Cognitive Sovereignty Self-Audit for Media and Publishing

1. When deciding which stories to assign, how much do you rely on AI engagement predictions versus your editors' news judgement?

AI engagement forecasts drive most story assignments. We assign stories that algorithms predict will perform well. We check AI predictions but editors often override them based on news value they see differently. Editorss assign stories based on news importance. We review AI data after decisions are made. Editorss have full authority over assignments. Engagement metrics play no role in editorial decisions.

2. How do your journalists research and draft stories when they use AI writing tools like ChatGPT or Claude?

Reporters use AI to do initial research and drafting. They edit and publish AI output with source verification. Reporters do their own research but use AI to draft sections, then rewrite substantially. Reporters complete full reporting and drafting. AI is used only for copy editing or formatting. Reporters work without AI writing tools. All research, reporting, and drafting is done by human journalists.

3. When your organisation uses the same AI tools as competitors, how much does your coverage resemble theirs?

Our story selection, angles, and explanations are often similar to other outlets using the same AI tools. We notice some overlap in how we cover certain topics because we use shared AI research tools. Our stories differ from competitors even when we cover the same events, due to different editorial choices. Our coverage is distinctly different from other outlets. We maintain separate editorial perspectives.

4. How do junior journalists develop investigative instincts when AI handles research and early drafting?

Junior journalists mainly learn to use AI tools and edit their output. They do less independent reporting. New reporters do some original reporting but rely on AI for background research they would have learned to do. Junior journalists do most of their own research with AI as a final reference check. We have a structured training programme where junior journalists complete full investigations without AI before using tools.

5. How do you know whether your readers can tell the difference between AI-assisted and human-reported stories?

We do not disclose AI use or test whether readers notice. We let AI-written content appear identical to human reporting. We label some AI-assisted stories but readers often assume all stories are human-written. We label all AI-assisted content clearly and have surveyed readers about their trust in those stories. We disclose AI use transparently. We only publish human-written investigative and opinion journalism.

6. When a breaking story requires fast decisions, who decides whether it is accurate enough to publish?

AI systems flag and prioritise breaking stories. Editorss publish based on AI confidence scores with minimal human review. AI systems gather facts and editors verify them before publication, but the process is rushed. Editorss make verification decisions for all breaking stories, using AI only for fact collection. A named editor personally verifies sources for every breaking story before publication, regardless of speed pressure.

7. How does your organisation know whether AI is biasing story selection toward particular viewpoints or demographics?

We do not audit AI systems for bias. We assume they are neutral because they are algorithmic. We occasionally review whether AI recommendations miss certain stories but do not formally test for bias. We review AI output quarterly to check whether it covers diverse perspectives and communities fairly. We conduct regular bias audits on all AI systems. We have public policies on what we found and how we corrected it.

8. When readers lose trust in your news, can you point to decisions that were made by your journalists versus decisions that came from AI?

We cannot clearly distinguish. AI and human decisions are so integrated that we cannot identify where trust broke down. We could trace some decisions back to editors or AI but our records are incomplete. We can identify whether AI or editors made most publishing decisions but we have not reviewed this systematically. We log which decisions came from editorial judgement versus AI systems. We review this regularly with staff and readers.

Cognitive Sovereignty Self-Audit for Media and Publishing

Cognitive Sovereignty: How To Think For Yourself When AI Thinks For You