---
language: en
title: "Payment per inference"
description: "Payment per inference links compensation to the use of content in an AI output. Learn how it differs from retrieval and what evidence is needed for billing."
canonical: "https://tropk.ai/en/resources/glossary/payment-per-inference"
last_modified: "2026-10-12T16:10:54.641Z"
---

# Payment per inference

Reviewed on 2026-10-07.

## What it means

Payment per inference, or pay-per-inference, refers to a model that compensates the rights holder when their content contributes to an AI-generated output. The compensated activity is the use of the material, not merely its [retrieval](/en/resources/glossary/information-retrieval). The [RSL](/en/resources/glossary/really-simple-licensing) 1.0 specification provides for the use modality for contributions to an output, including inference, [grounding](/en/resources/glossary/grounding) and generation, without demonstrating universal adoption.

## Why it matters for marketing

For marketers and publishers, this proposal shifts the commercial discussion from the number of automated visits to the use of their content libraries. This requires defining what counts as a contribution, how it will be recorded and which event triggers compensation. A revenue forecast based only on public citations may omit uses without citations or count mentions without actual use of the content.

## Example in practice

Suppose a publisher negotiates compensation for the use of equipment data sheets in technical answers. The partner’s hypothetical report identifies the data sheet, the answer and the contribution event; the retrieval record alone does not show this connection. The publisher decides to require this association to reconcile charges for the data sheets before accepting the reported total.

## How to apply or verify

Before signing a contract, establish the unit of use, how content is identified, how copies are handled and the rule for answers with multiple sources. Request sample events and a dispute procedure. In RSL, payment terms and reporting requirements can be declared, but the adopted profile needs to specify the format, fields and how the evidence works.

## Important distinctions

This payment to the rights holder differs from the computational cost of running a model. It also differs from paying for the request that retrieves a page. A licensing declaration does not, by itself, set up measurement or billing. [Citation](/en/resources/glossary/ai-citation), retrieval and contribution are distinct observations; the presence of one does not automatically imply the others or allow amounts owed to be calculated.

## Sources and references

- [Really Simple Licensing (RSL) 1.0 Specification](<https://rslstandard.org/rsl>) — RSL Collective — Technical Steering Committee

## Continue a leitura

- [Content licensing agreement](</en/resources/glossary/content-licensing-agreement>)
- [Pay-per-crawl](</en/resources/glossary/pay-per-crawl>)
- [Really Simple Licensing](</en/resources/glossary/really-simple-licensing>)

## About the author

[Luiz Gardelin](</en/team/luiz-gardelin>)

Luiz Gardelin is the founder of Tropk.ai and works in GEO/AEO, product vision, strategy, and growth, helping brands understand and improve their presence in AI responses.

---

[All pages in Markdown](<https://tropk.ai/sitemap.md>)