AI Ready Check
RU Русский

Security · Trust · AI readiness

How AI sees a website — and whether to trust it

One check, two answers. Visitors learn whether a site is safe before they pay or sign in. Owners see how AI sees the site — and exactly what to fix.

  • Rendered in an isolated browser
  • VirusTotal · Google Safe Browsing · URLhaus · OpenPhish
  • Probed as GPTBot, ClaudeBot and PerplexityBot
01 / Security & trust

Before you trust it

A real browser opens the homepage. We check threat databases, domain age, lookalike addresses, fake login and card forms, data leaks and the company behind the site.

02 / How AI reads it

What AI agents see

We request the site as eight AI crawlers and check robots rules, raw-HTML content, structured data, llms.txt, MCP and API discovery.

03 / Independent opinion

A second look by a model

A separate model estimates trust from identity, transparency and reputation. It never overrides a security warning.

For visitors

Check a site before you trust it

Fake shops and phishing pages look polished. The warning signs are in the evidence: a domain registered last week, a card form that sends data to another server, a sign-in page imitating a bank. We open the page in an isolated browser so you do not have to.

  • Threat databases: VirusTotal, Google Safe Browsing, URLhaus, OpenPhish
  • Domain age, registrar and certificate
  • Lookalike addresses and brand impersonation
  • Sign-in and card forms that send data elsewhere
  • Company identity, checked in public registries
  • Redirects, cloaking and fake “update” downloads
Example
secure-account-verify.top
8/100
High risk
  • Imitates a bank sign-in page
  • Card form sends data to another domain
  • Domain registered 6 days ago
  • Listed by OpenPhish
Do not enter passwords or card details on this site.
For owners

Ready when you are

AI Site Scorer is a public readiness scanner. Give it any domain and it runs real HTTP probes — as GPTBot, ClaudeBot, PerplexityBot and other agents — then checks the emerging standards that decide whether AI systems can find, read and cite your site: llms.txt, robots Content Signals, structured data, MCP and API catalogs. You get a graded scorecard, the exact failing checks, and how to fix them.

→ llms.txt · robots Content Signals · MCP
→ Real HTTP probes across 8 AI bot user-agents
→ Independent evidence-based AI trust evaluation
37%
Overall readiness
67 / 182 F
Poor
✓ Homepage reachable over HTTPS
✕ No CAPTCHA / bot challenge on homepage
✕ llms.txt (recommended)
Why it matters

Search is turning into answers

The web is shifting from ranked links to AI-generated answers — and those answers are assembled by machines that crawl, parse and judge your pages on their own terms. If an agent can’t reach, read or trust your site, you never appear in the reply a user sees, and you never find out why. This check shows exactly what today’s AI systems can and cannot do with your site — before a competitor’s page becomes the default answer instead of yours.

Answers, not ten blue links
Google’s top result is now a Gemini answer and DuckDuckGo shows AI summaries. Most users read the generated answer and never click through — if you are not in it, the visit never happens.
Meta tags are fading
Agents fetch your raw HTML, summarize it and rank it with their own algorithms. Google is deprecating FAQ rich results in 2026 — hand-tuned meta tags matter less than whether the content is actually there and readable.
Readiness is measurable
Bot access, llms.txt, Content Signals, structured data and MCP are concrete, testable signals. This tool probes each one as a real AI crawler and scores exactly what agents see — so “AI-ready” becomes a checklist, not a guess.
Where this is going

AI agents will soon browse — and buy

The next web is not just read by AI, it is acted on by it. Emerging standards like x402 (HTTP 402 Payment Required, with machine-readable terms) let autonomous agents complete a purchase end to end — discover a product, evaluate the seller and pay — without a human ever opening the page. Protocols such as MCP, agent cards and API catalogs are how an agent decides you are worth transacting with in the first place. Your business will depend directly on how AI systems see, trust and act on you.

x402
Agents that pay
With x402 and agentic-commerce rails, an agent can settle a transaction the moment it decides — checkout becomes a machine-to-machine call, not a form a person fills in.
MCP
Chosen before you’re seen
Agents shortlist and rank options for the user. If your site is unreadable or blocked, you are filtered out before a human is ever shown a choice.
REV
Revenue tracks AI-readability
As discovery, comparison and payment move to agents, your pipeline depends on how AI can read, trust and act on your site — readiness becomes revenue infrastructure, not SEO hygiene.
FAQ

Questions, answered

01 What does the check cover? +
Two independent checks run together. The security check opens the homepage in an isolated browser and looks for threats: threat-database listings, domain age, lookalike addresses, fake sign-in or card forms, data sent to third parties, malicious downloads and the company behind the site. The AI-readiness check requests the site as eight AI crawlers and tests the standards that decide whether AI systems can reach, read and cite it.
02 How is the security score calculated? +
Each piece of evidence adds or removes points from a neutral 50. Hard evidence — a listing in Google Safe Browsing, VirusTotal or a phishing feed, a sign-in form that posts to another domain, data sent to a collector — caps the score at high risk, however polished the site looks. Missing data is shown as lower evidence coverage, not as danger.
03 Does a good score mean the business is honest? +
No. Automated checks find known warning signs and confirm public facts; they cannot prove that a shop delivers or that a company is honest. Check payment and refund terms before you buy.
04 Is this the same as SEO? +
No. Classic SEO optimizes for ranked links; the readiness check measures whether autonomous agents can fetch and use your site at all. A site can rank well in Google and still be invisible to AI assistants — for example if a WAF challenges every bot user-agent.
05 Do I need to install anything? +
No. Both checks read only public pages, the way a visitor or a crawler would. No code, no tags, no access to your server.
06 How is the readiness score calculated? +
Each check carries points weighted by severity (Critical, Major, Minor) and grouped into five categories: Discoverability, Content, Bot access, Policy and Capabilities. Your score is the sum of passed checks against the maximum for your profile, converted to a percentage and an A–F grade.
07 What is the AI trust opinion and why is it separate? +
A separate language model estimates trust from identity, transparency and reputation. It is an opinion, not a measurement, so it never changes the readiness grade and never overrides a security warning.
08 How often can I re-scan? +
Saved results are reused for 7 days. A fresh check is free within a small daily limit per visitor, which keeps the load on scanned sites reasonable.
09 Are results public, and what do you keep? +
Scores and summaries are public so anyone can look a site up. A security scan keeps a homepage screenshot and page evidence for 40 days; after that only aggregate scores remain. Site owners can ask us to remove a security report.

Check any website

One check, no sign-up: a security verdict for visitors and an AI-readiness grade with exact fixes for owners.