🇺🇦 90% of paid reports support Ukraine — how

AI Agent Readiness Score

Will browser-using AI agents actually use your site?

A 0–100 score across 18 checks for how well your page is structured for ChatGPT Agent Mode, Claude Computer Use, Perplexity Comet, and Operator-class agents. Free scan, runs in seconds.

Free · No signup · Optional paid report supports Ukraine 🇺🇦

What we check

Two layers — browser-level UX + metadata discovery — for a complete agent-readiness picture.

UX layer · 7 checks

Semantic Interactive ElementsAI agents identify clickable things by tag name. <button> and <a> are unambiguous; styled <div> is not.

ARIA Roles & TabindexWhen semantic HTML is unavailable, role="button" + tabindex="0" gives the agent the same affordance.

Label–Input Associations<label for="…">, aria-label, or wrapping label. Placeholder text is not an accessible name.

Tap Target SizeInteractive elements should be at least 24×24px. Smaller targets are filtered out by visual analysis.

Ghost OverlaysTransparent layers covering interactive elements cause agents to click the wrong thing.

cursor: pointerA small but consistent visual signal that the element is interactive — both for humans and visual agents.

Layout Stability (multi-URL)When primary CTAs move >200px between pages, agents have to re-scan the entire viewport every time.

Discovery layer · 11 checks

robots.txt existsCrawlers expect /robots.txt at the root. Missing or non-200 is treated as ambiguous.

robots.txt — AI bot rulesExplicit Allow/Disallow for GPTBot, ClaudeBot, Google-Extended, PerplexityBot, ChatGPT-User, anthropic-ai, CCBot.

sitemap.xmlParseable XML with at least one <url> entry. Agents enumerate via sitemap instead of crawling link-by-link.

llms.txtAI-native equivalent of robots.txt per llmstxt.org — H1 title, blockquote summary, curated link sections.

llms-full.txtBonus: full content corpus as markdown. Agents can ingest the whole site in one request.

Markdown content negotiationServer responds with text/markdown when Accept: text/markdown is sent. Saves agents from HTML parsing.

MCP Server Card/.well-known/mcp.json describing tools an agent can invoke without parsing forms.

Agent Skills manifest/.well-known/agent-skills.json per agentskills.io — capability advertising.

WebMCP detectionwindow.WebMCP + [mcp-action] attributes let agents invoke page actions without simulating clicks.

OAuth discovery (RFC 8414)/.well-known/oauth-authorization-server so agents negotiate auth programmatically.

API catalog (RFC 9727)/.well-known/api-catalog so agents discover available APIs at a standard location.

Free vs Paid report

Free gives you the diagnostic. Paid report gives you a personalized treatment plan.

	Free	Paid
UX checks (7)	✅	✅
Discovery checks (11)	✅	✅
Tech stack detection (25+ techs)	✅	✅
Rule-based quick fixes	✅	✅
Personalized AI audit (Markdown)	—	✅
Stack-specific code examples	—	✅
Developer Spec for AI coding agents	—	✅
Email delivery + 12-month re-access	—	✅
Supports Ukraine	—	✅ 🇺🇦

Pay what you want — from $1 up to $50. 90% to Ukrainian charity, 10% to infrastructure. See where every UAH goes →

See an example paid report →

How charity works

1. You pay what you want$1 to $50 (or $2 to $100 for bundle). Slider on the report page, charged via Mono in UAH at NBU rate.

2. Your report is preparedA personalized audit, tailored to your detected stack. Usually takes 20-40 seconds.

3. 90% goes to UkraineTransferred to the Sternenko Community Foundation. Every transfer documented with a bank confirmation.

Open the live transparency report →

Why it matters

AI agents that drive a real browser (not just scrape HTML) are getting capable enough to buy, book, and submit on behalf of users. The same accessibility hygiene that helps screen readers helps these agents — but most marketing pages were never tested against them. A site that's hard for an agent to navigate quietly loses to one that isn't.

Compared to Cloudflare's tool

Cloudflare's isitagentready.com focuses on metadata standards (discovery, bot rules, commerce protocols). We focus on browser-level UX — semantic HTML, ARIA, layout stability — plus core discovery checks. We also detect your tech stack to personalize our paid reports. Run both for a complete agent-readiness picture.

Built by Anton Krokhmal

SEO Team Lead · AI Operations · Lisbon, Portugal

antonkrokhmal.com·LinkedIn