An AI answer gives quitting advice: check the citation before trusting the claim

Separate the claim from the fluent explanation

Choose one factual statement that would affect your decision, rather than trying to verify every sentence at once. Write it in plain language and note the source the answer gives. A long bibliography or confident tone does not establish that the cited documents exist or contain the asserted finding.

WHO’s guidance on large AI models in health describes risks from inaccurate, false, biased, or incomplete outputs. It does not conclude that every AI answer is wrong or validate any particular tool. The reading exercise here addresses source checking; it does not test a model or turn an AI response into individualized clinical advice.

World Health Organization: WHO releases AI ethics and governance guidance for large multi-modal models

Find the actual document through an independent route

Use an official publisher, recognized public-health site, or legitimate research database to locate the citation. Compare the title, author information, date, and identifier with the answer. If a link leads to a different paper or a nonexistent page, mark the citation unresolved rather than assuming that a similar title is the intended study.

In a hypothetical example, an answer claims that a progress-tracking app has been proven to help adults quit vaping, but its link opens smoking research about another intervention. Even if the paper is genuine, it does not establish the app-specific claim. The mismatch is about evidence scope, not simply whether the webpage opens.

Read the passage supporting the specific statement

Find the result, recommendation, or limitation that the answer appears to use. Check whether it concerns vaping, smoking, a different population, or a different outcome. A paper discussing future research is not proof that the proposed intervention works. If only an abstract is available, keep that reading limitation visible.

Ask what the source does not establish: individual results, treatment suitability, a specific app function, or an exact recovery date. You can use a qualified professional to clarify a health decision while keeping an unsupported citation out of your plan. Do not provide private medical records to an unfamiliar AI service just to ask it to verify its own answer.

Write a conclusion proportionate to the check

Your note might say: “The source exists, but I have not found support for this app-specific statement.” If the passage does support a narrower claim, preserve that narrower wording and the source link. Avoid replacing the AI’s overstatement with an equally sweeping claim that no relevant evidence could ever exist.

The useful outcome is one verified or unresolved claim, not a score for the AI’s overall reliability. Repeat the process only for information that matters to your next decision. Practical preparations can continue while a citation remains open, but do not use uncertainty as permission to improvise medication instructions, diagnose symptoms, or promise a clinical benefit from tracking.