AI attention
A versioned public prompt corpus is sampled across supported answer engines. Recommendation frequency, prominence, citations, category breadth, and provider breadth contribute to visibility.
Measure the signal.
Show the source.
Every public ranking is a versioned snapshot—not an opaque live score. This page describes the principles the production data pipeline is designed to enforce.
What makes public rankings trustworthy
Compute markets
Verified rental listings are normalized to price per physical GPU hour. Accelerator model, memory, topology, contract type, geography, and availability are preserved.
Public equity prices
Licensed end-of-day closes are attached only to reviewed exchange-traded securities. The displayed move compares two available closes; it is not an intraday quote.
Private-company valuations
A private valuation appears only when the company or transaction participants disclose a financing valuation. It is labeled by transaction date and is never presented as a stock price.
Prediction signals
Prediction-market prices remain raw market signals unless a versioned, reviewable transformation defines an implied valuation. They are never mixed with disclosed transaction valuations.
Company financials
Public filing facts are mapped into a versioned financial taxonomy. Debt, cash, leases, EBITDA, and capital expenditure retain source period and filing provenance.
Model economics
Published API pricing and measured influence are separate metrics. Model influence does not represent intelligence, safety, or benchmark quality.
Confidence
Every score includes observation coverage, source freshness, provider success, and entity-resolution confidence. Low-confidence values are withheld from ranking.
Corrections
Entity merges, source corrections, and methodology changes are logged. Historical snapshots remain tied to the methodology version used at publication.
Commercial relationships do not determine corpus membership, scoring, or rank. Rank.ai customers receive analysis and recommendations—not favorable public scores.
Customer prompts remain private
Public rankings use a separately governed Rank.ai benchmark corpus. Private customer prompts and competitive sets are never published or used to identify public ranking entities without explicit consent.
Context, not investment advice
Stock, debt, and capital-expenditure data is presented as factual market context. Rank.ai scores do not predict securities performance and should not be treated as investment recommendations.