Can You Price AI Usage? A Documentation Audit of Credit-Based AI Products

Abstract Many AI products sell a named credit that users spend as they run actions. Whether a buyer can judge cost depends on what the public documentation shows: how credits map to actions, what a credit costs, and how spending is attributed afterward. This study audits first-party documentation for 22 AI prod-ucts and asks three questions. Before a run, can a buyer work out the credit cost of a named action? Can the buyer turn those credits into dollars on the same pages? After use, do the pages show where the credits went? The main finding is that documentation is stronger at stating a credit number than at supporting a decision. 17 of 22 products publish an exact credit amount for at least one action. Of these, 7 also state a dollar rate for that action, and 5 of those also have a known or buyer-set quantity, so only 5 reach a computable dollar quote for one event (6 if a rate derived from a same-page plan price counts). On coded pages the usual gap is the dollar rate: 9 products have a known quantity but state none there. A dated read of other same-domain pages (Section 2) raises the products with a dollar rate or a plan price with a credit allowance from 11 to 19 of 22; the added figures are plan prices, not event quotes. Spend after use is less gran-ular still: only 5 of 25 surface records describe an amount per action, job, query or session, and most show totals or a balance only. A credit can carry different prices by route, so a figure needs its route. Providers could close common gaps by stating the dollar value of a credit beside each credit amount, publishing an action-level rate table, and showing credits per action in the usage record. These are implications, not measured effects. The results describe public documentation for a frame of 76 candidates, not product behavior, billing accuracy or market prevalence.

Authors

Publication Details

Journal
Research Square
Published
2026-11-12
DOI
https://doi.org/10.21203/rs.3.rs-11707201/v1
Primary Topic
Ethics and Social Impacts of AI
Type
preprint
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
preprint

Can You Price AI Usage? A Documentation Audit of Credit-Based AI Products

Pratik Gupta
Research Square
Ethics and Social Impacts of AI
preprint

Can You Price AI Usage? A Documentation Audit of Credit-Based AI Products

Pratik Gupta
preprint en

Abstract

Abstract Many AI products sell a named credit that users spend as they run actions. Whether a buyer can judge cost depends on what the public documentation shows: how credits map to actions, what a credit costs, and how spending is attributed afterward. This study audits first-party documentation for 22 AI prod-ucts and asks three questions. Before a run, can a buyer work out the credit cost of a named action? Can the buyer turn those credits into dollars on the same pages? After use, do the pages show where the credits went? The main finding is that documentation is stronger at stating a credit number than at supporting a decision. 17 of 22 products publish an exact credit amount for at least one action. Of these, 7 also state a dollar rate for that action, and 5 of those also have a known or buyer-set quantity, so only 5 reach a computable dollar quote for one event (6 if a rate derived from a same-page plan price counts). On coded pages the usual gap is the dollar rate: 9 products have a known quantity but state none there. A dated read of other same-domain pages (Section 2) raises the products with a dollar rate or a plan price with a credit allowance from 11 to 19 of 22; the added figures are plan prices, not event quotes. Spend after use is less gran-ular still: only 5 of 25 surface records describe an amount per action, job, query or session, and most show totals or a balance only. A credit can carry different prices by route, so a figure needs its route. Providers could close common gaps by stating the dollar value of a credit beside each credit amount, publishing an action-level rate table, and showing credits per action in the usage record. These are implications, not measured effects. The results describe public documentation for a frame of 76 candidates, not product behavior, billing accuracy or market prevalence.

Research Square
Ethics and Social Impacts of AI
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon