph | No |
| Pricing Transparency | Machine-readable pricing, visible plans, /pricing link in nav | No |
| Agent Infrastructure | MCP endpoint, A2A, OpenAPI, Auth.md, OAuth, agents.json, x402 | Partial |
| Trust & Identity | Contact page, About page, Privacy policy, Terms of service | No |
| Grounding Quality | llms.txt content depth, OKF completeness, structured data accuracy | No |
Based on Faro's 30+ check audit engine vs. Cloudflare's public agent readiness score documentation.
The Pricing Transparency Problem Is Bigger Than You Think
Pricing transparency is one of the checks that most surprises people when they see it in a Faro audit. It seems obvious: if you sell software, put your pricing on your website. But the version of that check relevant to AI agents is more specific.
An AI agent evaluating whether to recommend your product needs to answer the question: "How much does this cost, and is there a plan that fits the buyer's budget?" If your pricing page uses a JavaScript-rendered toggle between monthly and annual, requires a modal click to see enterprise pricing, or buries the per-seat cost in the FAQ, the agent cannot answer that question reliably. It will either skip your product or return an inaccurate price.
Faro's pricing transparency check looks for machine-readable pricing (in llms.txt, in pricing.md, or in structured data), a direct link to your pricing page from the main navigation, and visible plan names with associated prices. This is the kind of check that directly affects whether an AI agent recommending software to a buyer includes your product in the comparison set.
Structured Data: The Signal Most Sites Get Wrong
Structured data is not a new concept. Schema.org has defined markup standards for organizations, products, pricing, and articles for over a decade. Google has supported JSON-LD for nearly as long. But most sites still get it wrong in ways that specifically matter for AI agents.
The most common issue Faro finds is missing sameAs links. A well-structured organization schema includes links to your LinkedIn page, Twitter/X profile, GitHub repository, and Crunchbase entry. These sameAs links let an AI agent cross-reference multiple sources to confirm that what it found on your site is accurate. Without them, the agent has only one source for your claims — your own website, which carries less evidentiary weight than corroborated sources.
Beyond basic JSON-LD, Faro checks for the Open Knowledge Format (OKF) bundle — a structured directory at /okf/ that organizes machine-readable versions of your pricing, team, products, and API documentation in a format purpose-built for LLM consumption. Very few sites have this yet. The ones that do give AI agents a far richer signal to work with.
MCP and the Action Layer
The Model Context Protocol (MCP), published by Anthropic and now supported across Claude, GPT-4o, and other leading models, defines how AI agents can take actions on your behalf. An agent with MCP access can query your product catalog, check pricing, start a trial, or retrieve account information — without a human clicking through your website.
Cloudflare checks whether your site has an agents.json file (which declares agent compatibility) and whether you support x402 payment protocols. Faro checks four states: MCP card and live endpoint both present, card only (declared but not active), endpoint only (active but not declared), and neither. The distinction matters because an agent that finds a card pointing to a non-functional endpoint fails silently — the agent moves on to your competitor.
In Faro's current scan data, fewer than 6% of sites have a functional MCP endpoint. That gap represents a significant near-term competitive opportunity for the businesses that move first.
How to Use Both Tools
The right approach is not either/or. Start with Cloudflare's agent readiness check to confirm the infrastructure layer is sound. If you're blocking AI crawlers, fix that first — nothing else matters until agents can access your site. Then run a Faro audit to understand what those agents find once they get in.
The sequence is: access, then signal quality. Cloudflare handles the access layer well. Faro audits the signal quality layer. A business that only passes the Cloudflare check but fails the structured data, pricing transparency, and llms.txt checks is in the same position as a shop that unlocks the front door but leaves the shelves empty.
What the Average Site Scores
Across Faro's scan database, the average AI readiness score is 41 out of 100. Fewer than 12% of sites score above 70. The most common failures, in order of frequency, are: missing or incomplete llms.txt (63% of sites), no machine-readable pricing (71%), missing structured data beyond basic JSON-LD (58%), and no MCP or OpenAPI endpoint (94%).
These are not infrastructure access failures — they are signal quality failures. They exist independently of whether Cloudflare is in the picture at all.
In Short
Cloudflare's agent readiness score is a useful infrastructure check. If you block AI crawlers, it will find that. But the checks that determine whether an AI agent understands you, trusts your pricing, and recommends you to a buyer are in a different layer — one that requires a dedicated AI readiness audit. The two tools solve adjacent problems. Neither replaces the other.
Scan your site at byfaro.ai to see your full score across all 30+ checks. Results in 30 seconds, no account required.
FAQ
Does Cloudflare's agent readiness score replace Faro?
No. Cloudflare checks infrastructure access: can an agent reach your site at all? Faro audits signal quality: what does that agent find when it gets there? Both checks are necessary, and they cover different failure modes.
What are the most important checks for AI agent visibility?
In order of impact, based on Faro scan data: AI crawler permissions (robots.txt), llms.txt presence and quality, machine-readable pricing, JSON-LD structured data with sameAs links, and MCP/OpenAPI endpoint availability. Missing any of the first three will significantly reduce an agent's ability to understand and recommend your business.
How does Cloudflare's AI bot management affect agent readiness?
Cloudflare's AI bot management lets you control which AI crawlers can access your site. If you use it to block GPTBot or ClaudeBot, those agents cannot index your site. If you allow them, Cloudflare's role in agent readiness ends there — the quality of what they find is entirely determined by your content, structured data, and documentation.
What is an OKF bundle?
The Open Knowledge Format is a structured directory at /okf/ on your site that provides machine-readable versions of your pricing, team, products, and API documentation in a format designed for LLM consumption. Use Faro's OKF Generator to build one in minutes.