Investigate a health headline
What kind of study sits behind a confident headline?
25-40 minutes
Start with these sources
Your first move: Search sleep exercise in PubMed and choose two relevant publications with different study types.
Without AI
- Record each publication’s question, participants, study type, and measured outcome from the available text.
- Mark what you could not establish from the abstract and inspect the full methods when accessible.
- Write a headline for each that matches the evidence and names a key limitation.
With GenAI
Bring your own AI tool and the source material. This atlas does not send your data anywhere or run an AI model.
- Supply the two abstracts, identifiers, and any methods excerpts you can access to your AI tool.
- Ask for a comparison of study design and a cautious headline for each, with evidence locations.
- Return to both records to verify the classifications and remove advice or causal claims the methods do not support.
A prompt to make your own
Compare only the supplied biomedical records and excerpts. Make a table of research question, study design, participants, outcome, and limitations. Cite the supplied PMID or URL and relevant excerpt for each claim. If design or a number cannot be established, say missing or unclear; do not fabricate numbers. Propose a headline faithful to each study. Do not offer diagnosis, treatment, or personal medical advice, and do not turn an association into causation. Records and excerpts: [Paste here.]
Make something: A study-comparison card with a careful rewritten headline and links to the records.
Keep your footing
- A systematic review and an individual trial answer questions differently.
- An abstract may not contain the methods needed to judge a claim.
- Check correction and retraction notices and the population studied.
Why follow this question?
Why this is interesting
A headline has little room for uncertainty; research needs room for methods, populations, and limits. Reading two kinds of study side by side makes that compression visible. The aim is to become a better reader, not to decide what someone should do about their health.
The PubMed user guide explains publication types and routes to full text. If you cannot read the methods, your comparison can still be useful when it names that limit. “Not established from the available text” is a result worth keeping.
Follow another thread
Follow a reference from a review to an included study. Compare the review's description with the original record and note what context was compressed. You can then carry a DOI into OpenAlex to explore the wider citation trail while keeping questions about research quality separate from citation counts.
Go a little further
Compare your rewritten headline with an actual news headline and underline the words that change its meaning.