TRANSCRIBE
GREATMachines eat repetitive production work.FIELD GUIDE · JOURNALISM, AUTOMATION & ACCOUNTABILITY
SHOULD AI BE ALLOWED TO WRITE THE NEWS?
AI can turn transcripts into summaries, suggest headlines, translate copy, and produce clean prose in seconds. None of that answers the old newsroom question: who actually knows this happened?
UPDATED 2026-09-22
THE PROBLEM
THE REPORTER WAS NEVER THERE
Beautiful copy. Excellent grammar. Now where did the fact come from?
News production includes reporting, sourcing, verification, judgment, writing, editing, translation, transcription, headlines, graphics, and distribution. Generative AI is much better at some of those tasks than others. Treating the whole bundle as 'writing the news' hides the difference between producing sentences and producing verified knowledge.
SUMMARIZE
CHECK ITCompression can introduce mistakes.REPORT
WHO CALLED?New facts require contact with reality.PUBLISH
YOUR NAMEAccountability does not autocomplete.THE FIGHT
IS WRITING THE NEWS THE SAME THING AS REPORTING IT?
Newsrooms face brutal economics and repetitive production work, so AI can buy back time for reporting. But journalism's scarce asset is not grammatical prose. It is trustworthy observation, source relationships, editorial judgment, and accountability when facts are disputed.
THE TOOL CASE
AI CAN REMOVE A LOT OF LOW-VALUE PRODUCTION WORK
AP and Reuters both permit bounded uses of AI for tasks such as research assistance, summaries, transcription, translation, headlines, and workflow support while requiring human verification and accountability.
The printing press changed production without becoming the reporter.
THE REPORTING CASE
THE NEWS PRODUCT FAILS IF SYNTHESIS OUTRUNS VERIFICATION
AI systems can generate plausible but wrong, outdated, or poorly sourced accounts of current events. Large international evaluations of AI news answers continue to find substantial accuracy and sourcing problems.
A perfect rewrite of a false rumor is still a false rumor.
THE WEIRD SHIT
THE PROSE CAN LOOK BETTER WHILE THE SOURCE CHAIN GETS WORSE
AP NOW FORMALLY ALLOWS BOUNDED AI ASSISTANCE
AP's 2026 standards allow AI for early research, document summarization, transcription, translation, headline suggestions, summaries, grammar, and search optimization, while requiring journalist review and preserving human responsibility for reporting and verification.
A major newsroom draws the line at accountability, not at typing.REUTERS USES THE SAME CORE RULE
Reuters standards embrace approved generative-AI assistance while requiring independent verification of AI-generated facts and maintaining journalist accountability for published work.
The machine can assist; the newsroom still owns truth claims.AI NEWS ANSWERS STILL HAVE SYSTEMIC ERROR PROBLEMS
A 2025 EBU-coordinated study had journalists evaluate more than 3,000 AI answers about news and current affairs; 45% contained at least one significant issue, including sourcing and accuracy failures.
Smooth current-events synthesis remains a risky substitute for verified reporting.AUDIENCES MAY NOT EASILY DISTINGUISH AI-WRITTEN COPY
A 2025 study of journalism students in Portugal and Spain found participants often rated AI-generated news prose highly and struggled to distinguish it from human-written news.
Readable prose does not reveal how the information was gathered.THE PEOPLE WITH A STAKE
WHO OWNS THE SENTENCE WHEN IT'S WRONG?
JOURNALISTS
They can spend less time on transcription and commodity rewrites, or see employers use the same tools to cut reporting capacity.
READERS
They may get faster, clearer, more accessible news or more synthetic-looking certainty detached from original reporting.
PUBLISHERS
They gain production efficiency while retaining legal, ethical, and reputational responsibility for every published claim.
SOURCES + SUBJECTS
AI-assisted mistakes can spread quickly and may be harder to trace when nobody is sure which human actually checked the statement.
THE UNANSWERED QUESTION
WHAT PART OF JOURNALISM ACTUALLY REQUIRES A JOURNALIST?
- Newsroom studies showing AI assistance consistently increases original reporting time rather than merely reducing headcount and increasing output volume.
- Reliable source-tracing systems that make every generated factual claim auditable back to primary reporting or documents.
- Clear disclosure norms for materially generated stories that remain understandable to readers without turning routine assistive use into noise.
- Evidence that AI-only news production can independently gather, verify, challenge, correct, and take responsibility for contested facts rather than only synthesize them.
TAKE THIS TO DINNER: The machine can help write a story. It cannot be the person who knows why the story is true.
RECEIPTS
Newsrooms are already using AI. Their own standards reveal where they think the human job still begins.