15 points
DOM and layout stability
Initial structure, hydrated DOM and visual rendering remain coherent enough to prevent action errors.
Reference model
Measure whether a website is actually ready for AI agents
The Agentic Readiness Score turns a technical intuition into an observable grid: can an agent find the right resource, understand the interface, choose an action and execute it without critical ambiguity?
- Semantic HTML
- Accessibility
- Stability
- Controlled hydration
- Agentic web
Scoring model
A score out of 100, structured around execution risk
The score does not predict SEO ranking. It measures the operational ability of a site to be understood, navigated and operated by agents.
15 points
Visual / DOM / accessibility coherence
What is visible, what is coded and what is exposed in the accessibility tree express the same intent.
15 points
Actionable semantics
Links, buttons, fields and states use native elements and explicit programmatic names.
15 points
Agentic accessibility
Roles, labels, relationships, states and form errors let an agent understand available actions.
10 points
Interaction determinism
A critical action produces a predictable, readable result attached to the right context.
10 points
Hydration risk
Critical actions and content do not depend on late, fragile or divergent client-side state.
10 points
Visual hierarchy and affordances
Primary, secondary and dangerous actions are distinguishable without creating false interaction signals.
5 points
Machine discoverability
Key resources are available through sitemaps, internal links, structured data and readable machine surfaces.
5 points
WebMCP / agent APIs readiness
Complex transactional journeys are evaluated for future exposure as structured tools.
Lighthouse vs Pagup Score
What Lighthouse measures and what the Pagup Score adds
Lighthouse Agentic Browsing makes selected signals visible. The Pagup Score places them inside a broader grid: journeys, proof, governance, editorial coherence and real action.
Technical signal
Lighthouse detects indicators
Accessibility, stability, programmatic names, surfaces such as llms.txt and WebMCP readiness can signal an agentic construction level.
Journey risk
Pagup verifies real use
An agent must find the right service, understand context, choose a CTA, fill a form and interpret confirmation without critical ambiguity.
Governance
The score must remain inspectable
The result should prioritize remediation and prevent regressions, not produce a decorative grade or a ranking promise.
Related reading
Read the Lighthouse Agentic Browsing audit article and the Lighthouse vs Pagup Score matrix to separate tool audit, public score and journey verification.
Reading the result
Four levels of agentic maturity
The result should prioritize remediation, not create a cosmetic grade. A useful bad score is better than an unverified good one.
0-39
Fragile for agents
The site may be visually correct, but agents may misinterpret or mis-execute critical journeys.
40-59
Partially interpretable
Some signals are present, but the interface still relies too much on human context, JavaScript or fragile inference.
60-79
Agent-compatible
The main journeys are readable and actionable, but governance and testing still need to reduce regressions.
80-100
Agent-ready
The site exposes a stable, semantic, coherent and testable structure for humans, search engines and AI agents.
Self-diagnostic
Rapid estimate before an audit
Only check what you can technically verify. This estimate does not replace an audit: it helps determine whether the risk should be handled now.
Technical sources
Why this score exists now
The grid aligns with a documented shift: agents read pages through multiple channels and need to act on stable interfaces.
Agentic reading channels
Google describes screenshots, raw HTML and the accessibility tree as major reading paths for agents.
Read the web.dev documentationStability, accessibility and machine surfaces
Lighthouse now documents audits related to agentic browsing: layout stability, accessibility for agents, llms.txt and WebMCP.
Read the Chrome documentationNext step
Does your score deserve technical verification?
The agentic readiness audit checks real journeys, not only declared intentions.