Library › Marketing AI › Marketing and content

Catch the unsupported claims in an AI-written blog draft

Practice: Catching mistakes and made-up facts. About 5 minutes.

About this exercise

A freelancer used AI to draft a blog intro for Quillpad. Before it goes live, you need a numbered list of every claim in the draft that the official fact sheet doesn't support, with one line per problem: the claim, then what the fact sheet actually says (or "not in fact sheet"). Don't flag claims that check out, and don't rewrite the draft. Keep it under 100 words, plain text. Write a prompt that gets that list.

Common mistakes when you ask AI for this

  • Saying "check this blog post". The AI gives writing feedback on tone and flow, and may rewrite the whole thing, instead of checking facts. Try: Be specific: "List only the claims the fact sheet doesn't support."
  • Not naming the source of truth. The AI checks claims against what it thinks it knows about apps like this, which can't tell it anything about Quillpad. Try: Add: "Treat the fact sheet as the only source of truth."
  • Flagging everything. To look thorough, the AI flags claims that are fine, like the $8 price, so you waste time on non-problems. Try: Add: "Do not list claims that match the fact sheet."
  • Trusting the AI to count words. AI is bad at counting. Ask for "under 100 words" and it often lands well over. Try: Cap each part instead, like "each line under 18 words," so the total lands well under 100.
  • Getting a formatted report. AI uses bold claim names and headings, which you then have to clean up before pasting into the edit doc. Try: Add: "Plain text, one numbered line per problem, no bold."

What a strong prompt includes

  • Names the fact sheet as the source
  • Asks for only unsupported claims
  • Says what each line contains
  • Says not to rewrite
  • Sets a word limit

Practice this now ↑

Related exercises