Dispatch 31
Journalism costs money before anyone reads it. Reporters gather evidence, editors check it and newsrooms maintain the capacity to cover what happens next. An AI service can draw value from that work long before it displays a quotation or sends a reader to the original article.
That is the business question behind SPUR’s new content reporting standard: what should publishers count, what can they check, and which agreements can sustain the work? Finland gives the question immediate weight, as the Finnish news agency STT considers ending its current news and photo operations. The value of reliable information and the ability to finance its production deserve equal attention.
This edition examines both. It also considers copyright, the reliability of AI answers and the practical choices that help newsrooms use AI responsibly.
I. On the value of journalism in AI
Measurement helps a negotiation. The contract determines what the supplier earns.
The Spur Coalition- logo
31.1 — SPUR gives publishers a common language for use
SPUR has published Content Telemetry 1.0, a shared standard for usage reporting. It distinguishes when a system retrieves content, uses it to support an answer, cites it, presents it or records engagement with it, while implementation and audit work continue. Its scope document defines the limits: a usage record alone establishes neither permission nor price, and checking its accuracy and completeness requires separate work. The specification covers use when producing answers; it does by itself prove how a company uses material for model training.
The need for telemetry becomes tangible in Google’s AI contribution pilot, where participating publishers receive a monthly earnings figure with limited detail about its calculation. Publishers need enough information to reconcile an invoice with the material supplied and the rights granted. A contract should specify which activities count, who records them and how each party can challenge the figures.
Cloudflare illustrates two commercial choices: its Monetization Gateway lets sellers price access requests, including APIs and tools, while Pay Per Use lets buyers define qualifying uses and prices, then report that use. Publishers can accept or decline those offers, but their negotiating room depends on the reporting and payment terms. SPUR can help parties describe what happened; they still need to agree what it is worth.
Reflections
Which uses of your journalism should a commercial agreement recognise? What evidence would let you check the customer’s account of those uses?
31.2 — Publishers need contracts that finance production
Axios is preparing direct feeds for institutional customers and AI systems, with annual pricing for its first offering and a personal AI agent feed at an earlier stage. The New York Times sets three tests for partnerships: a fair exchange of value, sustainable collaboration and control over content use. In media monitoring, PMG reports uptake of an AI rights extension and validation logo, a useful case whose precise protections depend on the licence terms.
A publisher can offer a professional machine service alongside its human website: an identified customer, an agreed collection, defined permissions and a predictable bill. Access to a reliable, regularly updated collection has economic value from the first intake into the customer’s system. My view is that small payments for selected uses alone offer too little certainty to finance a newsroom’s continuing costs; a collection licence or minimum commitment gives both parties a clearer basis for planning.
Yahoo’s Jim Lanzone also questions whether micropayments can adequately compensate publishers, while Florent Daudens describes an agent buying access in a small test. The experiment demonstrates a possible transaction with test money, leaving commercial demand and scale to establish. Publishers should judge these models by their contribution to production costs, their protection of existing customer relationships and the control they retain over reuse.
Reflections
What payment structure recognises the value of keeping a trusted collection available? How would you test whether an AI deal adds income or substitutes for an existing product?
31.3 — Newsrooms make origin easier to check
Ukraine’s Suspilne has integrated Content Credentials into its publishing system, with support from BBC Media Action. The implementation checks credentials from compatible cameras and preserves origin information through edits such as cropping. This gives customers a practical way to examine where material came from and how the newsroom handled it.
Other suppliers are at different stages: Fotoware has launched C2PA support, Reuters and Sony demonstrated a video workflow, and AFP and Dalet plan a video proof of concept for early 2027. The Bergen Media Provenance Summit shows how widely the sector is working on the problem. Each deployment still needs to prove its usefulness in everyday production.
Origin, accuracy and permission remain separate checks. As Microsoft’s documentation explains, provenance signals establish neither truth nor ownership. A valuable news service combines identifiable material with editorial verification and clear contractual rights.
Reflections
Which information about origin would help your customers make better decisions? How will you explain the limits of a credential alongside its benefits?
II. On rights and bargaining power
Legal boundaries and commercial agreements give publishers different, complementary forms of certainty.
31.4 — Copyright gives negotiations a legal basis
The European Commission is consulting on technology’s effects on copyright, including AI, until 3 November. In the United States, an appeals court upheld Thomson Reuters’ copyright victory against Ross Intelligence. Reuters reports that the court’s reasoning remains sealed and that the case concerns a legal search tool using non-generative AI, which limits broader conclusions.
Publishers also need practical ways to exercise their choices. More than 300 media executives are supporting proposed US legislation requiring bots to identify themselves and their purpose. Such identification would help a publisher distinguish customers, search services and other automated visitors before deciding what access to allow.
For European news businesses, copyright and the CDSM framework are the legal reference points. Courts clarify disputed boundaries, negotiations establish workable prices, and technical controls help parties apply their agreements.
Reflections
Where does your ability to express a preference exceed your ability to enforce it? Which legal and practical safeguards would make a licensing negotiation more balanced?
III. On keeping news production viable
Finland deserves particular attention. The consequences of losing a news agency extend across national and international organisations that rely on it.
31.5 — STT puts the cost of shared reporting into focus
STT has begun restructuring negotiations affecting 95 employees, after prolonged losses and unsuccessful efforts to secure a sufficiently broad group of committed owners. The European Federation of Journalists reports that current news and photo operations could end on 31 March 2027, while a smaller operation remains possible. The Finnish government has paused preparations for state support pending the owners’ decision, so the public record still describes a threatened closure. STT’s profitable press release distribution business falls outside the restructuring plan.
The distinction matters because distributing a company’s message and producing independent reporting serve different purposes. The EFJ warns that smaller outlets would face higher costs to replace STT’s services. A shared newsroom lets several customers finance independent reporting that each would struggle to reproduce alone. When a customer assesses its subscription solely against its own immediate needs, it underestimates the collective cost of losing that capacity.
The response needs durable commitments from owners and customers, with clear safeguards for editorial independence in any public support.
EANA’s World News Day statement links trustworthy journalism to rigour, independence and accountability; financing those qualities requires decisions as concrete as the values themselves. I will bring this perspective to next week’s MINDS meeting in Stockholm, where we will discuss STT and a possible common position.
Dispatch 32 will return to that discussion: for now, Finland gives every news organisation reason to examine which shared reporting services its helps sustain.
Reflections
How should customers account for the reporting capacity they collectively depend on? What combination of ownership, commercial commitments and independent governance could sustain it?
IV. On trust in the answer
A reader needs evidence that supports the claim. A recognisable source name is only part of that assessment.
31.6 — Researchers test what AI citations actually support
A study of Google AI Overviews examined more than 55,000 searches and compared claims with their cited pages. The researchers found stronger credibility scores for cited domains alongside gaps in support for individual claims. Their university’s account notes important limits, including incomplete page capture and information that changed during collection.
Verification therefore requires checking the relationship between an answer and its evidence. CJR’s Ask Anika column examines efforts to influence the material chatbots may encounter, adding a separate concern about deliberate manipulation. Publishers and AI services need checks on both source selection and the claims an answer attributes to those sources.
Euronews’ examination of fabricated German election videos offers a concrete example of verification through ballot details and election officials. This is the work that World News Day asks audiences to value. A useful AI product should make that evidence easier to inspect and corrections easier to apply.
Reflections
How would you assess whether an answer accurately represents the sources it cites? What should happen when a publisher corrects material that an AI service has already used?
V. On making AI work in the newsroom
Responsible adoption depends on management choices. Teams need time, skills and clear authority to make those choices well.
31.7 — Media organisations organise responsibility for AI
The J·Index database documents publicly reported AI adoption across news organisations, offering examples to examine within the limits of public disclosure. A survey of AI for Media Network members records broad use alongside concerns about skills and dependence on technology suppliers. Its respondents also express strong interest in cooperation with other media and researchers.
Teams can give that cooperation a practical form. DPG Media and Utrecht University’s DEDA for Media method brings relevant colleagues together to examine ethical questions in data and AI projects. Such a review gives editors, product teams and legal colleagues a shared opportunity to assess benefits, risks and responsibility.
At APA, Maria Scholl’s confirmation as chief executive accompanies plans to develop its AI strategy. APA’s strategy illustrates the growing importance of AI for news agencies, both in producing journalism and in supplying reliable information to customers’ AI systems.
Reflections
Which decisions about AI belong with individual teams, and which require common rules? How will you measure the effect on journalistic quality alongside time saved?
31.8 — Local needs shape useful collaboration
Research across Pacific newsrooms reports gaps in training, policy and support for local languages. Media leaders meeting in Cape Town called for cooperation, stronger ownership of intellectual property and better ways to earn revenue from audiences. These are distinct markets with shared questions about resources and control.
A newsroom needs to test tools against the languages, subjects and communities it serves. Shared purchasing or training can reduce costs, provided participants retain the ability to set their own editorial standards. Support from a technology supplier also deserves clear terms for data, future costs and continued access.
At the Ibero-American journalism and AI forum in Valencia, speakers likewise emphasised governance, training and human oversight. Cooperation becomes useful when participants can apply it to a specific production problem. Its value depends on what each newsroom can do better afterwards.
Reflections
Which capabilities could your organisation build more effectively with others? What must remain under local editorial control for that cooperation to serve your audience?
Conclusion
A newsroom’s budget includes the cost of being ready: maintaining expertise, checking an uncertain claim and assigning a reporter before anyone knows how widely the story will matter. That capacity gives a collection of journalism much of its value. An AI licensing agreement should recognise the work that makes reliable answers possible.
The commercial task is to agree permissions, evidence and payments that support continuing production. Publishers need enough visibility to check the agreement and enough predictable income to plan their work. Replacing a shared reporting service would require others to commit people, money and time.
For news organisations, regulators and AI companies, a workable agreement needs three documents beside one another: the licence, the usage record and the newsroom budget. Together they connect permission, accountability and the cost of producing reliable information. That is the basis for a sustainable relationship between those who make journalism and those who use it.