This turns Claude into a fast research assistant that does four things in order: finds whether a real published study exists, pulls the actual numbers (sample, who was studied, how long), tells you if it is correlation or genuine cause, and names what the study does NOT prove. That last part is the gold, because the viral version almost always overreaches past what the researchers actually found.
This only works if Claude can actually search the live web, otherwise it answers from memory and can invent a plausible-sounding source. Web search is free on every Claude plan, including the free one, since 27 May 2025 (Anthropic announcement).
🔧 Open a new chat at claude.ai, click the slider icon in the message box, and switch Web search on (how to enable it). When Claude uses the web it adds direct source links under its answer, so you can click straight through to verify.
Run these one after another in the same chat so Claude keeps the context. Swap the example claim for whatever you are checking.
Step 1: does a real study even exist
Search the web and tell me: is there a real, published, peer-reviewed study behind this claim? "A study tracked 5,000 people for five years and found your partner's personality predicts your salary more than your boss does." Give me the actual study title, the authors, the journal, and the year, with a link. If you cannot find a real source, say so plainly instead of guessing.
Step 2: pull the real numbers
For that study, tell me the exact sample size, who was actually studied (ages, relationship status, country), and how long it ran. Then compare those real details to how the viral post described it, and flag anything the post got wrong or exaggerated.
Step 3: correlation or cause
Does this study show a correlation or did it prove cause and effect? Explain in plain terms which one it is and why that matters for how confidently I can repeat the claim.
Step 4: what it does NOT prove
List what this study does NOT prove, and where the viral version overreaches beyond the actual finding. Give me one honest sentence I could post that stays true to what the research really showed.
The reel's example is a genuinely real study, which is what makes it a perfect teaching case: the study is real, but the viral retelling bends it.
📚 The real study is "The Long Reach of One's Spouse" by Brittany Solomon and Joshua Jackson, published in Psychological Science in 2014 (study page, WashU write-up). It followed nearly 5,000 married people aged 19 to 89 over five years, with both partners working in about 75% of couples.
✅ What it actually found: people with a more conscientious partner reported higher job satisfaction, higher income, and more promotions, even after accounting for their own conscientiousness.
⚠️ Where the viral version overreaches: it is a correlation, not proof of cause. The researchers proposed three theories for WHY (a conscientious partner does more at home, models good habits, and lowers stress), but they call these explanations, not proven mechanisms (Scientific American). So "choose your partner like your salary depends on it" is a fun hook, but the honest version is "people with conscientious partners tended to do a bit better at work over time, and we are not certain exactly why."
🕵️ Spot the funding bias:
Who funded this study, and does the funder have a stake in the result? Flag any conflict of interest I should know before I share it.
📉 Check if it replicated:
Has this finding been repeated by other independent studies, or is it a single study that never replicated? A one-off result is much weaker than a repeated one.
🎙️ Trace the telephone game:
Find the original study, then show me how the headline and the social media version each changed the claim. Where did the exaggeration creep in?