“In a recurring theme, we see that LLMs don’t know what they don’t know; they work decently well with the information they’ve got but don’t know how to conduct clinical interviews, and in the hands of the lay public can easily give bad advice because the proper questions never get asked, either by the patient or the LLMs.”
Our reading:
LLMs don't know what they don't know: they work decently well with the information they've got but don't know how to conduct clinical interviews, and in the hands of the lay public they can easily give bad advice because the proper questions never get asked, either by the patient or the LLM.
Assessment: Can’t be checked
What happened
Not graded on evidence
Outcome: Can’t be checked
An opinion, a recommendation, or too vague to test. Listed, not scored.
Evidence up to . Checked .
Why
Marcus offers a qualified warning and a proposed cause. The studies provide relevant context: performance depends on the interaction setting and the clinical reasoning task. But grading the whole statement would require turning 'can easily' into a frequency claim or treating the interview language as a universal assertion. Neither is supplied by the text. Under the rule for hedged claims, this should not receive a supported or contradicted grade. That decision does not establish that the warning is wrong.
Caveats
- The Nature page reader exposed only the abstract despite marking the response FULL TEXT.
- The assessment does not separately grade the categorical fragments after removing their surrounding qualifications.
- There is no defined binary outcome to mark as known at the evaluation date.
Evidence (2 items)
Claims like this aren’t scored, so there is no evidence to show.
How we judged it
- Kind of claim
- Causal claim
- Deadline
- None given. We don’t add one.
Conditions
- in the hands of the lay public (their words)
- work decently well with the information they've got (their words)
- can easily give bad advice (their words)
Where we had to interpret
- Read in context, this is a qualified warning about possible bad advice and a proposed explanation involving missing questions.
- I do not convert 'can easily' into a claim that this happens typically, at a particular frequency, or across all models.
- I do not read 'never get asked' as asserting that no LLM ever asks an appropriate question. It describes the proposed failure pathway.
- The statement combines categorical descriptions of interview ability with qualified possibility language. Scoring the whole statement would require choosing a scope and threshold that Marcus did not supply.
We don’t grade claims like this. We only check the quote and who said it.
What would settle it
A separately specified claim naming the models, clinical setting, meaning of 'easily', and causal comparison could be assessed. Adding those requirements here would change the original claim.
We found 7 claims in this post. See them all.
Full record
Source and dates
Marcus on AI, Please don’t trust your chatbot for medical advice, paragraph 10.
- Said
- We read the source
- live page, full text, public
- Evidence up to
- Checked
We read this wording on 2026-10-28. The post may have been edited after 2026-04-21.
Where it was said (1)
Said once in the posts we read.
: Marcus on AI, Please don’t trust your chatbot for medical advice
In a recurring theme, we see that LLMs don’t know what they don’t know; they work decently well with the information they’ve got but don’t know how to conduct clinical interviews, and in the hands of the lay public can easily give bad advice because the proper questions never get asked, either by the patient or the LLMs.
- Where
- paragraph 10
- Published
- 2026-04-21
- We read it
- , live page
- Access
- full text, public
History
Versions
- Current: Can’t be checked. Evidence up to 2026-10-28.
What happened to this record
PublishedBy us
Published after both AI reviews and the final check passed.
Held backBy us
A reviewer said our test wasn’t fair to what was actually said.