More than 80 publishers in over 20 countries have agreed on a common way to count what AI systems take from their work. On 13 October 2026 The New York Times, Agence France-Presse, The Economist and The Atlantic joined the coalition behind it. None of this covers video creators yet — but the problem it attacks is exactly yours: an AI answer repeats what you made, and the view never reaches you.
What SPUR is and who is in it
SPUR stands for Standards for Publisher Usage Rights. The coalition launched in February 2026 in the UK and is funded entirely by its members — no AI developer pays into it.
Founding members include the BBC, the Financial Times, Guardian Media Group, Sky News, Telegraph Media Group and Mediahuis. The Associated Press joined as a founder on 9 July 2026. After the 13 October intake — the NYT, AFP, The Economist and The Atlantic — the group passed 80 members across more than 20 countries.
Their goal is deliberately commercial: make AI usage of published material visible, countable, and therefore licensable.
Content Telemetry: the five events now being counted
On 2 October 2026 the coalition shipped version 1.0 of the Content Telemetry standard. It is an open technical specification, published on GitHub under the Apache 2.0 licence, and anyone can read or implement it.
It defines five points in the life of a piece of content.
| Event | What actually happened |
|---|---|
| Retrieved | A crawler fetched the page over HTTP |
| Grounded | The text was loaded into the model's generation context |
| Cited | The answer explicitly references the source |
| Presented | The citation or snippet was made visible to a real person |
| Engaged | The person clicked, copied, shared, or told the agent to act on it |
The value for a publisher is in the gaps between those events. Content that is grounded but never cited shaped the answer while earning no attribution and no visit. That gap is what nobody could measure before.
Two design choices worth knowing
First, reporting is post-hoc, not pre-declared: an agent reports what it did, not what it intends to do. Second, the standard only describes observable boundaries — fetch, cite, display. It deliberately stays out of re-ranking, reasoning chains and anything else inside the model.
Events are bundled into sessions with a shared session ID. Each one carries a timestamp, a content identifier, character and token counts, the citation type and the presentation format.
Why publishers moved now
Search traffic. The coalition's own press release cites a 33% drop in organic Google referrals to more than 2,500 sites worldwide between November 2024 and November 2025. The search engine increasingly answers on the page, and the reader never arrives.
Suing one company at a time is slow — the NYT's case against OpenAI and Microsoft has been running for years. A shared standard is the other route: measure the usage first, sell the licence second.
Will the AI companies play along?
That is the open question. SPUR invited OpenAI, Google, Anthropic, Meta and Microsoft onto an advisory board that is meant to seat around 20 members. According to Digiday, none had publicly confirmed; Google said only that it "frequently engages with SPUR and other associations on a variety of topics".
There is also a verification hole. Version 1 has no mechanism to confirm that a report is accurate — the agent reports on itself. Auditing and verifiable credentials are deferred to later versions, and Digiday reported that the first technical committee meeting on exactly that was set for 15 October 2026.
What this means for a short-form creator
Right now, nothing directly. The standard targets publisher web pages, no short-video platform is in the coalition, and an individual creator has no seat in it.
What matters is the precedent. The largest media companies in the world did not ask for a ban — they asked for a meter. Count the usage, then charge for it. Video is heading for the same fork: AI search already summarises clips inside the answer, and the visit to your channel is optional. We looked at that in how AI answers cite YouTube in shopping queries.
What you can do this week
- Check your AI training setting on YouTube. In YouTube Studio under Settings → Channel → Advanced settings there is a "Third-party training" control. It is off by default; you can allow everyone or pick individual companies from the list.
- Watch your traffic sources. The external-sources breakdown in your analytics is where a referral decline shows up first — months before it shows up in your income.
- Make things that do not survive a summary. A face, a voice, a real hands-on demo. Advice that fits in one paragraph gets absorbed into an AI answer without you; a demonstration does not.
- Sign your work. Handle on screen, a consistent thumbnail style, a mention in the first seconds. That is attribution that survives a re-upload.
- Keep project files and dates. Your edit project and first publication date are the proof the footage is yours. Platforms run their own version of this check — see how YouTube demotes reused footage in Shorts.
- Brands: put usage in the brief. Where the video may run, for how long, and whether it may be fed to model training — that belongs in the campaign terms, not in an argument afterwards.
What is still unknown
- Whether AI companies will report at all. The standard is voluntary and nobody can be compelled to adopt it.
- How a report gets verified. Version 1 has no verification; it is deferred.
- Whether money reaches individual creators. Licensing is being negotiated between publishers and platforms, and the person who filmed the clip is not at that table.
- Whether anything similar will exist for video and audio. No short-video platform has announced a comparable scheme.
FAQ
What is the SPUR coalition?
A group of large publishers building shared rules and technical plumbing so they can see how AI systems use their journalism and charge for it.
What is the Content Telemetry standard?
An open reporting format: an AI agent tells the publisher that it fetched an item, grounded an answer on it, cited it, showed it to a person, and that the person engaged with it. Version 1.0 was published on 2 October 2026.
Does it protect an individual creator?
No. It is built for publisher websites and only works between those publishers and the AI companies that agree to report.
Did Google traffic really fall 33%?
That figure comes from the coalition's press release: organic Google referrals to more than 2,500 sites fell 33% between November 2024 and November 2025. It is an average across that sample, not a per-site guarantee.
Can I stop AI from using my content?
On your own site, partly — AI crawlers are blocked through robots.txt, though not all of them honour it. On someone else's platform you only get the switches that platform offers, such as the third-party training setting in YouTube Studio.
Related reading
- AI answers cite YouTube in shopping queries: what it means for creators
- YouTube Shorts and reused footage: how clippers should adapt
- How to check if an image is AI-generated: what SynthID does
While the industry argues about metering, the clearest way to get paid for views is still a brief with a published rate. On Prime Oracles a brand sets one up through "Create campaign", and a creator opens the "Content rewards" section, hits "Take on task" and sees exactly what a thousand views pays.