# Best AI for Finance evidence methodology

Version: 2026-09-04

Reviewed: 2026-09-04

License: CC BY 4.0

## Scope

This dataset is a repeatable audit of first-party product, pricing, security, privacy, and help documentation for the products compared at https://llm-stats.com/best-ai-for-finance.

It is not a standardized hands-on product-performance test. A published vendor claim is evidence that the claim exists, not proof that the feature performs well. Missing public evidence remains missing and is not converted into a negative product score.

## Reproduction procedure

1. Open every first-party URL in the evidence file.
2. Record whether a numeric US price is publicly visible. Classify the result as public_numeric_price, partial, or quote_only.
3. Record whether the cited materials state concrete security, certification, encryption, retention, or model-training controls. Otherwise mark contract_confirmation_required.
4. Trace each listed capability to the cited product documentation.
5. Record the review date and retain the unmodified evidence file so later changes can be compared.

## Recommendation method

Rank and use-case recommendations combine documented source grounding, workflow fit, security controls, pricing transparency, and the selection criteria published on the page. They are editorial judgments rather than a mathematical product score. LLM Stats accepts no payment for ranking position.

## Required field test

Before purchase, run every finalist on the same representative source files and known-answer tasks. Keep model and product versions fixed. Measure accuracy, source validity, recall, false positives, reviewer correction time, repeatability, latency, and total cost. For high-consequence work, require review by the appropriately qualified professional.