Quick Fact-Check Assistant.
A 30-second accuracy read on any short piece โ every claim flagged ๐ข/๐ /๐ด with the search term to confirm it.
Example result
๐ Needs checking
โ ๏ธ "Email marketing returns $42 for every $1 spent" โ this figure comes from older Litmus/DMA reports and gets repeated everywhere; newer estimates are lower and vary by source (search: "email marketing ROI $42 latest study").
โ "ChatGPT launched in November 2022" โ correct, OpenAI released it on 30 November 2022 (search: "ChatGPT launch date").
โ ๏ธ "The average person sees 5,000 ads a day" โ widely quoted but the number is disputed and often traced to a marketing exec's guess, not a study (search: "average ads seen per day source").
โ "Instagram is the most-used social platform in the world" โ Facebook and YouTube report more monthly users; Instagram isn't first (search: "most used social media platform monthly active users").
โ ๏ธ "68% of consumers say they'd pay more for good customer service" โ plausible but source unclear; different surveys give different numbers (search: "consumers pay more good customer service statistic").
โ "Mailchimp was founded in 2001" โ correct, founded by Ben Chestnut and Dan Kurzius (search: "Mailchimp founded year").
โ ๏ธ "Studies show colour increases brand recognition by 80%" โ this stat floats around unsourced and is often misattributed (search: "colour increases brand recognition 80% source").
Unverifiable: "Our competitor lost 30% of their subscribers last quarter" โ internal or private claim, no public data to confirm.
Bottom line: two solid, one real error (the Instagram claim), and a pile of stats that need a source before you ship. Do this โ swap every "studies show" for the actual study, or cut it.
What this prompt does
- Flags every empirical claim ๐ข/๐ /๐ด in one pass
- Hands you a search term to confirm each
- Marks unverifiable claims โ never guesses or invents sources
- Catches accuracy errors before anyone else sees them
Why it works
Most 'AI slop' complaints are really accuracy complaints. This catches the obvious errors in one pass, no setup โ before anyone else sees them.
Tips for this prompt
- Paste your near-final draft โ not a rough first pass.
- Search every ๐ด and ๐ term before you ship.
- Feed it stats-heavy copy โ that's where it earns its keep.
- Add "flag vague sourcing too" for numbers with no source.
- Trust "Unverifiable" โ that's the point, not a miss.
How to use the prompt
- 1Paste the prompt into your AI assistant.
- 2Drop your text into the {{PASTE YOUR TEXT HERE}} slot.
- 3Run it.
- 4Read the verdict โ ๐ข, ๐ , or ๐ด.
- 5Search each flagged claim's term.
- 6Fix what needs fixing before you ship.
More like thisEditing prompts
Browse all โ