Most messy AI answers start the same way: you dump a long email, PDF excerpt, or meeting notes into ChatGPT, Claude, Grok, or Perplexity… then bury your actual ask somewhere in the middle.
The model can’t tell which lines are rules and which lines are source material. So it “helps” by rewriting your paste, inventing missing pieces, or ignoring the one constraint you cared about.
Today’s free tip (works on every major platform):
Put a hard fence between instructions and pasted text.
1. Write your ask and constraints first (short).
2. Add a clear delimiter line, like:
---SOURCE START---
3. Paste the raw material.
4. Close it:
---SOURCE END---
5. End with the output you want (bullets, table, email, etc.).
Even that alone usually stops the model from treating your source as fresh instructions.
What’s locked for Premium ($5/mo) in Classroom → AI That Pays:
• Copy-paste templates for ChatGPT, Claude, Grok, and Perplexity
• The 5-part “fence” checklist (including what to do when the paste is huge)
• Failure fixes: when the model still rewrites your source, skips citations, or invents quotes
• A 60-second before/after example you can reuse this week
Upgrade to Premium ($5/mo) to unlock the full tip in Classroom → AI That Pays.