Agentic Website Readiness Lab
Score a bundled sample marketing site against an agent-readiness checklist inspired by Chrome WebMCP and agent-readable web surfaces, then read the cheapest diff that moves it one tier up.
What this proves
Agent-readability is a six-dimension scorecard with a diff, not a vibe. The lab proves the rubric using bundled sample hosts so every score is reproducible.
How it works
What This Proves
Google I/O 2026 made the browser and the web feel more agent-operable. The follow-up question is harder than it sounds: is my site actually ready? Most teams cannot point at a number.
This lab pins down a number. Pick a bundled sample host, get a six-dimension score (llms.txt, markdown twins, agent UA policy, schema, server structure, answerability), and read the minimum diff that would push the next-quarter score one tier up.
What This Build Does
- Loads one of the bundled dummy hosts (no live fetch, no crawl, no claim about a real site)
- Runs a deterministic rubric so the same host always returns the same score
- Prints the cheapest upgrade plan as a copyable artifact
Safety Boundaries
- No live fetch. The lab never opens a connection to a real site.
- No model call. The rubric is a fixed lookup so the score is reproducible.
- No PII. The form only accepts a host name from the sample list.
What Would Come Next
- Add the conversion plan view that emits the diff as a one-PR checklist
- Wire the rubric to a side-by-side compare for two sample hosts
- Land the rubric as a static JSON so other surfaces can score against the same definition
Get new builds, breakdowns, and useful AI updates.