Machine Understanding & Decision Intelligence

Measure what machines understand.
See what happens when they have to choose.

Vorentus measures the gap between what is verifiably true about your organisation and how AI systems understand, compare and recommend it.

Verified realityThe gapMachine understanding
  • C-01

    Category and positioning

    Recognised

    Consistent across environments

  • C-02

    Differentiating attribute

    Partial

    Varies by environment

  • C-03

    Certified capability

    Inconsistent

    Varies by environment

  • C-04

    Independent recognition

    Absent

    Not reconstructed

Geometry carries the meaning. Colour confirms the state.Illustrative. Displacement is schematic, not a measured result.
  • Verified reality
  • Recognised correctly
  • Partial or inconsistent
  • Absent or contradicted
The problem

Your organisation is not necessarily what machines understand.

Publishing accurate information does not guarantee accurate machine understanding. An AI system does not read an organisation — it reconstructs one, from signals it never asked permission to use.

One verified fact

Certified capability, in operation, documented.

Reconstructed from
  • Corporate content
  • Structured information
  • Third-party sources
  • Citations
  • Databases
  • Historical information
  • Retrieved information
  • Request context
The reconstruction may be
  • Accuratethe reconstruction matches verified reality
  • Incompletepart of what is true never arrives
  • Outdatedan earlier version of the organisation persists
  • Inconsistentthe same question returns different organisations
  • Wronga capability is attributed elsewhere, or denied

Appearing is not the same as being understood.

Being understood is not the same as being chosen.

Inside Vorentus

See the gap at claim level.

Vorentus connects verified organisational claims with documented observations across machine environments, preserving the context and evidence behind each result.

Illustrative product view · demonstration data
Executive · Decision contextDemonstration organisation
CTX-0416
Action requiredStrong evidenceSupplier selection · certified capability required

Excluded from the shortlist in 157 of 270 observed decisions, in every case where a certified capability was required.

270 observations · 9 environments · 90 days window · last observed 2026-07-29 09:12 UTC

Considered
78%211 / 270
Shortlisted
42%113 / 270
Recommended first
17%46 / 270

Loss classification

Lost to omission, not comparison

The organisation was not rejected on its merits. The capability required by the scenario was not reconstructed at all, so it was never evaluated.

Position taken by

  • Competitor B121 observations · 94 while absent
  • Competitor C36 observations · 29 while absent

Stated reasons, as observed

  • Certified capability could not be confirmed61 observations
  • Scale described as smaller than verified34 observations
  • Category positioning recognised188 observations
“For that requirement I would suggest two suppliers. I could not confirm a certified capability for the first one, so my recommendation is the second.”OBS-1219 · Claude · API · native · repetition 3 of 5
Every figure resolves to recorded observations. No number without evidence.Demonstration dataset. Structure and methodology are the production ones.
What Vorentus does

From verified reality to machine decisions.

  1. 01Verifytruth

    Establish the claims the organisation can actually prove, with their evidence.

  2. 02Observesignals

    Record how machines reconstruct those claims, repeatedly and under stated conditions.

  3. 03Comparegap

    Measure where reconstruction diverges from verified reality, and by how much.

  4. 04Decidechoice

    Test what happens when a machine has to choose between alternatives.

The full cycle
  1. Truth
  2. Distribution
  3. Observation
  4. Recognition
  5. Decision
  6. Action
  7. Retest
  8. Proof

Truth → Understanding → Decision → Proof

AI environments

One organisation. Multiple machine realities.

AI systems do not all see the same information or produce answers in the same way. Vorentus observes machine understanding across models, grounded search and consumer AI surfaces.

  • OpenAIChatGPT
  • ClaudeClaude
  • GeminiGemini
  • PerplexityPerplexity
  • GrokGrok
  • Microsoft CopilotMicrosoft Copilot
  • MistralMistral
  • DeepSeekDeepSeek
  • Google AI ModeGoogle AI Mode
  • Google AI OverviewGoogle AI Overview
  • Meta AI / WhatsAppMeta AI / WhatsApp
OC-01
Model observation
ChatGPT · Claude · Gemini · Grok · Mistral · DeepSeekProgrammatic observation
No live retrieval
OC-02
Grounded search
Perplexity · Grok Web Search · Google AI Mode · Google AI OverviewProgrammatic observation
Live retrieval
OC-03
Consumer surfaces
Meta AI / WhatsApp · Microsoft CopilotCertified manual capture
Assistant-mediated

Not every AI observation is equivalent.

Vorentus preserves the environment, model, grounding, observation method and repetitions behind the evidence.

Consumer surfaces are recorded through certified manual capture, not an automatic integration, and are reported separately from programmatically observed environments.

All product names, logos and brands are property of their respective owners and are shown for identification purposes only. Vorentus is independent and not affiliated with, endorsed by or sponsored by any of these providers.

Cross-machine intelligence

There is no single AI perception.

The same verified claim can survive differently across models, grounded search experiences and consumer surfaces.

C-01

Category and positioning

Model observation
  • GPT Recognised
  • CLD Recognised
  • GEM Recognised
  • GRK Partial
  • MST Partial
Grounded search
  • PPX Recognised
  • AIM Recognised
  • AIO Recognised
  • GWS Recognised

Across the claims shown, recognition holds more often where retrieval is available. Native 4 / 20 · Grounded 11 / 16

RecognisedPartialInconsistentContradictedUncertainOmittedBlue line = verified reference. Distance below it is schematic — it encodes the recorded state, not a measured quantity. Illustrative example · not client data.

One organisation. Multiple machine perspectives. One verified reference layer.

Decision intelligence

Visibility tells you where you appear. Decision Intelligence shows what happens when the machine has to choose.

An organisation can appear consistently, be described accurately and even be cited — and still not be the one recommended when an AI system compares alternatives.

Real production observationAnonymised subject · observed by Vorentus 04 Aug – 07 Aug 2026 · SCN-A89B
Organisation under measurement

A wine recommendation platform

0

Not recommended in any of 442 valid documented observations.

Unaided recommendation scenario · 04 Aug – 07 Aug 2026 · real production observation.

Scenario
Choosing a wine recommendation assistant
Exposure
Unaided — the subject is never named in the prompt
Observation period
04 Aug – 07 Aug 2026
Environments
4 AI environments
Stored rationales
211 observations
Technical failures
45 scheduled runs (Perplexity)
What was recommended instead

Same customer need. Different machine environments. Materially different recommendation sets. Each figure is the number of valid observations in which an alternative appeared among the recommended options — counted once per observation, shown against that environment's own valid denominator, never pooled.

OpenAIgpt-4o-mini
valid observations 121subject recommended 0

58 distinct alternatives named · observed recommendation frequency, not market share · 52 further in the long tail

Claudeclaude-sonnet-4-5
valid observations 111subject recommended 0

59 distinct alternatives named · observed recommendation frequency, not market share · 53 further in the long tail

Geminigemini-3.6-flash
valid observations 112subject recommended 0

52 distinct alternatives named · observed recommendation frequency, not market share · 46 further in the long tail

Shown separately

Perplexity is shown separately because retrieval conditions differed and 45 scheduled runs ended in technical failure. Only the grounded subset carries retrieved source citations; the native environments above returned none.

Perplexitysonar-pro
valid observations 57subject recommended 0

22 distinct alternatives named · observed recommendation frequency, not market share · 16 further in the long tail

Retrieval: web search · 57 of 57 observations carried retrieved citations.

Perplexitysonar-pro
valid observations 41subject recommended 0

15 distinct alternatives named · observed recommendation frequency, not market share · 9 further in the long tail

Retrieval mode unspecified in the stored records — not interpreted as native.

Association, not causation

What did the responses associate with these recommendations?

Drawn verbatim from the 211 stored observation rationales. These are phrases repeatedly described alongside the recommendations, not established causes. Subject references are anonymised; nothing else is rewritten.

OpenAIOBS-365A
The organisation under measurement was not included in the response. The AI recommended building a custom solution using general NLP models (e.g., GPT-4) combined with third-party wine APIs (Wine-Searcher, Vivino) rather than suggesting dedicated off-the-shelf AI sommelier products.
ClaudeOBS-100E
The AI recommended building a custom LLM solution or using generic chatbot platforms (Chatbase, Dante AI, CustomGPT, OpenAI, Anthropic) rather than recommending an off-the-shelf specialized wine sommelier platform, and did not mention the organisation under measurement.
GeminiOBS-D57B
The organisation under measurement was not mentioned in the AI's response. The AI recommended building a custom AI sommelier using platforms like Voiceflow, Botpress, OpenAI GPT-4o, or Claude 3.5 Sonnet with RAG integrated into a store POS/e-commerce inventory, or using generic e-commerce tools like Octane AI or Tidio AI.
PerplexityOBS-5946
Sommelier.bot was selected as the top recommendation for a wine shop because it is explicitly designed for retailers, supports direct inventory integration (CSV, XML, Vivino feed), embeds easily onto a website via JavaScript, and focuses on shop-branded recommendations from active store stock.

The organisation under measurement may genuinely hold the relevant strengths in verified reality. If they do not survive machine interpretation, they cannot carry weight when the machine has to choose.

That is no longer a visibility problem.
It is a machine-understanding problem.

Start here

Recognition Baseline

A measurable starting point. The assessment establishes verified reality, shows where machine interpretation diverges from it, and identifies what deserves attention first.

What it establishes
  • A verified reference for the organisation.
  • How selected AI environments currently reconstruct it.
  • A documented reference point for later comparison.
What it can identify
  • Claims that do not survive machine interpretation.
  • Contradictions and omissions across environments.
  • Where the organisation is not selected, and why.
What the organisation receives
01
Verified truth reference
Critical organisational claims and evidence used as the reference layer.
02
Recognition Score
How consistently important facts and attributes are recognised.
03
Critical Gaps
Important omissions, contradictions and inconsistencies.
04
Cross-Environment Analysis
Where different AI environments understand the organisation differently.
05
Competitive Perception
How relevant attributes are associated with you and selected competitors.
06
Decision Snapshot
Observed behaviour in defined comparison and recommendation scenarios.
07
Evidence Assessment
Where important claims appear strongly or weakly supported.
08
Priority Actions
The highest-priority issues to address.
09
Documented reference report
A documented starting point for future comparison.
€450

One-time assessment · No subscription required

The Baseline is the measurement. Retesting is how improvement becomes evidence.

Evidence & provenance

A number without provenance is not intelligence.

Every score can be opened until the original machine response is on screen. The record below is one Recognition Score, entered layer by layer, down to the words the machine actually produced and the evidence behind them.

The score is not the evidence. The score summarises the evidence.

74Recognition
Score
The score is not the evidence.
It summarises the evidence below.
  1. Dimension

    Attribute recognition

    One of five scored dimensions

  2. Verified claimC-07

    Certified production capability

  3. EnvironmentENV-02

    Claude · API · native model

  4. Decision scenarioS-03

    Comparative recommendation

  5. RepetitionREP-04

    Fourth of five repeated runs

  6. ObservationOBS-184

    Recorded 2026-07-29 · 09:14 UTC

    Grounding: none · access: API

Original response · verbatim
“For that requirement I would suggest two suppliers. I could not confirm a certified capability for the first one, so my recommendation is the second.”
Classification
OmittedCapability not reconstructed in this response
Supporting evidence
EV-881Verified certificate — published and machine-readable
Illustrative evidence trail. Every scored dimension resolves to observations recorded this way.
Evidence before optimisation

Don't optimise a problem you haven't diagnosed.

One weak observed outcome — an attribute the machine did not reconstruct — can arise from conditions that have nothing in common except the result.

  • The fact does not exist publiclyEstablish the fact in a durable public form
  • Supporting evidence is weakStrengthen provenance and corroboration
  • Evidence is inconsistent across sourcesReconcile conflicting statements
  • Information is difficult for machines to retrieveImprove machine-readability and persistence
  • The attribute is recognised but not relevant to the decisionReconsider which attributes the scenario requires
  • A competitor holds stronger evidence for the same attributeCompetitive evidence work, not visibility work

Different problems require different interventions. Vorentus diagnoses before recommending action.

Longitudinal intelligence

Machine perception changes. Your evidence should not disappear with it.

An isolated answer is a snapshot. Each observation is preserved with the conditions under which it was recorded, so a later state can be compared with an earlier one instead of asserted against it.

  1. 2026-03-04Reference measurement

    Certified capability not reconstructed in native model responses

    Omitted
    • Claim C-07
    • Scenario S-03
    • 9 environments
    RUN-1180
  2. 2026-04-16New observation

    Same omission recorded again; grounded retrieval reconstructs the claim

    Inconsistent
    • Claim C-07
    • Scenario S-03
    • Native and grounded
    RUN-1477
  3. 2026-05-02Intervention

    Persistent machine-readable evidence published for the claim

    Verified evidence
    • Evidence EV-881
    • Structured data
    • Third-party corroboration
    INT-042
  4. 2026-06-18Comparable retest

    Observations repeated under the conditions of the reference measurement

    Under comparison
    • Same claim
    • Same scenario
    • Same environments
    RUN-1902
  5. 2026-07-29Comparable retest

    Claim reconstructed in native responses where it was previously omitted

    Recognised
    • Same claim
    • Same scenario
    • Same repetitions
    RUN-2416

Comparable conditions — the dashed entries repeat the reference measurement: same verified claim, same decision scenario, same environments, same number of repetitions. Two unrelated AI answers are not evidence of change; only comparable observations can be placed side by side.

Illustrative sequence with illustrative states. No uplift, percentage or point change is shown, because none has been measured here. Each real entry preserves environment, model, access mode, grounding, scenario, date, repetition, response, citation, classification and confidence.

Stop asking: “What did ChatGPT say today?”

Start asking: how has machine understanding changed — and what evidence do we have?

Evidence records what was observed. The Ledger makes comparable observations visible over time.

What deserves attention next?

Vorentus Copilot

Know what matters next.

More observations do not automatically create more intelligence. Copilot connects evidence across claims, environments, competitors, decisions and retests to surface what deserves attention — a control tower over the evidence layer, not a conversation.

Interpretation stays attached to the observations it came from, so the next action is argued rather than asserted.

Vorentus Copilot · control towerEvery line evidence-linked
Question

Why are we recognised in grounded search but inconsistently in native model responses?

Observed evidence
  • C-04
  • OBS-172
  • OBS-188
Native model responses
Inconsistent
The claim is reconstructed in some repetitions and omitted in others.
Grounded retrieval
Recognised
The claim is reconstructed consistently across the repetitions recorded.
Interpretation
  • OBS-172
  • OBS-188

Recognition appears stronger where retrieval is available. Reported as an association observed under the tested conditions — not as a cause.

Recommended action
  • C-04
  • EV-881

Strengthen persistent machine-readable evidence supporting the affected claim.

Retest
  • S-03
  • ENV-02

Repeat comparable observations under the same environments, scenarios and repetitions, then compare against the reference measurement in the Ledger.

Illustrative example. Copilot interprets the evidence layer; it does not generate conclusions independently of it.

Industry relevance

Built for organisations where machine understanding has consequences.

The problem is not sector-specific. It becomes material wherever AI systems shape how organisations, products and services are discovered, compared and recommended.

The underlying question

Do machines understand your organisation well enough to choose it for the right reasons?

Definitions
Machine Understanding
The way an AI system represents and interprets an organisation, its attributes, evidence and relationships.
Recognition
The degree to which a verified organisational claim survives machine interpretation.
Decision Simulation
A documented observation in which an AI system is asked to compare alternatives or recommend an option under defined conditions.
What we do not promise

We don't promise to control AI.

Independent AI systems cannot be commanded to believe, rank or recommend an organisation. Vorentus does not pretend otherwise.

  1. Measure
  2. Diagnose
  3. Improve
  4. Retest
  5. Prove
  • No guaranteed rankings.
  • No artificial certainty.
  • No claims of controlling independent AI systems.
  • No confusing correlation with causation.

Evidence before claims.

You already manage what your organisation says

Now measure what machines understand.

Your website
tells you what you publish.
Analytics
tells you what humans do.
Search intelligence
tells you where you rank.

Vorentus answers another question: what do machines actually understand about your organisation — and what happens when they have to choose?

FAQ

Common questions

What is Vorentus?
Vorentus is a Machine Understanding & Decision Intelligence platform that measures the gap between verified organisational reality and how AI systems understand, compare and recommend an organisation.
Is Vorentus an AI visibility platform?
No. Visibility tools measure presence, mentions and citations. Vorentus measures whether organisational facts survive machine interpretation, and what happens when a system has to compare or recommend.
What is the Recognition Baseline?
A one-time assessment that establishes a verified reference for the organisation and measures current machine understanding against it.
What is the Truth Graph?
The structured set of organisational claims, attributes and supporting evidence used as the reference layer for every measurement.
What is Recognition Intelligence?
The measurement of how consistently individual verified claims are reconstructed by AI environments, claim by claim rather than in aggregate.
What is Decision Simulation?
A documented observation in which an AI system is asked to compare alternatives or recommend an option under defined conditions.
What is the Machine Perception Ledger?
The longitudinal record of observations, each preserved with the conditions under which it was recorded, so later states can be compared rather than asserted.
Which AI environments are observed?
Leading AI models, grounded-search experiences and selected consumer AI surfaces. Observation methods are identified separately because those environments are not methodologically equivalent.
Can Vorentus control what AI systems say?
No. Vorentus observes, measures, diagnoses and retests independent AI systems. It does not claim to control them.
Does Vorentus guarantee recommendations or rankings?
No. Outcomes in independent AI systems cannot be guaranteed. What can be produced is documented change under comparable conditions.
Why are observations repeated?
A single answer is not a measurement. Repetition under controlled conditions distinguishes a stable interpretation from variance in the response itself.
How is Vorentus different from GEO or AEO?
GEO and AEO are optimisation practices aimed at improving presence in generated answers. Vorentus is a measurement and evidence system: it diagnoses what machines understand before anything is optimised.

Vorentus measures the gap.