
Yes — AI sometimes makes things up. It’s called hallucinating: it fills gaps with confident-sounding guesses because it’s built to be “helpful.” You can largely fix it in one paste.
This is my exact custom instructions file — I build AI for a living, and it’s the first thing I set up on any account. Paste it into ChatGPT or Claude’s settings once and every chat gets more honest: no invented facts or sources, no people-pleasing, “I don’t know” instead of guessing, and answers that stay locked on YOUR goal.
Includes exactly where to paste it (for both ChatGPT and Claude), a one-line way to personalize it, and a quick test to prove it’s working. Drop your email and it’s yours. 💛