Check Whether a Health Headline Matches the Study Behind It

Paste a headline and its study abstract to get a 1-5 accuracy score, a check table on study design and risk figures, and a plain-English rewrite.

You getVerdict, check table and what the headline gets wrong

Prompt

Prompt
You are a science journalist and research methods editor who fact-checks health headlines against the studies behind them.

I will paste a headline and the study's abstract right after this message. Judge how well the headline matches what the study found.

Rules:
1. For each check, quote the abstract wording you rely on. If the abstract does not state something, write "Not stated" instead of guessing.
2. Check five things: study type (observational, randomized trial, meta-analysis, lab), subjects (cells, animals or humans), sample size, absolute vs relative risk, and whether the headline claims cause when the study shows only association.
3. If only relative risk is reported, say so and explain what that means for a reader. Do not invent absolute numbers.
4. Rate the headline 1-5 (1 = misleading, 5 = accurate).
5. Keep the rewrite under 20 words, with no claim stronger than the abstract supports.

Output format:
- Verdict: score plus one-sentence reason
- Check table: Check | Headline implies | Abstract says | Match (Yes/Partly/No)
- What the headline gets wrong (2-4 bullets)
- Plain-English rewrite
- What to look for in the full paper (2-3 bullets)

End with one line on what an abstract alone cannot tell you. This is not medical advice; talk to a doctor before changing diet, medication or treatment based on a study.

Example output

Example for: the headline "Two Cups of Coffee a Day Cuts Liver Cancer Risk in Half" and a fictional cohort study abstract.

Verdict: 2/5. The study found an association in one observational cohort, but the headline states cause and presents a relative reduction as if it were a large personal benefit.

Check table
CheckHeadline impliesAbstract saysMatch
Study typeProven effect"prospective cohort study"No
SubjectsPeople"adults aged 40-69"Yes
Sample sizeNot mentioned"12,406 participants; 61 incident cases"Partly
Risk typeYour risk halves"hazard ratio 0.51 (95% CI 0.29-0.89)"No
Cause vs associationCoffee cuts risk"was associated with lower risk"No
What the headline gets wrong
  • "Cuts" implies cause. A cohort study tracks people who already drink coffee; it cannot rule out that coffee drinkers differ in other ways (alcohol use, liver health, income).
  • "In half" is relative risk. A hazard ratio of 0.51 halves a risk that was already small. With only 61 cases among 12,406 people, the absolute difference for an individual is likely a fraction of a percentage point. The abstract does not report absolute risk.
  • The estimate is uncertain. The confidence interval runs from 0.29 to 0.89, so the true effect could be much smaller than "half."
  • Few events. 61 cancer cases is a thin base for a firm conclusion.
Plain-English rewrite

Coffee drinkers had fewer liver cancers in one long-term study; it's unclear if coffee was the reason.

What to look for in the full paper
  • Which factors were adjusted for, especially alcohol, hepatitis B/C and body weight.
  • Absolute case rates in each coffee group, usually in a results table.
  • Whether earlier randomized or pooled studies show the same pattern.

An abstract cannot show how participants were selected, which confounders were missed, or who funded the work. This is not medical advice; ask a doctor before changing your diet based on one study.

Written by AI from this prompt. Your results will vary.

Tips and common edits

Why this works

  • Requiring a direct quote or "Not stated" for every check stops the model from filling gaps with plausible-sounding guesses.
  • The five named checks target the most common distortions: animal results, tiny samples, relative risk and implied causation.
  • Capping the rewrite at 20 words with no claim beyond the abstract forces a usable, honest alternative headline.

Use when a news story or social post cites a study. Have the headline and the full abstract (from PubMed or the journal page) ready to paste into ChatGPT, Claude or Gemini.

Common edits

  • Ask for a short reply you can post under a social media share of the article.
  • Paste the full methods section instead of the abstract for a deeper check on confounders and funding.

Make it repeatable

Keep using this prompt without pasting it in every time.

Prompt details

By
Prompts-GPT editors
Published
How prompts are written and checked

Did this prompt work?

Keep exploring

More in Research