AI Content Licensing Models: Who Holds the Meter | Elacity
AI companies pay for content by flat fee, per crawl, per volume or per use. The model decides what reaches you, and in most of them the buyer does the counting.
AI Content Licensing Models: Who Holds the Meter Decides What You Get Paid
AI companies pay for content in four main ways: a flat fee for an archive, a price per crawl, a fee per volume of text delivered, or a fee each time your work feeds an answer. The model decides how much of your work's ongoing value ever reaches you, and in most of them the buyer is the one doing the counting.
That question got sharper on November 10. VerticalScope, a Toronto company that says it runs more than 1,200 online communities holding over 2 billion user-generated posts, announced data agreements with two frontier AI labs. Its fees can follow the volume of text the labs accept, or how much the content contributes to AI answers, with that usage as determined by the counterparty.
Read the last phrase again. The lab using the posts is also the one measuring the use. The announcement does not mention the members who wrote them.
We argued in October that AI training data licensing is a real market now. This guide is about how that market pays, and who it pays.
The Four AI Content Licensing Models
- Flat fee: a lump sum or annual payment for access to an archive. Reddit's deal with Google was reported at about $60 million a year. It is simple, and it prices yesterday's archive rather than tomorrow's use.
- Per crawl: a price each time a bot fetches a page. Cloudflare's pay per crawl answers an AI crawler with HTTP 402 Payment Required and a price, and lets the site owner allow, charge or block each crawler.
- Per volume delivered: payment by the quantity of text a lab accepts, measured in tokens. This is the first of VerticalScope's two fee streams.
- Per use: payment each time content contributes to an AI-generated output. The RSL 1.0 specification defines this as its own payment type, and Microsoft says publishers in its Publisher Content Marketplace will be paid on delivered value, with usage-based reporting.
Read the list as a ladder, not a menu. Each step moves payment closer to the moment your work is actually used, which is where its value sits.
Where Each Model Puts the Meter
A flat fee needs no meter. That is its appeal and its limit: once the archive is delivered, nothing written afterwards is priced unless someone negotiates again.
Per crawl puts the meter at your front door. It counts fetches, though, and a page fetched once can sit inside a model for years after the single payment.
Per use is the closest to fair and the hardest to count. In the RSL Collective's process, the AI system tracks its usage and sends a usage report with its payment. When RSL launched, The Verge noted there is no simple way to tally an inference fee without the AI company's participation.
That is the gap in one sentence: pay per use only pays as well as the count, and today the count lives inside the buyer's systems.
Why Per-Use Pricing Is Coming Anyway
Software is starting to carry its own budget. On November 11, Rain launched Agent Vaults, where an AI agent spends from a balance its owner funds, within rules the owner sets, using a card issued for a single task.
When the buyer is an agent paying per task, a per-use price stops being a negotiation and becomes a checkout. The open question is who stands at the till.
Who Gets Paid: Platform, Publisher or Person
Every deal above is struck by whoever holds the archive: the platform, the publisher, the forum operator. A stricter version arrived the same day. The Korean Publishers Association's agreement with five AI data firms makes supply terms subject to the consent of copyright holders, and requires the data to be returned or destroyed when a project ends.
That is consent at the level of the rights holder, written into the deal. It still runs through an association. The person who wrote a forum answer an assistant quotes next year has no equivalent seat.
Elacity's founder, Sasha Mitchell, puts the principle plainly: "The people who create the value should own it. That is the entire reason Elacity exists."
The Elacity Approach: Put the Meter in the Owner's Hands
Elacity is building toward a different order of events. With Elacity dDRM, a creator packages a post, an article, a dataset or a body of work into a Wealth Capsule, with its terms and royalties written in.
The design aim is that using a Wealth Capsule and paying for that use become one event, recorded where the owner can see it. The count would then stop depending on a report from the buyer.
That would make per-use pricing something an individual can set, not only a publisher with a legal team. Royalties are a feature of the asset, not a promise of income: whether anyone uses your work is still up to the market.
This is what we are building, not a product you can use today, and it will not settle every question. Measuring how much one source contributed to a model's answer is genuinely hard, and no design removes the need for a buyer willing to license. What it can change is who holds the meter.
AI Content Licensing FAQ
Can an individual creator license content to AI companies today?
Mostly through a platform or publisher that aggregates content. The RSL group says any online publisher can start using its standard today to declare terms, but collecting depends on AI companies honouring them.
Does blocking AI crawlers stop training on my work?
Blocking stops future fetches by crawlers that respect it. It does not reach copies already taken, which is where opt-out stops.
Which licensing model pays creators best?
None guarantees payment. Per-use pricing ties pay most closely to value, but only when the count is trustworthy and visible to the owner.
Before you accept any AI licensing offer, ask one question: who counts the use? More analysis lives in Market Intelligence. To follow the argument as it develops, Follow Elacity on X.