← AnswerReady scannerMethodology, API, and evidence
AnswerReady rubric 2026-08-22.1 is a 100-point model. A score is emitted only when every criterion has source evidence and the point total can be recomputed. Score and coverage are separate: skipped, not applicable, unavailable, and error outcomes reduce coverage without being counted as passes or failures.
Scoring rubric
Pass earns the full weight, warn earns half, and fail earns zero. Skipped, not applicable, unavailable, and error outcomes are excluded from the score denominator and lower coverage instead. Category weights are derived from the registered criteria and total 100.
AI Discovery
25 pointsPublic files and policies that let answer engines discover the site.
llms.txt is available
A concise, root-level content map gives answer engines an explicit discovery surface.
llms-full.txt is available
A bounded full-text companion gives agents richer context without scraping the entire site.
AI crawler policy is explicit
Explicit crawler rules are more trustworthy than assuming access from an absent policy.
MCP discovery manifest is valid
A valid MCP manifest advertises real, agent-operable capabilities.
Agent policy is published
A machine-readable agent policy defines safe public and protected behavior.
AgentReady full test suite is verified
A second, purpose-built scanner provides independent evidence for agent discovery and safe operation.
Crawlability
20 pointsSignals that let crawlers safely find, select, and index canonical pages.
robots.txt is available
robots.txt is the baseline crawl-policy surface.
XML sitemap is available
A valid sitemap provides bounded, explicit URL discovery.
robots.txt references the sitemap
The reference joins crawl policy and URL discovery.
Canonical URL is declared
Canonical selection reduces duplicate and conflicting source identities.
Page is indexable
A page cannot be cited consistently if page-level policy blocks indexing.
Structured Meaning
18 pointsMachine-readable semantics used for entities, previews, and API discovery.
JSON-LD is syntactically valid
Valid JSON-LD is the strongest page-level machine-readable entity signal.
JSON-LD identifies useful entity types
Typed entities are more useful than syntactically valid but untyped graphs.
JSON-LD has complete, linked entities
Parseable markup still needs schema.org context, typed entities, core properties, and stable identifiers to be dependable.
Open Graph metadata is complete
Stable title, description, image, and type metadata improve source previews.
Social card metadata is available
A card declaration supports consistent previews across secondary discovery channels.
Machine-readable API schema is available
OpenAPI or MCP discovery lets agents verify available operations instead of guessing.
Answer Quality & Trust
23 pointsContent, attribution, and freshness signals that support defensible citations.
Title is descriptive
A concise title establishes the page's primary answer context.
Description summarizes the page
A useful description provides a compact statement of intent and scope.
A single primary heading is present
One clear H1 gives parsers a stable primary topic.
Content has useful depth
Substantive visible text is more likely to contain answerable, citable detail.
Content exposes answer structure
Question headings, lists, tables, and FAQ markup make answer units easier to extract.
Authorship or publisher is attributable
Attribution supports provenance and confidence in a citation.
Claims link to supporting sources
Outbound source links provide inspectable provenance for factual claims.
Freshness is machine-readable
Published or modified dates help engines reason about time-sensitive claims.
Reach & UX
14 pointsSyndication, locale, transport, mobile, and response-time readiness.
Document language is declared
Language metadata improves parsing, accessibility, and locale selection.
RSS or Atom feed is available
Feeds provide durable, incremental discovery for changing content.
Mobile viewport is configured
Mobile usability affects both readers and mobile-first indexing.
Primary HTML responds promptly
Fast server response improves crawl efficiency and user access.
Final page uses HTTPS
Secure transport is a baseline trust and integrity requirement.
External PlatPhorm validation
Insights and SitemapOps inputs are schema-validated before transmission. Responses must be successful JSON, remain inside a 4 MB bound, and pass the adapter schema. Otherwise the lens is marked degraded and contributes no claimed external score.
Scan modes and ownership
Quick runs the core HTTP suite plus local agent signals and the point-in-time AgentReady lane. Standard adds Insights, SitemapOps, SearchOps, TrustOps, AgentOps, and BrowserOps target validation. Full runs every configured lane and permits real BrowserOps execution. Browser emphasizes rendered Chromium and performance evidence. Network accepts bounded caller-network metrics separately. Developer emphasizes API, error, Markdown, MCP, versioning, rate-limit, OAuth, and SDK/CLI surfaces. A skipped lane remains visible in the unified provider matrix as not invoked; a failed invocation is reported as unavailable or unable to verify rather than as a target defect.
Provider ownership is explicit: SitemapOps crawls sitemap authority and sampled URLs; AgentReady evaluates point-in-time agent readiness; AgentOps counts and tracks machine-readable guidance; SearchOps evaluates external and on-site search quality; TrustOps evaluates identity, provenance, security, and well-known files; BrowserOps evaluates real browser behavior and contributes bounded browser timing/resource evidence; Insights evaluates performance and accessibility. AnswerReady reuses its bounded root sitemap artifact through authenticated same-target scan context when SitemapOps can accept it, while SitemapOps still owns child-sitemap discovery and route sampling.
Lifecycle and reproducibility
Every measured result retains source, method, observation time, applicability, confidence, and provider ownership. The expert evidence ledger is redacted. Use the API versioning guide for stable report negotiation and deprecation behavior.
Trusted scanners can submit public-safe criterion observations to MCP tool answerready.score_evidence for deterministic rubric validation without another network fetch. This returns a receipt to the caller and does not create a durable AnswerReady report; answerready.scan_site and POST /api/scan remain the report-producing surfaces.
Unified scanner topology
AnswerReady is the canonical rollup. AgentReady supplies point-in-time agent evidence; AgentOps owns longitudinal health; SitemapOps owns sitemap evidence; SearchOps owns external discoverability; TrustOps owns identity and provenance; BrowserOps owns real-browser and optional caller-network evidence; Insights supplies performance and accessibility measurements. Provider evidence is normalized with provenance, applicability, confidence, and failure ownership before rubrics are evaluated.
The optional caller network plane measures the browser-to-test-edge path only. It does not change target-site performance or readiness scores, and packet loss is reported as unavailable unless an approved TURN service is configured.
Schema quality and PlatPhorm compliance
JSON-LD quality checks parseability, schema.org context, typed entities, core properties for known entity types, and stable absolute identifiers. PlatPhormNews targets also receive a separate public read-only contract assessment for standard routes, health envelopes, discovery, OpenAPI and shared auth, trust, MCP ping, trace state, Vercel metadata state, and secret hygiene. Network-only checks do not change the general readiness score and are not applied to external sites.
API, MCP, feeds, and discovery
OpenAPI is available as YAML or JSON. MCP metadata and the JSON-RPC transport are public read-only surfaces. Stored reports are available as JSON or negotiated Markdown at /api/v1/scans/{scanId}.