llms.txt at root
Scanner inventory row: llms.txt, stored through aiso.llms-txt when TOOLBOX captures it.
How AISO maps to Lighthouse Agentic Browsing.
Chrome Lighthouse added experimental agentic audits for WebMCP, llms.txt, accessibility, and stability signals. AISO treats those as useful overlap signals inside a broader hosted scanner, profile, MCP, PDF, and repair-plan product.
This page is not a Lighthouse pass/fail claim. The AISO evidence-capped score is separate: it only moves when scanner rows, generated files, provider checks, or persisted events support the result.
AISO maps overlap with Chrome Lighthouse's experimental Agentic Browsing audits. This is not a Google ranking claim, not a Search Console metric, and not a Lighthouse pass/fail certificate. llms.txt is optional machine-readable context for agents, not a Google Search ranking lever. WebMCP checks remain planned/aligned unless a browser audit captures real tool registration evidence.
WebMCP declarative tools and forms
Scanner rows aiso.webmcp.declarative_tools, aiso.webmcp.forms, and aiso.webmcp.schema_validity. AISO public profiles also emit their own declarative Ask metadata on the profile form.
WebMCP imperative registration
Public profile Ask registers through guarded document.modelContext or navigator.modelContext when the browser exposes it. Detection of imperative_tools on third-party sites is a Phase 1 HTML-regex approximation; Phase 2 swaps in browser page.evaluate() introspection.
Agent-centric accessibility and layout stability
Scanner rows aiso.agent_a11y_tree and aiso.agent_layout_stability. Both are Phase 1 HTML-derived approximations today; Phase 2 swaps in real page.accessibility.snapshot() and PerformanceObserver CLS measurements.
Beyond the Lighthouse alignment above, the AISO scanner ships eleven additional modules in its offer. Each is registered today as a scanner row inside the current inventory, returning real evidence from TOOLBOX HTTP probes and HTML inspection. Two modules (Performance and Accessibility) currently use Phase 1 HTML-derived approximations for the browser-only signals (LCP / INP / CLS / axe-core); Phase 2 swaps those for Chromium measurements without changing the contract surface. No fake zeros, no fake scores: every row either has real evidence or honestly labels itself a Phase 1 approximation.
Website Health
HTTP / redirect / HTTPS, render success, JS console errors, broken links, failed resources, metadata, schema validity, forms, mobile viewport, security headers.
Performance & Core Web Vitals
LCP, INP, CLS, TTFB measured in the browser visit. Page weight, render-blocking CSS/JS, image optimization on the paid tier.
AI Bot Policy & Training Consent
noai / noimageai meta, ai.txt, X-Robots-Tag per-bot, TDM-Reservation, License-Allow-Train. Builds on the existing robots.txt per AI bot rows.
Structured Data Richness
Required vs recommended Schema.org field coverage, Google rich-result eligibility (Product, FAQ, HowTo, Article, LocalBusiness, etc.), schema graph connectivity.
Accessibility (Agent + Human Inclusive)
Contrast, alt text, heading hierarchy, ARIA correctness, keyboard navigation, focus indicators, prefers-reduced-motion. The same surface humans and agents both depend on.
Action Path Maturity
Booking, payment, contact, and quote flows detected from the page and scored for agent-usability. Cross-references WebMCP overlap. No flows are submitted.
Authority & Off-Page Signals
sameAs social graph, Wikipedia / Wikidata entity presence, third-party review platforms, Common Crawl backlink summary, RDAP domain age.
Privacy & Tracker Audit
Third-party tracker inventory, cookie banner correctness, privacy policy freshness, PII-in-URL patterns, do-not-sell links.
Identity & Business Verification
Logo / favicon / PWA manifest brand kit, physical address and contact block, optional LEI / Companies House / SEC EDGAR registration lookup, team-page evidence.
Citation Substrate
Author bylines, visible publish/update dates, FAQ density, citation-worthy claim count, glossary density. Predicts how citable the content is to an LLM.
Internationalization
hreflang correctness, html lang and language switcher detection, regional content surfacing. Deep tier only.