Publisher AI standards coalition SPUR has shared details of a proposed “common language” for tracking content usage by AI companies.
SPUR aims to come up with a standard technical foundation for how AI platforms report on use of the content they scrape. This could be used by publishers when agreeing licensing deals.
SPUR was launched at the start of the year by The Guardian, the Financial Times, BBC, Sky News and The Telegraph and has since added more than 20 other publisher members.
Financial Times chief executive Jon Slade said: “Tracking how AI systems use content creates benefits for everyone involved. A common standard means AI systems and their users benefit from better, more relevant outputs.
“For content creators, seeing how and where their work brings value within the AI environment helps them understand where to focus energy and resources for that audience.
“It’s a virtuous circle that we’ve seen work time and again across the industry: when publishers know what resonates with a particular audience, they produce better work and everyone up and down the chain benefits.”
The draft “signal format for content usage reporting” has been published so that feedback can be shared.
It tracks content through five stages:
– Retrieval from a website (content owners can already see this stage via bot activity)
– Grounding (content being loaded into an AI agent’s generation context)
– Citing (content being explicitly referenced in an AI response)
– Display (a user actually seeing the content via a reference or content embedded in an AI answer)
– Engagement (whether a user clicked, copied, shared or directed an AI agent to act)
Also published is a SPUR telemetry profile which sets out the proposed terms AI companies would need to meet to be compliant with the standards.
These include real-time delivery (sharing those five event stages as they happen), with even-level delivery and information routed to a destination chosen by the publisher.
Alex Springer, technical lead for the SPUR Coalition who is welcoming feedback via alex@spurcoalition.org until 10 July, said: “Our aim is to develop a common language for measuring and reporting on content usage, one that licensing and compensation frameworks can build on.
“We want the draft pulled apart before anything hardens, which is why the comment window is open to everyone – publishers, platforms, model labs and developers alike.”
Stephen Jones, chief executive at Maro which fed into the standards in terms of what would be needed to link up to publishers’ other analytics so the AI data is useful to them, said the draft announcement is a “major step forward”.
He said “proper, standardised measurement of how AI platforms use publishers’ content” and transparency from those platforms about how it is used are both “minimum requirements if there’s to be a sustainable long-term future in AI licensing for publishers. But they won’t just happen by themselves.”
Email pged@pressgazette.co.uk to point out mistakes, provide story tips or send in a letter for publication on our "Letters Page" blog
