Vista by Lara logo

Vista AI Authority Engine™ Documentation

AI Authority Score™ Methodology

The AI Authority Score™ is Vista by Lara's proprietary framework for evaluating how clearly, reliably, and credibly an organization's digital presence can be discovered, interpreted, verified, and considered by AI-powered search and recommendation systems.

This framework is independently developed by Vista by Lara. It is not an official metric created, endorsed, or used by Google, OpenAI, Anthropic, Microsoft, Perplexity, or any other third-party platform.

Version
v1.0
Released
2026-07-13
Last reviewed
2026-07-13
Status
Active
Review status
Internally human reviewed
Reviewing body
Vista by Lara Engineering
Independent external review
Not yet commissioned
View scoring categoriesView limitationsRequest an audit

Score scale (0-100)

0100
  • 0 to 20: Critically Underdeveloped -- The digital presence lacks basic clarity, accessibility, structure, evidence, or reliable entity identification.
  • 21 to 40: Limited Readiness -- Some foundations exist, but major technical, content, evidence, or authority gaps prevent reliable interpretation.
  • 41 to 60: Developing Readiness -- The organization has partial structure and useful content, but implementation remains inconsistent or insufficiently validated.
  • 61 to 80: Strong Readiness -- The organization is generally understandable, accessible, and well structured, with meaningful evidence and authority signals.
  • 81 to 90: Advanced Readiness -- The organization demonstrates mature technical architecture, strong entity clarity, evidence quality, and broad recommendation readiness.
  • 91 to 100: Exceptional Readiness -- The organization demonstrates highly developed technical, entity, content, evidence, governance, and authority architecture.

Full band definitions: see Score interpretation scale.

On this page

Overview

Why the score exists. Businesses increasingly need to know whether their digital presence can actually be understood by AI systems, not only by human visitors. The AI Authority Score exists to make that readiness measurable and explainable rather than assumed.

What it measures. Twelve weighted categories covering technical readiness, entity clarity, structured data, content quality, evidence, external authority, and governance. See Scoring Model.

Who it is for. Business owners, marketing and technical teams, and agencies who want a documented, evidence-based view of AI discovery readiness rather than an opaque number.

How it should be used. As a diagnostic and prioritization tool: identify the categories with the largest gaps, address them in weighted order, and re-evaluate on the documented review cycle.

What it must not be interpreted as. A ranking guarantee, a certification, or proof that any AI platform currently recommends the business.

The score evaluates readiness and evidence. It does not certify that a specific AI platform will recommend a business for a specific prompt.

Definitions

Terms marked Vista by Lara are proprietary to this methodology. All other terms reflect common industry usage and are not presented as official standards.

Search Engine Optimization (SEO)
Improving a website's technical structure and content so traditional search engines can crawl, index, and rank it for relevant queries.
Answer Engine Optimization (AEO)
Structuring content so that AI-powered answer engines can extract direct, accurate answers to specific questions.
Generative Engine Optimization (GEO)
Optimizing content, entities, and evidence so generative AI systems are more likely to reference or cite an organization in synthesized answers.
AI Visibility
The general degree to which an organization can be discovered, described, and cited by AI-powered systems.
Entity Optimization
Making an organization, its founder, and its services identifiable and disambiguated as distinct entities in structured data and content.
Machine Readability
The degree to which a page's structure and markup (schema, headings, semantic HTML) can be reliably parsed by automated systems.
Knowledge Authority
The credibility of an organization's published information, based on expertise, citations, and editorial standards.
Recommendation ReadinessVista by Lara term
Vista by Lara's term for how well-prepared an organization's digital presence is to be considered by an AI recommendation system, distinct from whether it is actually being recommended.
External Authority
Validation of an organization that comes from independent third parties rather than the organization itself.
Citation Visibility
Whether and how often an organization is referenced as a source by AI-generated answers.
Prompt VisibilityVista by Lara term
Vista by Lara's term for whether an organization appears in the response to a specific, logged AI prompt.
Evidence QualityVista by Lara term
The strength and verifiability of the evidence behind a claim, as classified by the four-tier evidence hierarchy documented in this methodology.
Trust Signals
Observable markers -- reviews, verified contact information, policies, transparency -- that indicate an organization is legitimate and accountable.

What the score measures

The score evaluates twelve dimensions of AI discovery readiness. Each is documented in full, with its evaluation criteria and evidence sources, in Scoring Model below.

  1. 01 Entity Architecture and Consistency
  2. 02 Technical Accessibility and Crawlability
  3. 03 Structured Data and Machine Readability
  4. 04 Content Clarity and Answer Extraction
  5. 05 Service, Industry, and Geographic Relevance
  6. 06 Content Authority and Expertise
  7. 07 Evidence Quality and Case-Study Strength
  8. 08 Independent Authority and Citation Signals
  9. 09 Reviews, Reputation, and Trust Signals
  10. 10 AI Accessibility and Public Knowledge Resources
  11. 11 Multilingual and Regional Readiness
  12. 12 Freshness, Monitoring, and Governance

What the score does not measure

The AI Authority Score does not directly measure or guarantee any of the following. Separate, explicitly labeled monitoring is required for visibility and business outcomes.

  • Google organic rankings
  • Google AI Overview inclusion
  • ChatGPT recommendations
  • Gemini recommendations
  • Perplexity citations
  • Claude responses
  • Microsoft Copilot recommendations
  • Traffic
  • Leads
  • Sales
  • Revenue
  • Conversion rate
  • Customer satisfaction
  • Market leadership
  • Legal compliance
  • Cybersecurity certification
  • Data privacy compliance
  • Future algorithm behavior

Scoring model

Twelve categories, weighted by importance to AI discovery readiness, sum to exactly 100 points.

CategoryWeightAutomationHuman review
Entity Architecture and Consistency12 ptsHybridRequired
Technical Accessibility and Crawlability10 ptsAutomatedNot required
Structured Data and Machine Readability10 ptsHybridRequired
Content Clarity and Answer Extraction10 ptsHybridRequired
Service, Industry, and Geographic Relevance8 ptsHybridRequired
Content Authority and Expertise10 ptsHybridRequired
Evidence Quality and Case-Study Strength10 ptsManualRequired
Independent Authority and Citation Signals10 ptsHybridRequired
Reviews, Reputation, and Trust Signals7 ptsHybridRequired
AI Accessibility and Public Knowledge Resources5 ptsAutomatedNot required
Multilingual and Regional Readiness4 ptsHybridRequired
Freshness, Monitoring, and Governance4 ptsHybridRequired
Total100 ptsWeights validated to sum to 100 at build time

Entity Architecture and Consistency

12 points

AI systems reason about organizations as entities, not pages. If the organization, founder, services, and locations are not clearly and consistently connected, machines cannot confidently identify who is being described.

Point allocation

CheckPoints
Organization entity exists and is schema-valid2
Stable organization @id used consistently sitewide1.5
Founder identity linked to the organization1
Brand name used consistently in text and markup1
Business description consistent across pages1
NAP (name, address, phone) consistency1.5
Service relationships explicit1
Geographic relationships explicit1
sameAs links present and verified1
Entity disambiguation from similarly-named organizations1
Category total12

Evidence sources

  • • Organization and Person JSON-LD with matching @id references
  • • Consistent NAP (name, address, phone) across the site and directories
  • • sameAs links to verified profiles

Strong implementation

Every service, location, and person page references the same Organization @id, and that entity's sameAs list resolves to verified, active profiles.

Weak implementation

The homepage uses one business name, the footer uses an abbreviation, and no page links the founder to the organization in markup.

Common failures
Different legal or trading names used across pages; founder referenced with inconsistent titles; sameAs links pointing to unclaimed or unrelated profiles.
Zero-score condition
No machine-readable Organization entity exists anywhere on the domain.
Limitations
Consistency can be verified programmatically; whether a human reader finds the entity credible still requires judgment.

Technical Accessibility and Crawlability

10 points

A business cannot be evaluated by any system, human or machine, if its pages cannot be reliably reached, rendered, and read.

Point allocation

CheckPoints
HTTP accessibility of key pages1
Robots directives do not block legitimate crawlers1
Sitemap completeness and accuracy1
Canonical tag implementation1
Status codes: no redirect chains, no soft 404s1.5
Server-rendered or fully crawlable content1
Mobile accessibility0.5
Page-speed characteristics0.5
Bot accessibility, including AI crawler user agents1
WAF and edge behavior toward known crawlers0.5
Crawl depth from the homepage0.5
Duplicate-content controls0.5
Category total10

Evidence sources

  • • Direct HTTP requests and header inspection
  • • robots.txt and sitemap.xml review
  • • Rendered HTML comparison against source HTML

Strong implementation

All primary navigation, footer, and sitemap links resolve directly to HTTP 200 with no redirect hops.

Weak implementation

Internal links routinely pass through one or more redirects before reaching a live page.

Common failures
Internal links that pass through 301/302/307/308 redirects instead of resolving directly; client-side-only rendering that hides content from crawlers that do not execute JavaScript; sitemap entries that 404.
Zero-score condition
The domain returns non-200 status codes for its primary pages or blocks all crawlers.
Limitations
Automated checks confirm accessibility at the time of the crawl; intermittent outages or edge-network issues may not be captured.

Structured Data and Machine Readability

10 points

Structured data is the most direct language a website can use to tell a machine what it is looking at. Valid, complete schema reduces ambiguity for search and AI systems.

Point allocation

CheckPoints
JSON-LD validity (parses without errors)1.5
Required and recommended properties present1.5
Schema type appropriateness for page content1
Entity linking via @id rather than duplicated data1
Organization schema present sitewide1
Person schema for named authors or principals0.5
Breadcrumb schema present and accurate0.5
Article and Service schema where applicable1
FAQ schema only where FAQ content is visibly present0.5
Public machine-readable resources (e.g. llms.txt)0.5
Consistency between visible text and schema content1
Category total10

Evidence sources

  • • JSON-LD extracted and validated against schema.org definitions
  • • Manual comparison of visible content to markup

Strong implementation

A single Organization entity is referenced by @id from every page that needs it, and FAQPage schema exactly matches the visible FAQ text.

Weak implementation

Multiple pages each define their own Organization block with slightly different data.

Common failures
Schema that describes content not actually visible on the page; FAQPage schema with no corresponding visible FAQ; duplicate Organization entities with conflicting @id values.
Zero-score condition
No valid JSON-LD is present anywhere on the domain.
Limitations
Schema validity is objective; schema appropriateness for intent requires editorial judgment.

Content Clarity and Answer Extraction

10 points

AI answer engines extract short, direct answers from longer pages. Content that never states a clear answer is difficult to cite, regardless of how accurate it is.

Point allocation

CheckPoints
Direct definitions of key terms and services1.5
Concise answers near the top of relevant sections1.5
Factual clarity and reduced ambiguity1.5
Logical heading structure1
Question-and-answer patterns where appropriate1
Clear service explanations1
Coverage of realistic user intent1
Content structured for extraction (short paragraphs, lists)0.5
Citation-ready factual statements1
Category total10

Evidence sources

  • • Page-level content analysis against the criteria above
  • • Manual review of a representative sample of pages

Strong implementation

A service page opens with a one-sentence definition, followed by a structured breakdown of what is included.

Weak implementation

A service page opens with a story about the company before ever stating what the service is.

Common failures
Answers buried in long, unstructured paragraphs; marketing language substituted for factual statements; no defined terms for technical concepts used repeatedly.
Zero-score condition
The domain has no indexable text content, or all content is inaccessible without interaction.
Limitations
Extractability can be assessed against known AEO/GEO patterns, but no methodology can guarantee a specific model will choose to extract a given passage.

Service, Industry, and Geographic Relevance

8 points

Recommendation systems favor sources that clearly match the intent and location of a query. Vague or geography-agnostic content is harder to match to local, industry-specific queries.

Point allocation

CheckPoints
Service specificity1.5
UAE, Dubai, and GCC relevance where applicable1
Industry relevance and terminology1
Explicit location-service relationships1
Sector-specific content depth1
Geographic evidence (addresses, service-area statements)1
Local business information completeness1
Local terminology usage0.5
Category total8

Evidence sources

  • • Content review against declared service and geographic scope
  • • LocalBusiness or Service schema review

Strong implementation

A service page explicitly names the cities and sectors it serves, with content specific to each.

Weak implementation

The same generic paragraph is republished across dozens of city or service permutations with no unique detail.

Common failures
Generic, location-agnostic service pages reused across unrelated markets; no distinction between service areas.
Zero-score condition
No service or geographic scope is stated anywhere on the domain.
Limitations
Relevance is evaluated against the organization's own declared scope, not against a universal industry standard.

Content Authority and Expertise

10 points

Content backed by identifiable expertise and editorial discipline is more defensible than anonymous, undated marketing copy.

Point allocation

CheckPoints
Named authors1.5
Demonstrated subject-matter expertise1.5
Publication and review dates present1
Reviewer identity where review has occurred1
Citations and references within content1
Editorial standards (consistency, correction process)1
Content depth relative to the topic1
Practical usefulness to the reader1
Originality and consistency of claims across pages1
Category total10

Evidence sources

  • • Author and reviewer bylines
  • • Publication and modification dates
  • • Citation review

Strong implementation

An article names its author, states a publication and last-reviewed date, and cites specific sources for factual claims.

Weak implementation

An anonymous article makes broad claims about results with no dates, sources, or named expertise behind it.

Common failures
No author attribution; identical claims made with no supporting detail; no visible publication or review date.
Zero-score condition
All content is unattributed, undated, and unreviewed.
Limitations
Expertise is assessed from what is publicly disclosed and verifiable, not from private credentials that cannot be confirmed.

Evidence Quality and Case-Study Strength

10 points

Case studies are only as credible as the evidence behind them. Unverifiable outcome claims reduce trust rather than build it.

Point allocation

CheckPoints
Verifiable project details1.5
Metrics with defined units and timeframes1.5
Named clients where permission exists1
Before-and-after evidence1
Dates attached to claims1
Client confirmation where available1
Disclosed methodology behind reported results1
Traceable sources for every figure cited1
Disclosed limitations of the case study0.5
Screenshots or artifacts where appropriate0.5
Category total10

Evidence sources

  • • Case-study content review
  • • Cross-reference of claimed clients or metrics against public sources where possible

Strong implementation

A case study names the client (with permission), states the engagement dates, and reports a specific, dated metric with its baseline.

Weak implementation

A case study claims dramatic growth with no client name, date, baseline, or method disclosed.

Common failures
Round, unsourced numbers ('300% growth') with no timeframe or baseline; anonymized case studies with no verifiable detail at all.
Zero-score condition
No case studies, portfolio evidence, or verifiable outcome claims exist anywhere on the domain.
Limitations
Vista by Lara can confirm evidence it directly holds. Independent verification of a client's own internal numbers is not always possible and is disclosed as such where relevant.

Independent Authority and Citation Signals

10 points

Independent, third-party validation is harder to fabricate than first-party claims, and both search and AI systems weight it accordingly.

Point allocation

CheckPoints
Third-party media mentions1.5
Earned links from independent sources1.5
Industry citations1
Independent articles referencing the organization1
Interviews and expert quotations attributed by name1.5
External professional profiles1
Partner acknowledgements and client-website references1
Conference or speaking profiles0.5
Directory presence1
Category total10

Evidence sources

  • • Direct verification of cited third-party sources
  • • Link and mention review

Strong implementation

A named journalist or independent publication references the organization, with a live, dated link.

Weak implementation

A 'featured in' badge with no working link to the claimed publication.

Common failures
Citing directories or press-release syndication as if they were independent editorial coverage; broken or unverifiable citation links.
Zero-score condition
No independently verifiable third-party mention of the organization exists.
Limitations
Only sources that are publicly accessible and verifiable at the time of evaluation are credited; sources may change or be removed later.

Reviews, Reputation, and Trust Signals

7 points

Reviews and transparent business information are a primary trust signal for both consumers and AI systems summarizing a business's reputation.

Point allocation

CheckPoints
Verified review volume1
Rating quality1
Business verification status on relevant platforms1
Transparency of address and contact information1
Review recency0.5
Diversity of review platforms0.5
Quality of responses to reviews0.5
Presence of policies and legal pages0.5
Team visibility0.5
A visible process for corrections or disputes0.5
Category total7

Evidence sources

  • • Direct review-platform verification
  • • Policy and legal-page review

Strong implementation

A verified profile with a meaningful review volume, recent activity, and visible owner responses.

Weak implementation

A handful of undated, unattributed testimonials embedded only as page copy.

Common failures
Reviews that cannot be traced to a real platform; no visible way to contact the business or resolve a dispute.
Zero-score condition
No verifiable reviews exist on any platform and no business-verification signals are present.
Limitations
Review platforms can remove or hide reviews outside Vista by Lara's control; this category reflects a point-in-time snapshot.

AI Accessibility and Public Knowledge Resources

5 points

Beyond traditional crawlability, AI systems benefit from clean, purpose-built resources that make a site's structure and offerings explicit.

Point allocation

CheckPoints
Confirmed AI crawler access (not merely assumed)1
Public knowledge endpoints (e.g. an /ai-data resource)1
Content accessible to AI crawlers specifically0.75
Clean content delivery without interstitials0.5
Endpoint availability0.5
Data freshness on public endpoints0.5
llms.txt where appropriate to the site's scale0.5
Clear separation of published versus draft resources0.25
Category total5

Evidence sources

  • • AI crawler user-agent access testing
  • • Public endpoint availability checks

Strong implementation

A documented, versioned public data endpoint exists, is reachable by named AI crawlers, and contains only published, reviewed data.

Weak implementation

robots.txt blocks all non-Google crawlers with no stated policy, and no machine-readable summary exists.

Common failures
Blocking AI crawlers by default without a documented policy reason; publishing draft or placeholder data on a public endpoint.
Zero-score condition
All AI crawler user agents are blocked and no public knowledge endpoint exists.
Limitations
Confirming that a crawler accessed a page requires server logs; access is not the same as inclusion in a model's output, which Vista by Lara cannot control or guarantee.

Multilingual and Regional Readiness

4 points

For a UAE and GCC audience, Arabic-language readiness is a distinct and material readiness factor, not a cosmetic translation layer.

Point allocation

CheckPoints
Correct hreflang implementation (resolves directly, no redirects)1
Arabic and English content parity for key pages1
Arabic-native terminology rather than transliteration alone0.5
Culturally appropriate content, not machine-translated copy0.5
Language-specific metadata0.5
Consistent multilingual navigation0.5
Category total4

Evidence sources

  • • hreflang and canonical review across language versions
  • • Native-language content review

Strong implementation

Every Arabic page's hreflang alternate resolves directly to a live, correctly matched English page and vice versa.

Weak implementation

Arabic pages exist but their hreflang alternates point to URLs that redirect or 404.

Common failures
hreflang tags that point to a redirecting or mismatched URL; Arabic pages that are thin machine translations of the English original.
Zero-score condition
Not applicable for organizations that do not target Arabic-language markets; scored only where multilingual delivery is in scope.
Limitations
This category is only scored for organizations where multilingual delivery is part of the declared scope of evaluation.

Freshness, Monitoring, and Governance

4 points

A methodology and the content it evaluates both decay without maintenance. Visible governance is itself a trust signal, and its absence is a real limitation.

Point allocation

CheckPoints
Last-reviewed dates on key content0.5
Version control on the methodology itself0.5
Named governance ownership0.5
A documented correction policy0.5
Broken-link monitoring0.5
Schema monitoring0.5
Defined content review cycles0.5
Citation and prompt monitoring where in scope0.5
Category total4

Evidence sources

  • • Version history review
  • • Last-reviewed date audit across sampled pages

Strong implementation

A visible version history documents every material change to scoring logic, with dates and a named owner.

Weak implementation

A score changes between two reports with no explanation and no version change.

Common failures
Content with no visible last-reviewed date; no named methodology owner; silent changes to scoring logic with no version bump.
Zero-score condition
No dates, ownership, or version history are disclosed anywhere on the domain.
Limitations
This category evaluates whether governance is visible and documented, not the internal quality of processes that cannot be externally observed.

Score interpretation scale

0-20

Critically Underdeveloped

The digital presence lacks basic clarity, accessibility, structure, evidence, or reliable entity identification.

21-40

Limited Readiness

Some foundations exist, but major technical, content, evidence, or authority gaps prevent reliable interpretation.

41-60

Developing Readiness

The organization has partial structure and useful content, but implementation remains inconsistent or insufficiently validated.

61-80

Strong Readiness

The organization is generally understandable, accessible, and well structured, with meaningful evidence and authority signals.

81-90

Advanced Readiness

The organization demonstrates mature technical architecture, strong entity clarity, evidence quality, and broad recommendation readiness.

91-100

Exceptional Readiness

The organization demonstrates highly developed technical, entity, content, evidence, governance, and authority architecture.

A high readiness score does not mean that every AI platform will recommend the organization for every query.

Score calculation

Every category has a maximum point value equal to its weight. Each category contains multiple checks, which may be binary, graded, quantitative, or manually reviewed. Category totals are normalized to the assigned weight, and the final score is the sum of all weighted category scores.

Final Score = Sum of all weighted category scores

The final score is rounded to the nearest whole number using standard rounding (0.5 rounds up). Missing evidence does not receive assumed credit. Unverifiable claims do not receive full credit. Draft and placeholder resources do not receive published-resource credit.

Unavailable data

Scored as a failed check for that criterion; not assumed to be correct.

Inaccessible pages

Retested once; if still inaccessible, scored as a failed check under Technical Accessibility.

Conflicting evidence

Resolved using the evidence hierarchy (higher tier wins) and documented in the evaluation record.

Duplicate evidence

Counted once. Repetition of the same evidence across pages does not multiply credit.

Stale evidence

Evidence older than the category's expected freshness window is flagged and weighted down, not excluded automatically.

Temporary technical failures

Retested before scoring; persistent failures across repeated checks are scored as findings.

Non-applicable checks

Marked not applicable and excluded from that category's applicable total, rather than awarded automatic full credit.

Evaluation process

Every evaluation follows the same twenty-step pipeline, in order.

  1. 1

    Scope confirmation

    Manual
    Input
    Domain, business context, declared markets
    Method
    Direct confirmation with the organization
    Output
    Defined evaluation scope

    Common failure: Scope not confirmed before evaluation begins

  2. 2

    Domain and route discovery

    Automated
    Input
    Root domain
    Method
    Sitemap and crawl-based route discovery
    Output
    Full route inventory

    Common failure: Orphaned pages missing from sitemap

  3. 3

    Crawl and rendering analysis

    Automated
    Input
    Route inventory
    Method
    Fetch and render comparison
    Output
    Rendered content per route

    Common failure: Content only available after client-side rendering

  4. 4

    Robots and sitemap inspection

    Automated
    Input
    robots.txt, sitemap.xml
    Method
    Direct fetch and rule parsing
    Output
    Crawler access policy summary

    Common failure: Contradictory or overly broad disallow rules

  5. 5

    Entity extraction

    Hybrid
    Input
    Rendered content, schema
    Method
    Structured-data and text-based entity extraction
    Output
    Entity map

    Common failure: Entity referenced with inconsistent names

  6. 6

    Structured-data validation

    Automated
    Input
    JSON-LD blocks
    Method
    Schema.org validation
    Output
    Validity and completeness report

    Common failure: Invalid or incomplete required properties

  7. 7

    Content and answer analysis

    Hybrid
    Input
    Rendered text content
    Method
    Manual and pattern-based review against AEO/GEO criteria
    Output
    Content clarity assessment

    Common failure: No direct answers near the top of key pages

  8. 8

    Internal-link analysis

    Automated
    Input
    Full route inventory
    Method
    Link extraction and status-code verification
    Output
    Redirect and broken-link report

    Common failure: Internal links routed through redirects

  9. 9

    Service and geographic mapping

    Manual
    Input
    Declared scope, content
    Method
    Manual review against declared markets
    Output
    Relevance assessment

    Common failure: Generic content reused across unrelated markets

  10. 10

    Evidence review

    Manual
    Input
    Case studies, claims
    Method
    Evidence-tier classification
    Output
    Evidence quality assessment

    Common failure: Claims with no traceable source

  11. 11

    Review and reputation analysis

    Hybrid
    Input
    Review platforms
    Method
    Direct platform verification
    Output
    Reputation summary

    Common failure: Reviews cited but not independently verifiable

  12. 12

    External authority analysis

    Manual
    Input
    Cited third-party sources
    Method
    Direct source verification
    Output
    Independent authority summary

    Common failure: Broken or unverifiable citation links

  13. 13

    AI accessibility review

    Automated
    Input
    robots rules, public endpoints
    Method
    Crawler user-agent testing, endpoint checks
    Output
    AI accessibility summary

    Common failure: AI crawlers blocked without a documented policy reason

  14. 14

    Multilingual review

    Hybrid
    Input
    Language versions
    Method
    hreflang and content-parity review
    Output
    Multilingual readiness assessment

    Common failure: hreflang alternates pointing to redirecting URLs

  15. 15

    Prompt testing (where included)

    Manual
    Input
    Defined prompt set
    Method
    Structured prompt logging across platforms
    Output
    Recommendation Visibility Module data

    Common failure: Single-run results treated as deterministic

  16. 16

    Human validation

    Manual
    Input
    All automated findings
    Method
    Reviewer verification of automated output
    Output
    Validated findings set

    Common failure: Automated false positives not caught before scoring

  17. 17

    Conflict resolution

    Manual
    Input
    Conflicting evidence or findings
    Method
    Documented resolution against evidence hierarchy
    Output
    Resolved findings

    Common failure: Conflicting evidence resolved inconsistently across reports

  18. 18

    Score calculation

    Automated
    Input
    Validated, resolved findings
    Method
    Weighted category scoring per this methodology
    Output
    Final score and category breakdown

    Common failure: Manual overrides applied without documentation

  19. 19

    Quality assurance

    Manual
    Input
    Draft report
    Method
    Second-pass review against this methodology
    Output
    QA-approved report

    Common failure: QA skipped under time pressure

  20. 20

    Report publication

    Manual
    Input
    QA-approved report
    Method
    Versioned publication with methodology reference
    Output
    Published report referencing this methodology version

    Common failure: Report published without the methodology version it was scored against

Automation versus human review

Automation detects patterns and implementation signals. Human review evaluates meaning, credibility, context, and evidence quality.

Automated checks

  • • Status codes
  • • Schema validity
  • • Metadata presence
  • • Sitemap discovery
  • • Internal-link status codes
  • • Heading structure
  • • Canonical tags
  • • Response time
  • • Public endpoint availability
  • • Date-freshness checks

Human-reviewed checks

  • • Case-study credibility
  • • Claim accuracy
  • • Entity ambiguity
  • • Author expertise
  • • Evidence strength
  • • Recommendation relevance
  • • Geographic appropriateness
  • • Testimonial context
  • • Conflict resolution
  • • False-positive review

Reviewer responsibilities: verifying automated findings, classifying evidence tier, checking claims against source material, and documenting judgment calls.

Disagreement resolution: reviewer disagreements are resolved against the documented evidence hierarchy and category criteria, not by majority vote, and the resolution is recorded.

Second reviewer: enterprise and high-stakes audits require a second reviewer to independently verify the evidence-quality and external-authority categories before a report is published.

Evidence standards

Every piece of evidence used in scoring is classified into one of four tiers. Lower tiers receive progressively less credit toward evidence-dependent categories.

1

Strong Independent Evidence

Verifiable by a party with no commercial interest in the outcome.

  • • Official government records
  • • Official business registries
  • • Official award-issuer pages
  • • Client-owned confirmations
  • • Verified platform reviews
  • • Recognized media coverage
  • • Independent research
  • • Official partner pages
  • • Analytics exports with dates
  • • Search Console evidence
2

Strong First-Party Evidence

Produced by Vista by Lara or the client, disclosed transparently, and internally consistent.

  • • Documented case studies
  • • Signed testimonials
  • • Published methodology
  • • Dated screenshots
  • • Public API records
  • • Published author profiles
  • • Technical documentation
  • • Version history
3

Supporting Evidence

Useful context that strengthens a claim but is not sufficient on its own.

  • • Directories
  • • Social profiles
  • • Presentation materials
  • • Event listings
  • • Partner announcements
  • • Structured citations
4

Weak or Unverified Claims

Receives little or no authority credit until upgraded with real evidence.

  • • Unsupported superlatives
  • • Unattributed ratings
  • • Anonymous testimonials
  • • Unverified awards
  • • Undated screenshots
  • • Unverifiable statistics
  • • Self-assigned leadership claims

Tier 4 evidence receives little or no authority credit.

Claim validation rules

Superlative and authority claims are held to explicit rules before they receive any evidence credit. This applies to terms including:

  • Best
  • Leading
  • Number one
  • Award-winning
  • Most recommended
  • Trusted
  • Certified
  • Official partner
  • Industry leader
  • Highest rated
  • • Claims must have a source.
  • • Sources must be publicly accessible where possible.
  • • Dates must be included.
  • • Scope and geography must be defined.
  • • Comparison methodology must be defined.
  • • Expired recognition must not be presented as current.
  • • Internal scores must not be presented as independent awards.
  • • Ratings must include platform and review count.
  • • Partner claims must link to official confirmation.

Benchmarking

Businesses are benchmarked against relevant sector peers, not against unrelated industries -- healthcare is not directly compared with restaurants, and local service businesses are not benchmarked identically to enterprise software companies.

Regulated industries require stronger evidence and compliance-aware review. Businesses with multiple locations require location-specific analysis. Arabic and English visibility may be scored separately. Benchmark datasets, where used, are date-stamped and their sample size is disclosed.

Weighting model for v1.0: the twelve category weights are fixed across every industry. Sector differences affect which checks apply, evidence expectations, benchmark interpretation, and diagnostic recommendations -- they never change the published category weights. A future version-controlled release could introduce industry-specific weighting tables; until then, a single weighting applies to every evaluation.

Local service business

Evaluated against local-market clarity and geographic evidence.

Medical and healthcare

Requires stronger evidence and compliance-aware review.

Education

Evaluated with attention to accreditation and outcome transparency.

Real estate

Location-specific analysis required for multi-listing organizations.

Retail and e-commerce

Applies stricter evidence expectations to product-level structured data while retaining the published category weights.

Hospitality

Applies stricter evidence expectations to review-platform signals while retaining the published category weights.

Professional services

Applies stricter evidence expectations to expertise and evidence-quality signals while retaining the published category weights.

Enterprise B2B

Applies stricter evidence expectations to case-study and external-authority signals while retaining the published category weights.

SaaS and technology

Applies stricter evidence expectations to technical documentation and structured-data depth while retaining the published category weights.

Multi-location organizations

Each location is evaluated individually before an aggregate view is produced.

Vista by Lara does not publish invented industry-average benchmarks; only date-stamped, disclosed benchmark data is used.

Recommendation Visibility Module (prompt testing)

Prompt testing is separate from the core AI Authority Score readiness score unless explicitly commissioned and disclosed as an included module. When included, it is documented, per prompt, as:

  • Platform
  • Model name where available
  • Date
  • Geography
  • Language
  • Logged-in or clean-session status
  • Prompt text
  • Result presence
  • Citation presence
  • Recommendation order
  • Competitor mentions
  • Response consistency across repeats
  • Source URLs
  • Repeated test count

Generative outputs are variable. Vista by Lara does not claim deterministic results from prompt testing, and where this module affects a published score, its exact weight and calculation is disclosed alongside the report.

Limitations

This section is deliberately prominent and must be read alongside any score.

  • • AI systems change over time.
  • • Outputs vary across models and sessions.
  • • Results vary by location and language.
  • • Personalization can affect responses.
  • • Crawling does not guarantee inclusion.
  • • Structured data does not guarantee citation.
  • • High authority does not guarantee recommendation.
  • • External sources may change after evaluation.
  • • Reviews may be removed by the platform hosting them.
  • • Rankings fluctuate independent of this methodology.
  • • Platform documentation may change without notice.
  • • Data may be incomplete at the time of evaluation.
  • • Some evidence cannot be independently verified.
  • • Technical errors may temporarily affect results.
  • • Proprietary scoring involves human judgment.
  • • Human reviewers may disagree; disagreements are resolved and documented.
  • • Scores are time-bound snapshots, not continuous measurements.
  • • Commercial performance requires separate measurement and is not implied by this score.

The AI Authority Score is an assessment tool, not a guarantee, certification, endorsement, ranking factor, legal opinion, financial projection, or promise of commercial performance.

Version history

VersionRelease dateStatusSummary
v1.02026-07-13ActiveInitial published version of the AI Authority Score methodology: twelve weighted categories, evidence tiers, scoring scale, evaluation pipeline, limitations, and governance.
Next review date
2026-10-13
Methodology owner
Vista by Lara
Technical owner
Vista by Lara Engineering

Any material change to category weights, scoring criteria, evidence requirements, normalization, rounding, human review, or benchmark logic creates a new methodology version. Historical scores are never silently changed; historical reports retain the methodology version in effect at the time of evaluation.

Corrections and appeals

To report a factual error, request an evidence review, challenge a score, submit missing evidence, correct outdated information, or escalate a dispute, contact Vista by Lara directly.

Contact method

Email solution@vistabylara.com or use the contact page, referencing the report date and methodology version.

Required information

The specific claim or score in question, the methodology version it was scored under, and any supporting evidence for the correction.

Review process and decision authority

Reviewed by Vista by Lara Engineering against this methodology. Disputed evidence is re-classified using the evidence hierarchy in Evidence Standards.

Audit trail

Corrections are dated and noted against the methodology version in effect; they do not silently alter past published reports.

A fixed response time is not promised here to avoid committing to a service level this page cannot guarantee; response timing is confirmed directly when a request is submitted.

Governance

Methodology owner
Vista by Lara
Technical owner
Vista by Lara Engineering
Editorial owner
Vista by Lara Engineering
Review frequency
Quarterly

Publication process: a draft passes quality assurance (evaluation pipeline step 19) before publication, and any change to scoring logic triggers a new methodology version.

Evidence retention and confidentiality: client-private evidence may be reviewed as part of an evaluation without being publicly exposed. Confidential evidence is never included in public structured data or on this page.

Data minimization: only evidence necessary to support a scoring decision is retained for that purpose.

Score publication policy: published scores reference the exact methodology version they were calculated under.

Frequently asked questions

What is the AI Authority Score?

It is Vista by Lara's proprietary framework for evaluating how clearly, reliably, and credibly an organization's digital presence can be discovered, interpreted, verified, and considered by AI-powered search and recommendation systems.

Is this an official Google score?

No. Google has no official AI Authority Score, and this framework is not created, endorsed, or used by Google.

Does OpenAI use this score?

No. OpenAI does not use, endorse, or recognize the AI Authority Score. It is independently developed by Vista by Lara.

Does a high score guarantee ChatGPT, Gemini, or other AI recommendations?

No. A high score reflects strong readiness and evidence. It does not guarantee that any specific AI system will recommend the business for any specific prompt.

How is the score calculated?

Twelve weighted categories, each scored against documented criteria and evidence standards, are summed to a final score out of 100. The full weighting is published in the Scoring Model section.

Why is human review required?

Automation can detect implementation signals such as valid schema or working links. It cannot judge whether a claim is credible, whether evidence is strong, or whether content is genuinely useful. Those judgments require a human reviewer.

How often is the score updated?

Per the review frequency documented in Version Control. Reviews may also be triggered by material changes to a business's digital presence.

Can the score decrease?

Yes. If technical issues, broken evidence, removed reviews, or content regressions are found on re-evaluation, the score can decrease.

What evidence is accepted?

Evidence is classified into four tiers, from independently verifiable records to weak or unverified claims. See Evidence Standards.

Are reviews included in the score?

Yes, within the Reviews, Reputation, and Trust Signals category, weighted at 7 of 100 points.

Are backlinks included?

Yes, as one signal within the Independent Authority and Citation Signals category, alongside media mentions, directories, and other independent references.

Is structured data alone enough to achieve a high score?

No. Structured Data and Machine Readability is one of twelve categories, worth 10 of 100 points. A high overall score requires strength across most categories.

Does publishing an llms.txt file increase the score?

It can contribute within the AI Accessibility and Public Knowledge Resources category, but it is one of several signals in a 5-point category, not a scoring shortcut on its own.

Does confirming AI crawler access guarantee visibility in AI answers?

No. Crawler access means a system can retrieve the content. It does not mean any model will choose to cite or recommend it.

How are case studies evaluated?

Against the Evidence Quality and Case-Study Strength category: named clients where permission exists, verifiable details, dates, metrics with defined baselines, and disclosed methodology and limitations.

How are unsupported claims treated?

Claims without a traceable source are classified as Tier 4 evidence and receive little to no authority credit. See Claim Validation Rules.

How are Arabic and English content assessed?

Under Multilingual and Regional Readiness, evaluating content parity, correct hreflang implementation, and native-quality Arabic content rather than machine translation.

How are different industries compared?

Organizations are benchmarked within their own sector class, not against unrelated industries. See Benchmarking.

What happens when data is missing?

Missing evidence does not receive assumed credit. Checks that cannot be evaluated are either marked not applicable and excluded from that category's applicable total, or scored as zero for that check, per the documented calculation rule.

Can a business or client challenge a score?

Yes. See Corrections and Appeals for the process to report an error, submit missing evidence, or request a review.

Are competitors scored using the same framework?

When a comparative or competitive analysis is explicitly commissioned, the same published methodology and version is applied consistently across all organizations being compared.

Does the score measure leads or revenue?

No. Business Impact metrics such as traffic, calls, consultations, and revenue are tracked separately from the AI Authority Score and are never blended into it without clear labeling.

What is the difference between readiness and visibility?

Readiness (the AI Authority Score) measures whether a business can be clearly discovered and understood. Visibility measures whether it is actually being surfaced or cited in real AI outputs, which is tracked separately as a Recommendation Visibility Module where included.

What is the difference between authority and recommendation frequency?

Authority reflects the strength of evidence and independent validation behind a business. Recommendation frequency reflects how often that business is actually mentioned or cited by AI systems in monitored prompts. A business can have strong authority and still be recommended rarely, particularly in a competitive category.

Why do AI answers change between sessions?

Generative AI outputs are inherently variable across models, sessions, personalization, and time. This methodology documents that variability rather than claiming deterministic results.

How are methodology changes handled?

Any material change to category weights, criteria, evidence standards, normalization, rounding, human-review requirements, or benchmark logic creates a new methodology version. Historical reports retain the version they were scored under. See Version Control.

Can a business receive a perfect score?

A score of 100 is mathematically possible. It would still not guarantee that an organization is recommended by every AI platform or for every relevant query.

Are private client documents included in the public methodology?

No. Confidential client evidence may be reviewed as part of an evaluation without being exposed in public structured data or on this page.

How are temporary technical errors treated?

A single transient failure is retested before being scored. Persistent failures across repeated checks are scored as findings.

How can a business improve its score?

By addressing the specific criteria, common failures, and evidence gaps documented in each category above, in priority order of category weight and current gap size.

References

This methodology is built on the following official standards and documentation.

Authorship and review

Author

Lara Eros Farbactian

Founder & Principal Architect at Vista by Lara

Technical review

Technical review: Vista by Lara Engineering

Published 2026-07-13 · Last reviewed 2026-07-13

AI Authority Score™ is a proprietary assessment framework developed by Vista by Lara.

See this methodology applied to your business

Request an AI Visibility AuditExplore the AI Authority EngineContact Vista by Lara