RSL (Really Simple Licensing)
An open standard for stating, in machine-readable form, what an AI system may do with a page and what it costs. Launched 2025-09-10 by the RSL Collective and declared version 1.0 in December 2025. Where llms-txt tells a model how to read a site, RSL tells it on what terms — the two occupy the same file-in-the-root slot and answer opposite halves of the question.
The license terms it can express
- Free — no compensation.
- Attribution — credit required, Creative Commons style.
- Subscription — access under a negotiated agreement.
- Pay per crawl — payment per indexing fetch, the same event Cloudflare prices.
- Pay per inference — payment each time the content is used in generating an answer.
- Pay per training — payment for inclusion in a training corpus.
The last two are what the crawl-level toll cannot reach, and they are the terms that matter once a page has been fetched once and reused indefinitely.
How it is published and enforced
Terms can travel in robots.txt, HTML, HTTP headers, RSS (via an RSL namespace), media files, and schema.org metadata — every place a crawler already looks. Enforcement runs through the RSL Open Licensing Protocol: the RSL document names a license server, and a system wanting the content queries that endpoint for a license before use, with payment handled server-side rather than by bilateral negotiation.
That is the design's ambition and its weak point. Publishing terms is unilateral and free; collecting on them requires the AI company to query the license server, which nothing in the standard compels. A robots.txt line has never bound anyone who chose to ignore it, and RSL inherits that.
Who has adopted it
The Collective reports 50+ partners and roughly 1,500 media organisations endorsing the standard, including People Inc, Yahoo and the Associated Press; Arena Group, BuzzFeed, USA Today Co and Vox Media joined later. The adoption figures are the Collective's own and count endorsements rather than paid licenses — no royalty totals have been published.
Why the collective matters more than the file format
A single site declaring "pay per inference" is asking to be ignored. The standard's theory is that a collective rights organization — the model music publishing has used for a century — negotiates once for many small publishers, which is the only way an independent site acquires any leverage at all. Whether that works is the open question in ai-content-monetization.
Related
ai-content-monetization · rsl-collective · pay-per-crawl · llms-txt · llmstxt-spec · wikimedia-enterprise · cloudflare-agentic-internet-report