Assessment integrity · Effective 27 July 2026

Ratings & Disclosures

A clear guide to the scope, evidence, limits, commercial separation, and correction process behind every Aurxo label.

Read a label as evidence, not a promise. Aurxo ratings are comparative, scope-specific assessments. “Greener” does not mean zero impact, climate neutrality, environmentally harmless, or officially certified. The grade is meaningful only together with its scope, date, evidence confidence, sources, and limitations.

1. What an Aurxo rating means

An Aurxo grade summarizes the evidence available for a defined AI product or reference model under a stated use and system boundary. Aurxo considers Energy, Carbon and Water under a consistent internal assessment model. Missing or non-comparable evidence can prevent an overall grade. Transparency and evidence confidence are shown separately and never turn weak evidence into a stronger environmental claim. Aurxo publishes the sources, metrics, scope, confidence and limitations needed to understand and challenge the result; exact thresholds, weights and implementation details remain proprietary.

2. What “greener” means

“Greener” is a comparative term. It means that, for the stated task and compared options, one result has a stronger supported environmental profile under Aurxo’s current methodology and evidence. It does not mean the selected product has no footprint, is sustainable in every respect, or will always use fewer resources. Prompts, reasoning depth, output length, modality, hardware, data-centre location, electricity mix, utilisation, and provider routing can materially change a result.

3. Not an official certification

Aurxo Baseline v1.0 is an early public methodology produced by Roitels Inc. It is not a government label, regulatory approval, ISO certification, independent laboratory certification, carbon-neutrality claim, or guarantee by the assessed provider. The use of a grade does not imply that the assessed company participated in or approved the assessment.

4. Scope matters

Current product assessmentApplies to a named consumer or enterprise product only when a sufficiently current, comparable product-level boundary is available.
Scoped reference labelApplies only to the named model, prompt, token counts, hosting assumptions, study date, and functional unit.
Withheld gradeA dash means the required evidence is incomplete, stale, or not comparable. It is not a failing grade.
Task recommendationA tool must first clear Aurxo’s task-fit threshold. Among qualified options, Aurxo prioritizes the strongest scope-compatible environmental signal. A lower-impact tool that cannot do the job well is never promoted.

5. Evidence categories

Externally reviewed evidence

Evidence reviewed or audited independently under a disclosed method. Its scope and limitations still apply.

First-party study or provider-measured data

Measurements published by the provider. Aurxo identifies them as first-party and does not imply that it audited the underlying telemetry.

Independent estimate

A modelled result created by researchers using public assumptions. It can support a scoped reference but is not equivalent to production telemetry.

First-party claim

A provider statement that lacks enough reproducible detail to be treated as a full measurement.

Unknown or not disclosed

No sufficiently comparable public evidence was identified. Aurxo does not turn absence of data into a favourable score.

6. Sources, dates, and limitations

Every public assessment should identify the product or model, assessed task or functional unit, system boundary, evidence date, methodology version, metric values, evidence type, confidence, source links, and known limitations. If the interface and a detailed assessment page appear to differ, the detailed public assessment page is the authoritative record for that version.

7. Commercial relationships and editorial separation

Roitels Inc. may in the future sell enterprise tools, reports, API access, or assessment services. Payment does not buy or alter a grade. Any material commercial relationship with an assessed provider will be disclosed where it could reasonably affect how a user interprets the assessment. Enterprise customers and evidence submitters do not receive preferential ratings.

8. Trademarks and provider relationships

ChatGPT, Claude, Gemini, Microsoft Copilot, Perplexity, Le Chat, DeepSeek, and other product and company names are trademarks of their respective owners. Unless explicitly stated, Aurxo and Roitels Inc. are not affiliated with, sponsored by, endorsed by, or approved by the companies assessed.

9. Corrections, appeals, and evidence submissions

Providers, researchers, and users may submit a correction request or new evidence. Include the assessment, disputed statement, proposed correction, supporting source, publication date, methodology, functional unit, and any relevant conflicts of interest.

How to submit: email aurxo@roitels.com with the subject “Aurxo rating correction or evidence submission.” Credible submissions will be considered in good faith. A submission does not guarantee a response by a particular date or a grade change. Aurxo may publish a correction note, revise the assessment, retain the result with added context, or decline unsupported requests.

10. Task recommendations

Aurxo recommendations compare documented product capabilities against the context detected in the user’s task. These capability profiles are editorial, dated and subject to change as products evolve. They are not warranties of output quality, availability, privacy, fitness for a regulated use, or access under a particular subscription. Aurxo first removes tools that cannot do the job well. Among the remaining qualified options, it prioritizes the strongest comparable environmental signal. If no greener recommendation can be verified, Aurxo retains task-fit order and makes the evidence gap visible.

11. Versioning and withdrawals

Assessments are dated and versioned. Aurxo may update, supersede, correct, or withdraw a rating when the product changes, evidence becomes stale, a source is corrected, or the methodology evolves. Historical references should not be presented as current universal product grades.

12. Questions

For methodology, rating, legal, or relationship disclosures, contact aurxo@roitels.com.

Evidence behind every label.

Every Aurxo assessment shows its scope, sources, confidence and known limitations.

How Aurxo ratings work · View current assessments