SentientWeb Readiness Scores · 2026-09-05.draft-1
Evidence before a score.
Four distinct assessments describe how a sampled website helps people, search systems, and supported agents understand and use a business.
Draft methodology, published 5 September 2026. These proposed weights require calibration and do not predict rankings, citations, leads, or revenue.
What 100 means
100 means all applicable criteria passed in the declared sample; it does not guarantee rankings, citations, sales or whole-site coverage.
We do not average the four scores: some criteria overlap, and the questions they answer are different. Noncustomers and customers use exactly the same criteria. Installing SentientWeb or another particular brand never earns points by itself.
Unknown results stay visible
A criterion can pass, partially pass, fail, remain unknown, or be not applicable. Partial credit is 50% of the criterion’s weight. A not-applicable finding requires evidence and a reason.
For applicable weight W, assessed weight T, unknown weight U, and earned points E, coverage is T/W. The score interval runs from 100 × E/W to 100 × (E + U)/W.
A worked example
60 earned points, 80 assessed weight, and 20 unknown weight produce 60–80/100 with 80% coverage. The score cannot become 75/100 by excluding the unknown checks.
A final single score requires every applicable weighted check to be assessed. No assessed evidence means “Not assessed.” Critical authorization or transaction failures remain visible independently of the number.
Agent readiness
Business and capability understanding · 20 points
Evidence: Services, constraints and meaningful action descriptions are accessible and consistent.
Applicability: Applicable to the declared business tasks and sampled pages; exclusions require current evidence and a reason.
Pass: All relevant checks supported: Services, constraints and meaningful action descriptions are accessible and consistent.
Partial: Business meaning is clear but material task constraints are incomplete.
Fail: Evidence demonstrates material failure of the relevant checks; inaccessible or untested checks are unknown, not failed.
agent.understanding · semantic · evidence expires after 30 days
Action discovery and input quality · 20 points
Evidence: Applicable tools can be discovered and their input schemas and required fields validate.
Applicability: Applicable to the declared business tasks and sampled pages; exclusions require current evidence and a reason.
Pass: All relevant checks supported: Applicable tools can be discovered and their input schemas and required fields validate.
Partial: Some applicable schemas or descriptions are incomplete.
Fail: Evidence demonstrates material failure of the relevant checks; inaccessible or untested checks are unknown, not failed.
agent.discovery · deterministic · evidence expires after 30 days
Verified task completion · 35 points
Evidence: Representative authorized tasks return authoritative outcomes in the declared environment.
Applicability: Applicable to the declared business tasks and sampled pages; exclusions require current evidence and a reason.
Pass: All relevant checks supported: Representative authorized tasks return authoritative outcomes in the declared environment.
Partial: Some representative tasks complete; others have evidenced failures.
Fail: Evidence demonstrates material failure of the relevant checks; inaccessible or untested checks are unknown, not failed.
agent.completion · authorized journey · evidence expires after 7 days · critical check
Permissions and reliable recovery · 15 points
Evidence: Identity isolation, confirmation, stale-state handling and meaningful recovery have been exercised.
Applicability: Applicable to the declared business tasks and sampled pages; exclusions require current evidence and a reason.
Pass: All relevant checks supported: Identity isolation, confirmation, stale-state handling and meaningful recovery have been exercised.
Partial: Identity boundaries hold but tested noncritical recovery paths fail.
Fail: Evidence demonstrates material failure of the relevant checks; inaccessible or untested checks are unknown, not failed.
agent.permissions · authorized journey · evidence expires after 7 days · critical check
Interoperability and human fallback · 10 points
Evidence: Claimed interfaces work in recorded clients and a usable human fallback exists.
Applicability: Applicable to the declared business tasks and sampled pages; exclusions require current evidence and a reason.
Pass: All relevant checks supported: Claimed interfaces work in recorded clients and a usable human fallback exists.
Partial: Only part of the claimed interface or fallback scope works.
Fail: Evidence demonstrates material failure of the relevant checks; inaccessible or untested checks are unknown, not failed.
agent.interoperability · authorized journey · evidence expires after 7 days
SEO readiness
Crawl access and indexability signals · 25 points
Evidence: Sampled status codes, robots, canonical and noindex signals permit intended indexing; actual index inclusion is not inferred.
Applicability: Applicable to the declared business tasks and sampled pages; exclusions require current evidence and a reason.
Pass: All relevant checks supported: Sampled status codes, robots, canonical and noindex signals permit intended indexing; actual index inclusion is not inferred.
Partial: Some intended indexable sample pages have inconsistent directives.
Fail: Evidence demonstrates material failure of the relevant checks; inaccessible or untested checks are unknown, not failed.
seo.indexability · deterministic · evidence expires after 30 days
Site structure and navigation · 20 points
Evidence: Relevant pages are reachable through useful internal links and declared sitemap entries are consistent.
Applicability: Applicable to the declared business tasks and sampled pages; exclusions require current evidence and a reason.
Pass: All relevant checks supported: Relevant pages are reachable through useful internal links and declared sitemap entries are consistent.
Partial: Some relevant pages are difficult to discover or sitemap entries are inconsistent.
Fail: Evidence demonstrates material failure of the relevant checks; inaccessible or untested checks are unknown, not failed.
seo.structure · deterministic · evidence expires after 30 days
Page meaning and content quality · 25 points
Evidence: Titles and headings accurately distinguish useful page content without material duplication.
Applicability: Applicable to the declared business tasks and sampled pages; exclusions require current evidence and a reason.
Pass: All relevant checks supported: Titles and headings accurately distinguish useful page content without material duplication.
Partial: Some pages are useful but meaning or differentiation is incomplete.
Fail: Evidence demonstrates material failure of the relevant checks; inaccessible or untested checks are unknown, not failed.
seo.content · semantic · evidence expires after 30 days
Mobile delivery and performance · 20 points
Evidence: Recorded mobile rendering and labeled lab measurements meet the evaluator's pinned thresholds; field data is separate.
Applicability: Applicable to the declared business tasks and sampled pages; exclusions require current evidence and a reason.
Pass: All relevant checks supported: Recorded mobile rendering and labeled lab measurements meet the evaluator's pinned thresholds; field data is separate.
Partial: Usability or lab performance misses some pinned thresholds.
Fail: Evidence demonstrates material failure of the relevant checks; inaccessible or untested checks are unknown, not failed.
seo.delivery · deterministic · evidence expires after 30 days
Applicable structured data · 10 points
Evidence: Relevant structured data validates and agrees with visible content.
Applicability: Applicable semantic types supported by the visible business content; documented absence of relevant types may be N/A.
Pass: All relevant checks supported: Relevant structured data validates and agrees with visible content.
Partial: Relevant data exists with noncritical validation or completeness issues.
Fail: Evidence demonstrates material failure of the relevant checks; inaccessible or untested checks are unknown, not failed.
seo.structured_data · deterministic · evidence expires after 30 days
GEO readiness
Clear business identity · 20 points
Evidence: Business, services, geography where relevant, and contact identity are clear.
Applicability: Applicable to the declared business tasks and sampled pages; exclusions require current evidence and a reason.
Pass: All relevant checks supported: Business, services, geography where relevant, and contact identity are clear.
Partial: Identity is recognizable but relevant details are ambiguous.
Fail: Evidence demonstrates material failure of the relevant checks; inaccessible or untested checks are unknown, not failed.
geo.identity · semantic · evidence expires after 30 days
Consistency of important facts · 20 points
Evidence: Material pricing, eligibility, policy and identity facts agree across the declared sample.
Applicability: Applicable to the declared business tasks and sampled pages; exclusions require current evidence and a reason.
Pass: All relevant checks supported: Material pricing, eligibility, policy and identity facts agree across the declared sample.
Partial: Some facts need clarification without material contradiction.
Fail: Evidence demonstrates material failure of the relevant checks; inaccessible or untested checks are unknown, not failed.
geo.consistency · semantic · evidence expires after 30 days
Original facts and useful evidence · 25 points
Evidence: Concrete explanations, original facts and important claims have relevant support.
Applicability: Applicable to the declared business tasks and sampled pages; exclusions require current evidence and a reason.
Pass: All relevant checks supported: Concrete explanations, original facts and important claims have relevant support.
Partial: Some useful claims are supported while others lack material evidence.
Fail: Evidence demonstrates material failure of the relevant checks; inaccessible or untested checks are unknown, not failed.
geo.evidence · semantic · evidence expires after 30 days
Source traceability and maintenance · 20 points
Evidence: Material claims are attributable with useful source links, ownership and appropriate update context.
Applicability: Applicable to the declared business tasks and sampled pages; exclusions require current evidence and a reason.
Pass: All relevant checks supported: Material claims are attributable with useful source links, ownership and appropriate update context.
Partial: Some sources or maintenance context are missing.
Fail: Evidence demonstrates material failure of the relevant checks; inaccessible or untested checks are unknown, not failed.
geo.traceability · semantic · evidence expires after 30 days
Accessible extractable information · 15 points
Evidence: Relevant public information is meaningfully extractable in the tested retrieval context; training opt-outs are not search penalties.
Applicability: Applicable to the declared business tasks and sampled pages; exclusions require current evidence and a reason.
Pass: All relevant checks supported: Relevant public information is meaningfully extractable in the tested retrieval context; training opt-outs are not search penalties.
Partial: Some relevant information requires unavailable rendering or lacks clear structure.
Fail: Evidence demonstrates material failure of the relevant checks; inaccessible or untested checks are unknown, not failed.
geo.accessibility · deterministic · evidence expires after 30 days
AEO readiness
Relevant buyer question coverage · 30 points
Evidence: A recorded category-specific question set is answered across relevant service pages, regardless of FAQ labeling.
Applicability: Applicable to the declared business tasks and sampled pages; exclusions require current evidence and a reason.
Pass: All relevant checks supported: A recorded category-specific question set is answered across relevant service pages, regardless of FAQ labeling.
Partial: Some relevant buyer questions lack answers.
Fail: Evidence demonstrates material failure of the relevant checks; inaccessible or untested checks are unknown, not failed.
aeo.coverage · semantic · evidence expires after 30 days
Directness and clarity · 25 points
Evidence: Answers to the recorded buyer questions are readily locatable and understandable.
Applicability: Applicable to the declared business tasks and sampled pages; exclusions require current evidence and a reason.
Pass: All relevant checks supported: Answers to the recorded buyer questions are readily locatable and understandable.
Partial: Some answers are indirect or ambiguous.
Fail: Evidence demonstrates material failure of the relevant checks; inaccessible or untested checks are unknown, not failed.
aeo.clarity · semantic · evidence expires after 30 days
Accuracy support and completeness · 25 points
Evidence: Answers include material qualifications and agree with supported business facts.
Applicability: Applicable to the declared business tasks and sampled pages; exclusions require current evidence and a reason.
Pass: All relevant checks supported: Answers include material qualifications and agree with supported business facts.
Partial: Some answers omit material qualifications or supporting evidence.
Fail: Evidence demonstrates material failure of the relevant checks; inaccessible or untested checks are unknown, not failed.
aeo.accuracy · semantic · evidence expires after 30 days
Structure and next-step usefulness · 20 points
Evidence: Answers are organized with useful context and appropriate working next-step links.
Applicability: Applicable to the declared business tasks and sampled pages; exclusions require current evidence and a reason.
Pass: All relevant checks supported: Answers are organized with useful context and appropriate working next-step links.
Partial: Some answers lack context, organization or useful next steps.
Fail: Evidence demonstrates material failure of the relevant checks; inaccessible or untested checks are unknown, not failed.
aeo.structure · semantic · evidence expires after 30 days
Technology is not an outcome
Integration cards distinguish Verified, Detected, Not detected, Not applicable, and Not tested. Verification describes a documented test scope. A responding endpoint does not prove a purchase can complete. Google channel approval is tracked separately from a UCP implementation.
Anthropic commerce-agents is an optional development reference, not a compatibility requirement. An unsupported browser leaves WebMCP untested. Expired evidence cannot support a current verified claim.
Scope, versions, and appeals
Each saved report records its sampled pages, assessment date, methodology version, evidence, and limitations. Public scans do not establish private account permissions, search indexing, or real-user performance. Historical scores retain their original methodology; changes between versions are not automatically improvement trends.
This is the first published draft. Future methodology changes will receive a new version and a documented change record. Semantic evaluations require anchored grading and evidence, with calibration against human review before scale.
To challenge a finding, use the correction and appeal contact options with the report link and supporting evidence. Owner declarations remain separate from independent findings. Directory inclusion does not require reciprocal links or payment.