A coalition of major publishers just released a technical standard for measuring AI content use โ and the most consequential decision in it is a unit of measurement.
No AI company has accepted SPUR’s invitation. Google gave a non-answer and the others did not respond; the standard is voluntary and no adoption is established.A dating discrepancy is reported but not resolved: the v1.0 release object records 2 September 2026 for both creation and publication, while the trade report describes the standard as released on 2 October. Nothing here is investment advice.
SPUR โ the Standards for Publisher Usage Rights initiative โ released Content Telemetry 1.0, a technical specification for tracking what AI systems do with publisher content. The coalition’s members include the Guardian, the Financial Times, the BBC, Sky, The Times of London, MediaHaus and the AP. This publication read the v1.0 release directly from the coalition’s own repository.
The specification ships with 171 fixtures, 15 validated worked examples and 59 mutation checks, run in continuous integration on every pull request. The public comment period ran from 12 June to 24 July 2026 and produced 29 specification threads, 3 profile threads and 5 pull requests. Every thread carries a recorded outcome on the issue tracker.
Alongside the release, SPUR has invited OpenAI, Anthropic, Google, Meta and Microsoft to help shape implementation through a new, invitation-only AI Licensing Advisory Board. Technical lead Alex Springer expects the board to reach about 20 members, with a first meeting planned for October. No AI company has accepted the invitation. A Google spokesperson said only that the company frequently engages with SPUR and other associations on a variety of topics. The others did not respond before publication.
The key insight: The most consequential line in Content Telemetry 1.0 is about units. The release designates chars_ingested and excerpt_chars as primary โ described explicitly as portable counts โ and relegates token counts to supplementary. A character is a character. A token is whatever a vendor’s tokenizer says it is.
Publishers did not write a manifesto. They wrote schemas with conformance tests, which is what makes the unit choice enforceable.
The Structural Read
This is a measurement standard written by the measured, not the measurer. That distinction shapes every design choice in the document.
Start with the unit. Count AI content use in tokens and the counterparty’s tokenizer defines the quantity. The same passage of text yields a different token count depending on whose model reads it. Count in characters and that variable disappears. The release calls chars_ingested and excerpt_chars portable counts for exactly this reason, and pushes token counts to supplementary status. The publishers chose the unit that doesn’t move when the vendor changes.
The same logic runs through the purpose field, which replaces the old bot_category. The new values are training, inference, search and advertising. Traffic is classified by what it is for, not by what it calls itself. A bot can rename itself; a purpose classification forces a declaration of intent.
Coverage declarations work the same way. Publishers must now explicitly state whether their data is complete, sampled, aggregated or selected. That forces a statement about how partial any reported figure is โ before anyone else has to ask.
One change runs against the privacy grain and belongs in the record. ip_hash was withdrawn from the edge and origin profiles. The manifest identifies a domain rather than a legal person. That is a narrowing of the standard’s reach, not an expansion.
SPUR Co-founder David Buttle
“This will ensure both sides of the equation have a seat at the table and can work together to drive through better, fairer and more sustainable ways of working,”
The board has been invited. No one has sat down yet.
Permission Layer โ Applied
Publishers are writing the meter before regulators do
The Permission Layer framework tracks how governance infrastructure gets built around AI capability. Content Telemetry 1.0 is a private-sector attempt to establish the measurement layer before any regulator mandates one. The standard is voluntary. No enforcement mechanism is described in the release. Whether that voluntary architecture becomes the default depends entirely on whether AI companies engage โ and none have, yet.
Three Implications
MEASUREMENT DESIGN IS NEGOTIATION DESIGN Choosing characters over tokens is not a technical preference โ it is a structural choice about who controls the unit of account. When the unit is portable, neither party can move it unilaterally. That is a negotiating property built into the specification itself, before any deal is on the table. The standard is voluntary, but if it is adopted, the meter travels with it.
PURPOSE CLASSIFICATION CHANGES WHAT IS AUDITABLE Replacing bot_category with a purpose field โ with declared values of training , inference , search and advertising โ means a crawler must declare what it is doing, not just what it is. Combined with the agent_fetched , agent_cached and third_party_sourced distinctions, the spec creates a more granular audit surface than existed before. What any party does with that surface is a separate question.
THE INVITATION IS THE STORY, NOT THE STANDARD Content Telemetry 1.0 is a technically complete specification with continuous integration, worked examples and a resolved public comment period. That work is done. What is not done is any engagement from OpenAI, Anthropic, Google, Meta or Microsoft. Co-founder David Buttle says the aim is for both sides to have a seat at the table. As of publication, one side has built the table. The other has not pulled up a chair.
Business Engineer Framework
The Permission Layer
The Permission Layer framework maps how governance infrastructure forms around AI capability โ and who builds it first. Content Telemetry 1.0 is a case study in private-sector standard-setting: publishers defining the measurement architecture before any regulator mandates one. The Map of AI traces where plays like this sit in the broader stack โ and which layers remain ungoverned.
SPUR has shipped a technically rigorous, voluntary standard โ engineered in public, tested in CI, and built around a unit of measurement that no vendor can quietly redefine. The invitation to OpenAI, Anthropic, Google, Meta and Microsoft is real, the AI Licensing Advisory Board is real, and the first planned meeting is in October. What is also real: no acceptance, one non-answer from Google, and silence from the rest. The meter is ready. No one on the other side has agreed to be measured by it.
The specification details above were read directly by this publication from the SPUR Coalition’s own GitHub release for Content Telemetry 1.0. The reporting on the advisory board comes from Digiday’s account of 2 October 2026 by Sara Guaglione.No AI company has accepted SPUR’s invitation. Asked whether it had accepted, a Google spokesperson said only that the company frequently engages with SPUR and other associations on a variety of topics, which does not answer the question.
OpenAI, Anthropic, Meta and Microsoft did not respond before publication. The standard is voluntary, no enforcement mechanism is described, and no adoption by any AI company is established.A dating discrepancy is reported above and not resolved. The v1.0 release object in the coalition’s repository records both a creation time and a publication time on 2 September 2026, and is marked neither draft nor pre-release; the trade report describes the standard as released on 2 October.
Both are stated as they stand, and nothing above declares either correct or attributes error to anyone.Nothing above estimates what any licensing arrangement would be worth, what any vendor would owe, or what revenue the standard might produce. No such figure appears in either source.Also absent, because the sources do not give them: any adoption or commitment by an invited company, any licensing rate, any enforcement mechanism, the advisory board’s final membership, whether any invited company will attend the October meeting, and the contents of the full specification document beyond its release notes. Nothing above predicts anything and nothing here is investment advice.
Gennaro Cuofano is a CRO and tech executive who has worked in AI since late 2015, starting with bringing NLP, natural-language generation, voice and chatbot products to the marketing industry. His background is in law and finance: he holds a Master's degree in Law and an International MBA with an emphasis on corporate finance (LUISS Business School and the University of San Diego, 2012), and worked as a financial analyst at a real-estate investment firm in San Diego and as an assistant controller. His work focuses on business model strategy, business engineering and, more broadly, structural analysis: how companies make money, read from their own filings. He created FourWeekMBA and leads research there and at The Business Engineer, his newsletter on AI and business strategy, with over 95,000 subscribers and more than 1,000 published analyses. His writing has also appeared on Entrepreneur (for example, How to Digitize a Family Business in Europe) and HackerNoon. Find him on LinkedIn and Substack. See how we source.
Scroll to Top
Discover more from FourWeekMBA
Subscribe now to keep reading and get access to the full archive.
