What is an Agent Readiness score?

Agent Readiness is a 0–100 score for how well a website lets autonomous AI agents (ChatGPT agent mode, Perplexity Comet) access, parse, trust, and act on it. GetVisus scores six pillars: agent access, machine-actionable schema, answer extractability, trust and identity, transaction integrity, and protocol interoperability (llms.txt and AGENTS.md).

What did GetVisus fix on its own site?

Three things: we added an AGENTS.md file (it was missing), added BreadcrumbList structured data, and exposed a top-level Offer in JSON-LD so buying agents can act on it. Together these took our Agent Readiness score from 84 to 100.

AGENTS.md is an emerging plain-text file at a site's root that tells AI agents how to interact with the site: entry points, available actions, identity, and contact. It is the companion to llms.txt — llms.txt summarises a site for indexing, AGENTS.md guides agents that take actions.

Case study

We ran GetVisus on GetVisus: Agent Readiness 84 → 100

We pointed our own AI-visibility tool at getvisus.com. It flagged three agent-readiness gaps. We fixed all three — an AGENTS.md file, BreadcrumbList schema, and a top-level Offer — and our Agent Readiness score went from 84 to 100. Here is exactly what we changed, with the code.

By Zeb Choudhry · GetVisus · Updated August 2026

The baseline

Before touching anything, we ran a full audit of getvisus.com through our own engine. The site already scored well — no surprise, it is what we build for — but the Agent Readiness score exposed real gaps in how ready we are for autonomous AI agents (ChatGPT agent mode, Perplexity Comet) that browse, compare, and act on a customer's behalf.

Metric	Baseline
Overall AI-visibility score	93 / 100 — Protected
Agent Readiness	84 / 100
— Machine-actionable schema	16 / 25
— Transaction integrity	5 / 10
— Protocol interoperability	3 / 5

The single open finding was blunt: "No /AGENTS.md — the emerging standard for telling agents how to interact with your site." The lower schema and transaction pillars pointed at missing structured data. So we had a clear, honest to-do list — on our own site.

Fix 1 — Add an AGENTS.md file

AGENTS.md is the companion to llms.txt. Where llms.txt summarises a site for AI models, AGENTS.md tells autonomous agents how to act on it: entry points, available actions, how to verify identity, and how to make contact. We publish an AGENTS.md generator for customers — and did not have one ourselves. We do now, at getvisus.com/AGENTS.md.

# AGENTS.md — GetVisus

## What an agent can do here
- Run a free audit: submit a URL on the homepage for a 0–100 score and fixes.
- Generate files: free tools produce a site's llms.txt and AGENTS.md.

## Key entry points
- Free audit (home): https://getvisus.com/
- Agent readiness checker: https://getvisus.com/agent-readiness-checker
...

Fix 2 — Add BreadcrumbList structured data

Breadcrumbs give both Google and AI agents a clean map of where a page sits. We had rich JSON-LD already (Organization, SoftwareApplication, FAQPage) but no BreadcrumbList. One small block closes that gap.

{
  "@context": "https://schema.org",
  "@type": "BreadcrumbList",
  "itemListElement": [
    { "@type": "ListItem", "position": 1, "name": "Home", "item": "https://getvisus.com/" },
    { "@type": "ListItem", "position": 2, "name": "AI Visibility Audit",
      "item": "https://getvisus.com/ai-visibility-audit" }
  ]
}

Fix 3 — Expose a top-level Offer

This one was subtle, and it taught us something about our own tool. Our pricing was already in structured data — but nested inside our SoftwareApplication schema. An agent scanning for a machine-actionable Offer (with price and availability) would not reliably pick it up. So we added a clear top-level Offer that a buying agent can act on directly.

{
  "@context": "https://schema.org",
  "@type": "Offer",
  "name": "GetVisus Fix Pack",
  "price": "19",
  "priceCurrency": "GBP",
  "availability": "https://schema.org/InStock",
  "url": "https://getvisus.com/upgrade"
}

Honest footnote: this also flagged an improvement for the audit itself — it should credit Offers nested inside a SoftwareApplication, not only top-level ones. Dogfooding your own product is the fastest way to find its blind spots.

The delta

We re-scored getvisus.com with the same engine after the fixes:

Metric	Before	After
Overall score	93	94
Agent Readiness	84	100
AI Visibility	91	94
Machine-actionable schema	16 / 25	25 / 25
Transaction integrity	5 / 10	10 / 10
Protocol interoperability	3 / 5	5 / 5
Open findings	1	0

Three small, standards-based changes — an AGENTS.md file and two JSON-LD blocks — took Agent Readiness from 84 to a perfect 100, with no redesign and no new content. That is the point: agent-readiness is mostly about exposing signals machines can trust, not rebuilding your site.

What is your Agent Readiness score?

Run the same free audit on your site. Score out of 100, the exact gaps, and a generated AGENTS.md.

Run free audit

Why this matters now

AI agents are moving from answering questions to completing tasks — booking, comparing, and buying on a user's behalf. The sites those agents can access, parse, trust, and transact with are the ones that win the task. Almost no business has prepared for that yet. Being agent-ready before your competitors is the opportunity, and the fixes are cheaper than most people expect.