The VerifiedLayer Trust Score is a structured, automated, evidence-based assessment of any software vendor operating in the B2B technology market. It is produced by running 53 structured prompts across 3 major AI platforms, combining community sentiment signals, and verifying product evidence, all without requiring vendor participation or submission.
The methodology is designed to be transparent, reproducible, and useful to software buyers making real purchasing decisions. Every component of the score is documented on this page. Vendors who understand the methodology can identify exactly what public evidence would improve their score; that is the intended behavior.
VTSS measures how leading AI models perceive software vendors based on publicly available evidence. It does not certify legal compliance, financial health, or factual truth.
DOCUMENT REFERENCE
VL-VTSS-V2 · Published Methodology
TOTAL EVALUATIONS
159 / vendor / month (53 prompts × 3 platforms)
SCORE RANGE
0–100, each axis
AXES PUBLISHED
Quality & Presence: never blended
UPDATE CYCLE
Monthly automated pipeline
HUMAN INVOLVEMENT
None in scoring
•Evidence precedes confidence, and confidence precedes score: never the reverse, on any single query.
•AI does not invent ratings. Every finding a model claims is checked against a predefined Evidence Rating Dictionary wherever a matching entry exists.
•Three independent model families evaluate every vendor. Agreement is measured, not assumed.
•Quality (how good) and Presence (how big/visible) are never blended into a single number.
•Every vendor is evaluated on the same monthly cycle, with the same prompts and formulas.
•Every published score is explainable and traceable back to the evidence that produced it.
Every vendor is evaluated across three independent signal-source layers. Weights are never fixed globally: they vary per dimension (Section 8).
AI Platform Signal
Prompts probing what leading AI models know from training and public sources
40 prompts × 3 platforms
Community Voice Signal
Prompts on Reddit, LinkedIn, Glassdoor, G2, TrustRadius sentiment
8 prompts × 3 platforms
Product Evidence Signal
Prompts verifying certifications, analyst recognition, customer references, financial stability
5 prompts × 3 platforms
Total53 prompts × 3 platforms = 159 evaluations / vendor / month
Three independent platforms
Three separate foundation-model companies, each with an independent training pipeline, so agreement between them means something real.
Two axes, never blended
Every piece of evidence is tagged as measuring Quality, Presence, or Both, and the two resulting scores are calculated and published separately.
A third, supplementary axis
Momentum tracks quarter-over-quarter improvement, independent of absolute size.
The fundamental problem with asking an AI model to "score this vendor 0–100" is that the model tends to assign a score first and generate justification for it afterward, producing overconfident, inconsistent results. VerifiedLayer inverts this order on every single query, without exception:
Collect Evidence
Collect evidence from public sources before considering any score. Sources are named and findings are quoted specifically. If fewer than 2 independent sources are found, the score for that query cannot exceed 45, regardless of what the evidence otherwise suggests.
Declare Confidence
Declare confidence (HIGH, MEDIUM, or LOW) based on what was found, before writing any score. HIGH requires 3+ independent sources, recent (within 24 months), of different types. MEDIUM requires 2 independent sources, or older sources. LOW means only 1 source, or indirect or contradictory evidence.
Assign the Score
Assign the score, constrained by the evidence and confidence already declared. A LOW-confidence query is still scored and published, but flagged.
EVERY QUERY RETURNS A CONSISTENT SET OF FIELDS
EVIDENCE_FOUNDsources named, findings quoted specifically
EVIDENCE_GAPSwhat was searched for but not found
CONFIDENCEHIGH / MEDIUM / LOW + reasoning
SCOREinteger 0–100 + reasoning
AXISQuality / Presence / Both
APPROVED PUBLIC SOURCES
Vendor sites & product documentationCertification bodies & regulatory registriesReview platformsGitHubNews publicationsCompany blogsPartner ecosystems
Vendor marketing copy is never treated as independent evidence.
Confidence is not a courtesy label; it actively constrains what a score is allowed to be:
| Confidence | Requirement |
|---|
| HIGH | 3+ independent sources, recent (within 24 months), of different types |
| MEDIUM | 2 independent sources, or sources older than 12 months |
| LOW | Only 1 source, indirect evidence, or contradictions found |
A vendor with fewer than 2 independent public sources on a given question cannot score above 45 on that question, no matter how positive the single available source is. This prevents a single self-reported claim from carrying the same weight as corroborated public evidence.
Common findings map to a predefined rating so that the same evidence is treated the same way for every vendor, every month:
| Finding | Rating |
|---|
| SOC 2 Type II Certified (verified via audit registry) | +95 |
| SOC 2 Type I only, or self-attested | +10 (maximum) |
| ISO 27001 Certified | +90 |
| Named Gartner MQ / Forrester Wave position, with year and report cited | +85 |
| Generic "recognized by Gartner," no specifics | 0 |
| GDPR Compliance documented | +40 |
| Named enterprise reference with quantified outcome | +30 |
| Anonymous case study / generic testimonial | 0 |
| Multi-Factor Authentication available | +20 |
| Public API documentation | +15 |
| Pricing fully published with tiers and inclusions | +15 |
| Contact-sales-only pricing | 0 (neutral) |
| "AI-powered" claimed in marketing with no technical substance | capped at 40 |
| Active Critical Security Incident (unresolved) | −50 |
| Active Regulatory Action | −70 |
| No Glassdoor presence | 45 (unknown, not treated as bad) |
A model does not decide what a certification or an incident is "worth" in the moment; it recognizes the finding, and this table supplies the anchor. This is also what makes a vendor's score improvable in a predictable way: an organization that genuinely earns SOC 2 Type II, publishes transparent pricing, or documents a named enterprise outcome knows in advance what that will do to its score.
Each of the 3 platforms answers every query independently. Their scores are combined using a reliability-weighted method, so agreement is rewarded with a simple average and disagreement is handled conservatively rather than blindly averaged away.
| Platform Agreement | Aggregation Method Used |
|---|
| High agreement | Simple average across all 3 platforms |
| Good agreement | Median (the middle score of the 3) |
| Moderate agreement | Trimmed mean (the one outlier score is dropped, remaining 2 averaged) |
| Low agreement | Conservative floor (lowest of the 3 scores used), and the query is flagged for review |
This produces one combined score per query, which feeds into its dimension along the axis (Quality, Presence, or Both) that evidence was tagged with.
Every vendor is scored across 8 fixed dimensions. Weights are derived from research on what actually drives enterprise purchase decisions, and sum to exactly 100%.
Feature completeness, stability, real-world effectiveness.
ANTI-GAMINGRequires actively investigating weaknesses; scores above 70 require documented negative-source review.
Analyst, media, and peer regard.
ANTI-GAMINGRequires a specific report name, year, and position; generic mentions are rejected.
What real practitioners say on Reddit and forums.
ANTI-GAMINGAbsence of signal scores 40, not 100; silence cannot be gamed into a good score.
Onboarding, support, post-sale reality vs. promise.
ANTI-GAMINGOnly user-reported timelines count, not vendor-stated ones.
Certifications, API quality, compliance.
ANTI-GAMINGSOC 2 Type II earns full credit; Type I is capped; self-attestation earns nothing.
Pricing transparency, documented ROI, true total cost of ownership.
ANTI-GAMINGMarketing ROI claims with no client attribution score 0.
Genuine innovation vs. marketing claims.
ANTI-GAMING"AI-powered" with no technical substance caps the score at 40.
Financial stability, lock-in risk, legal history.
ANTI-GAMINGLock-in difficulty must be documented by users who tried to exit, not by the vendor.
SIGNAL-SOURCE WEIGHTS PER DIMENSION
| Dimension | AI Signal | Community Signal | Evidence Signal | Why |
|---|
| Product Quality | 40% | 40% | 20% | Equally about perception and community voice |
| Market Reputation | 35% | 35% | 30% | Blends AI knowledge, community, and verifiable analyst position |
| Community Sentiment | 20% | 70% | 10% | Community is the primary signal |
| Customer Experience | 35% | 45% | 20% | Post-sale reality is best captured in community voice |
| Technical Credibility | 30% | 30% | 40% | Certifications are verifiable facts: evidence dominates |
| Value for Money | 35% | 35% | 30% | Blends perception, community feedback, and pricing evidence |
| Innovation & Vision | 40% | 30% | 30% | Best assessed through AI awareness and analyst citation |
| Buying Safety | 30% | 30% | 40% | Depends on verifiable facts: funding, legal, certifications |
Every finding is tagged as measuring Quality, Presence, or Both. A dimension therefore produces two sub-scores (a Quality-axis score and a Presence-axis score) built only from the evidence tagged to that axis. Evidence tagged Both contributes at half weight to each axis, so it is never counted at full strength twice.
FORMULAS
Quality Score = Σ (Dimension Weight × Dimension Quality-axis Score) across all 8 dimensions
Presence Score = Σ (Dimension Weight × Dimension Presence-axis Score) across all 8 dimensions
Both scores range from 0 to 100 and are published side by side, never combined into a single blended number. This is what prevents large, high-visibility vendors from automatically outranking better, smaller products: a vendor with excellent Quality and modest Presence is shown exactly as such, not averaged down.
This measures the reliability of the evidence base behind a vendor's score this month, not the vendor's quality.
| Factor | Description | Weight |
|---|
| AI Agreement | Average cross-model agreement across all queries | 40% |
| Evidence Coverage | Verified items collected vs. expected baseline for the category | 20% |
| Source Diversity | Independent source categories represented | 20% |
| Evidence Freshness | Average recency across all findings used | 20% |
Momentum rewards vendors improving quickly, regardless of where they started:
Momentum = 70% × (change in Quality Score over the last 2 quarters) + 30% × (growth in reviews, mentions, and references over the same period)
A vendor whose Momentum reaches +15 or more is shown with a "Rising" flag, regardless of how small it still is.
Four checks run automatically before any score is published. No human is involved in scoring (the pipeline always publishes, it never holds a score indefinitely) but low-confidence results are flagged directly on the public profile.
Cross-Model Reliability
Determines which aggregation method is used for each query (High / Good / Moderate / Low agreement).
Evidence Spot-Check
A sample of cited evidence is validated against a live web search. A high fabrication rate floors the affected score and flags it as low confidence; a moderate rate triggers a single re-run with a stricter check; a low rate passes cleanly.
Month-over-Month Change Gate
If a vendor's score changes sharply from the prior month, the published score is blended with the prior month's result rather than jumping immediately, and a "Significant movement detected" flag is shown. Prevents a single event (a viral post, a review bomb) from swinging a score overnight.
Confidence Gate
If a large share of a vendor's monthly queries return low confidence, the published score is capped and flagged "Limited Evidence" or "Partial Evidence." If the evidence base is too thin to trust at all, the leaderboard shows "Unranked: Insufficient Evidence" instead of a numeric score.
Badge: a grid of Quality × Presence, never a single blended score.
| Quality Score | Presence Score | Badge |
|---|
| ≥ 75 | ≥ 65 | |
| ≥ 65 | < 65 | |
| 50–74 | ≥ 65 | |
| 50–64 | < 65 | |
| < 50 | any | |
| (any of the above) | Momentum ≥ +15 | |
A vendor with high Quality but modest Presence earns High Performer, not a diluted mid-tier badge. A vendor with high Presence but mediocre Quality earns Established, not an inflated Leader badge.
CATEGORY RANK FORMULA
Rank Score = 70% × Confidence Score + 20% × Review Volume Score + 10% × Category Relevance
Category Rank orders vendors within a category page only; it does not affect a vendor's Quality, Presence, or badge.
CATEGORY ASSIGNMENT PRIORITY
| Priority | Source | Why |
|---|
| 1: Highest | G2 Category Page | Research-backed, reflects real buyer evaluation behavior. |
| 2: High | Gartner / Forrester Report | Independent analyst classification, names category directly. |
| 3: Medium | Technology Press | Journalists generally research before publishing. |
| 4: Lowest | Vendor's Own Website | Vendors self-describe aspirationally; used only if 1–3 yield nothing. |
Vendors matching no approved category are placed in "Emerging AI Software" and may correct their category assignment by claiming their profile.
Trigger the monthly evaluation for every vendor
Execute all 53 standardized prompts across all 3 platforms
Apply the evidence-first protocol to every query
Aggregate cross-model scores; compute dimension, Quality, and Presence scores
Compute Confidence and Momentum scores
Run all automated quality gates and apply any caps
Assign the badge and Category RankPublish the leaderboard, each month archived as an immutable historical snapshot VTSS measures AI-perceived, evidence-based reputation, not absolute truth. Rankings:
•Reflect only publicly available information as of the evaluation date.
•Depend on the current Evidence Rating Dictionary; ratings can be revised going forward but are never applied retroactively to past monthly snapshots.
•Are not legal, financial, security, or compliance certifications, and should never be treated as such.
•Can be temporarily capped or withheld when evidence is thin; this reflects the confidence of the evaluation, not a judgment on the vendor.
•Should be read as a structured, monthly AI-driven assessment, not a guarantee of vendor performance or a substitute for a buyer's own due diligence.
Query
One of the 53 standardized monthly prompts, asked of all 3 platforms.
Signal Source
The AI Platform, Community Voice, or Product Evidence category a query belongs to.
Evidence Rating Dictionary (ERD)
The fixed, versioned table mapping findings to predefined ratings.
Confidence (query-level)
The HIGH/MEDIUM/LOW label a model assigns a single query, based on evidence found.
Confidence Score
The separate, aggregate measure of how reliable the whole month's evidence base was.
Axis (Quality / Presence / Both)
Which of the two never-blended scores a piece of evidence contributes to.
Quality Score / Presence Score
The two core published scores; never combined into one number.
Momentum Score
Rate of quarter-over-quarter improvement.
Category Rank Score
The separate formula used only to order vendors within a category page.
Badge
The label (Leader, High Performer, Established, Verified, Emerging/Caution, + Rising) from the Quality × Presence grid.
QUALITY-AXIS DIMENSION SCORES
| Dimension | Weight | Quality Score |
|---|
| Product Quality | 18% | 88.00 |
| Market Reputation | 16% | 82.00 |
| Community Sentiment | 15% | 79.00 |
| Customer Experience | 14% | 84.00 |
| Technical Credibility | 13% | 87.00 |
| Value for Money | 11% | 75.00 |
| Innovation & Vision | 8% | 80.00 |
| Buying Safety | 5% | 90.00 |
PRESENCE-AXIS DIMENSION SCORES
| Dimension | Weight | Presence Score |
|---|
| Product Quality | 18% | 70.00 |
| Market Reputation | 16% | 85.00 |
| Community Sentiment | 15% | 90.00 |
| Customer Experience | 14% | 60.00 |
| Technical Credibility | 13% | 65.00 |
| Value for Money | 11% | 55.00 |
| Innovation & Vision | 8% | 72.00 |
| Buying Safety | 5% | 68.00 |
Quality Score = 0.18(88)+0.16(82)+0.15(79)+0.14(84)+0.13(87)+0.11(75)+0.08(80)+0.05(90) = 83.03
Presence Score = 0.18(70)+0.16(85)+0.15(90)+0.14(60)+0.13(65)+0.11(55)+0.08(72)+0.05(68) = 71.76
CONFIDENCE SCORE
well above the evidence floor, published without a cap
MOMENTUM
below the +15 threshold, no "Rising" flag this month
BADGE: QUALITY 83.03 (≥75) AND PRESENCE 71.76 (≥65) →
Published profile: Quality 83.03 · Presence 71.76 · Confidence 88.60 · Momentum 11.72 · Badge: LEADER
Browse the leaderboard to see how vendors perform under VTSS v2.0, or run your own vendor audit.