Artificial intelligence is now part of the daily workflow of newsrooms around the world, including our own. As a publication that covers the impact of AI on journalism, we have an obligation that goes beyond commentary: we must demonstrate, through our own practice, what responsible AI-assisted reporting looks like. This page sets out the standards that govern every piece of analysis and news we publish at Editors Weblog.

What AI tools we use, and how

We use large language model assistants as writing aids. Specifically, they help with drafting structures, suggesting headline alternatives, checking logical flow in argument-heavy analysis pieces, and surfacing relevant regulatory text for cross-reference. We do not use AI to generate original reporting, to produce fabricated quotations, or to manufacture statistics. Every factual claim in our articles is verified by a human editor against a named, traceable source before publication.

We do not use AI image generators for editorial photography. Where illustrations are AI-generated, we label them as such in the caption. Our coverage of the AI disclosure debate reflects the same transparency standards we apply to ourselves.

Human oversight is non-negotiable

Every article published under the Editors Weblog masthead is written, edited, and approved by a named human journalist or editor. AI output is treated as a first-draft resource, not a publishable product. Our editorial chain requires that a senior editor reviews all AI-assisted copy before it reaches the CMS. The responsibility for accuracy, fairness, and editorial judgment rests with people, not tools.

This is not simply a matter of internal policy preference. The EU AI Act imposes transparency obligations on publishers using AI systems to produce content that reaches the public. We treat compliance with those obligations as a floor, not a ceiling.

Accuracy and source integrity

We do not invent original research, run benchmarks, or conduct tests that we have not actually carried out. We do not attribute findings to unnamed internal sources or present AI-generated summaries as if they were our own empirical data. Every statistic we cite comes from a named study, a named organisation, or a named regulatory document. If we cannot verify a claim to that standard, we omit it or describe it in general terms, with that limitation made clear to the reader.

This commitment matters especially when covering AI tools themselves, a beat where vendor claims are often unverified and hype cycles move faster than peer review. Our readers rely on us to hold the line on evidentiary standards even as the technology moves quickly. For a deeper look at how we and other outlets approach this, see our analysis of synthetic media ethics and editorial guidelines.

Disclosure to readers

When AI tools have played a meaningful role in the production of a specific article, we note that in a short disclosure line at the foot of the piece. We define "meaningful" as any instance where AI-generated text has been incorporated into the final draft, even in heavily edited form. We do not consider AI-assisted headline generation or structural suggestions to require article-level disclosure, but we document those uses internally.

Our disclosure language is drawn from the framework being debated and adopted across the industry. Newsrooms looking for practical templates may find our collection of newsroom AI policy templates useful as a starting point for their own standards documents.

What we will not do

  • Publish AI-generated content without human review and approval.
  • Use AI to fabricate sources, quotes, or data points.
  • Present AI-generated analysis as original empirical research conducted by our staff.
  • Allow commercial relationships with AI vendors to influence editorial coverage of those vendors.
  • Suppress corrections when AI-assisted drafts have introduced errors that reached publication.

How these standards will evolve

The regulatory environment is shifting. The current status of the EU AI Act and its obligations for media organisations continues to develop, and our standards will be updated to reflect changes in law, platform behaviour, and professional consensus. We commit to reviewing this page at least twice a year and to noting the date of each revision at the foot of this document.

We recognise that publishing editorial standards is only meaningful if those standards are enforced. Readers who believe we have fallen short of what is stated here are encouraged to contact our editorial desk. We will respond to all substantive challenges and publish corrections prominently.

Sources

  • European Parliament and Council, Regulation (EU) 2024/1689 (EU AI Act), Official Journal of the European Union.
  • World Editors Forum, "AI in the Newsroom" guidance series, WAN-IFRA.
  • Reuters Institute for the Study of Journalism, Digital News Report, University of Oxford.
  • Associated Press, "AP Stylebook" guidance on AI-generated content and disclosure, The Associated Press.