RP RoyaltyProof

Royalty-audit rigor for AI content licensing

AI platforms pay you on numbers they report themselves.
Who audits the results?

Publishers now collect royalties from Microsoft Publisher Content Marketplace, ProRata, Cloudflare pay-per-crawl, TollBit, and direct model-maker deals — each one settled on self-reported citation and crawl counts, on separate and uncorrelated rails, with no independent verification. RoyaltyProof reconciles those statements against your own edge logs and an independent citation panel. We work on contingency, so a review costs you nothing up front.

See the reconciliation demo Request an audit
Music-audit precedent
10–20%underpayment routinely recovered
Cloudflare disclosure
~50%of crawls never surface in answers
Your upfront cost
$0we are paid from what we recover

The problem

A new revenue line with no audit function attached

The money is real and growing: content-licensing programs from the major AI platforms are now cutting royalty checks to publishers. But the checks are sized by numbers the payer generates — how many times your content was crawled, cited, or surfaced in an answer. The publisher receives a statement and a payment. There is no invoice to contest, no meter you control, and no third party attesting that the count is complete.

Every other high-value royalty stream — mechanical, performance, sync — is audited as a matter of course. This one arrived without that discipline.

  • Uncorrelated statements Each rail — PCM, ProRata, pay-per-crawl, TollBit, direct deals — reports differently, on its own cadence, with no shared definition of a "citation." Errors don't cancel out; they hide in the gaps between systems.
  • The payer holds the meter Counts are self-reported. Under-counting looks identical to lower demand. Without your own logs beside the statement, there is nothing to reconcile against.
  • Known leakage Cloudflare has publicly noted that roughly half of AI crawls never surface in a user-facing answer. Whatever the settlement basis, the gap between what's fetched and what's credited is exactly where value goes missing.
  • No one is checking Finance teams book the revenue and move on. "Who audits the results?" is, today, an open question — the same open question the music industry answered decades ago with routine royalty examinations.

The service

Statement-versus-log reconciliation, on contingency

Statement reconciliation

We line up each platform's royalty statement against your own edge and server logs, day by day and rail by rail, and quantify where credited activity falls short of what your infrastructure actually recorded.

Independent citation sampling

A sampling panel queries the AI products directly and records where your content is surfaced and cited, giving an outside measurement that doesn't depend on the payer's own instrumentation.

Contingency pricing

We are paid a percentage of what we recover — typically around 30%. If a review finds nothing, you owe nothing. The incentive to be thorough is entirely ours.

You risk nothing to find out. The audit pays for itself out of recovered underpayment, or it is free.

How an engagement runs

Four steps, no upfront fee

  1. 01

    Scoping

    We identify which payout rails you are on and confirm you can export edge/server logs for the audit window. One short call.

  2. 02

    Reconciliation

    We ingest your logs and each platform's statements, normalize the counts to a common definition, and build the day-by-day diff.

  3. 03

    Sampling & findings

    The citation panel corroborates the log gaps. You receive a findings report with the discrepancy quantified per rail and in aggregate.

  4. 04

    Recovery

    We support the claim back to each platform. Our fee comes out of what is recovered — never before.

Interactive demo

Underpayment estimator & reconciliation demo

Two ways to see the method. First, a back-of-envelope estimate of what a review might recover for you. Then a full day-by-day reconciliation on a synthetic month of data — or paste your own.

Do you retain edge / server logs for the audit window?

Range anchored to the 10–20% underpayment routinely recovered in music royalty audits, adjusted down when fewer rails or weaker logs reduce what can be substantiated. Illustrative — not a projection of your results.

Precedent

The music industry already ran this experiment

For decades, labels, publishers, and artists have collected royalties settled on counts the payer reports. The response was not to trust the statement — it was to audit it. Royalty examinations became standard practice precisely because self-reported statements systematically understate what's owed, and audits routinely recover 10–20% underpayment. AI-licensing royalties are the same structure — self-reported counts, a payer who holds the meter, a publisher with no invoice to contest — arriving before the audit function did. RoyaltyProof brings the established method to the new rail.

Find out what your statements are missing

Send us the rails you're on. We'll scope a contingency audit — no upfront fee, and nothing owed unless we recover.

Request an audit — ebo@testmachine.ai